spis treści - contents - eksploatacja i niezawodność
Transkrypt
spis treści - contents - eksploatacja i niezawodność
SPIS TREŚCI - CONTENTS Science and Technology Abstracts............................................................................................................................................................................................................. III Andrzej LOSKA Remarks about modelling of maintenance processes with the use of scenario techniques Uwagi o modelowaniu procesów eksploatacyjnych z wykorzystaniem technik scenariuszowych................................................................ 92 Sebastian Marian Zaharia, Ionel Martinescu, Cristin Olimpiu Morariu Life time prediction using accelerated test data of the specimens from mechanical element Prognozowanie czasu pracy elementu mechanicznego z wykorzystaniem danych z badań przyspieszonych............................................. 99 Hubert Dębski, Grzegorz Koszałka, Mirosław Ferdynus Application of fem in the analysis of the structure of a trailer supporting frame with variable operation parameters Wykorzystanie mes w analizie struktury nośnej ramy naczepy o zmiennych parametrach eksploatacyjnych...............................................107 Xiao-Ling ZHANG, Hong-Zhong HUANG, Zhong-Lai WANG, Ning-Cong XIAO, Yan-Feng LI Uncertainty analysis method based on a combination of the maximum entropy principle and the point estimation method Metoda analizy niepewności oparta na połączeniu zasady maksymalnej entropii i metody oceny punktowej.............................................114 Zhisheng ZHANG, Yifan ZHOU, Yong SUN, Lin MA Condition-based maintenance optimisation without a predetermined strategy structure for a two-component series system Optymalizacja zależnego od stanu technicznego utrzymania urządzeń dla dwuskładnikowego systemu szeregowego nie wymagająca z góry ustalonej struktury strategii..............................................................................................................................................120 Wenjie WU, Hong-Zhong HUANG, Zhong-Lai WANG, Yan-Feng LI, Yu PANG Reliability analysis of mechanical vibration component using fuzzy sets theory Analiza niezawodnościowa mechanicznego elementu wibracyjnego z wykorzystaniem teorii zbiorów rozmytych.....................................130 Diego GALAR, Anna GUSTAFSON, Bernardo TORMOS, Luis BERGES Maintenance decision making based on different types of data fusion Podejmowanie decyzji eksploatacyjnych w oparciu o fuzję różnego typu danych...............................................................................................135 Łukasz JEDLIŃSKI Multi-channel registered data denoising using wavelet transform Odszumianie danych rejestrowanych wielokanałowoz użyciem transformaty falkowej....................................................................................145 Andrzej TOMPOROWSKI Stream of efficiency of rice grains multi-disc grinding Strumień wydajności wielotarczowego rozdrabniania ziaren ryżu........................................................................................................................150 Leonas Povilas LINGAITIS, Sergey MJAMLIN, Denis BARANOVSKY, Virgilijus JASTREMSKAS Prediction methodology of durability of locomotives diesel engines Metodologia prognozowania trwałości silników diesla w lokomotywach............................................................................................................154 Zdzisław Chłopek Testing of hazards to the environment caused by particulate matter during use of vehicles Badania zagrożenia środowiska cząstkami stałymi podczas eksploatacji pojazdów samochodowych............................................................160 Monika HARDYGÓRA, Henryk KOMANDER, Ryszard BŁAŻEJ, Leszek JURDZIAK Method of predicting the fatigue strength in multiplies splices of belt conveyors Metoda prognozowania trwałości zmęczeniowej złączy wieloprzekładkowych taśm przenośnikowych.........................................................171 Krzysztof PARCZEWSKI, Henryk WNĘK Make use of the friction coefficient during braking the vehicle Wykorzystanie przyczepności podczas hamowania pojazdu...................................................................................................................................176 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 I SPIS TREŚCI - CONTENTS Daniel PIENIAK, Agata M. NIEWCZAS, Paweł KORDOS Influence of thermal fatigue and ageing on the microhardness of polymer-ceramic composites for bio-medical applications Wpływ zmęczenia cieplnego oraz starzenia na mikrotwardość kompozytów polimerowo – ceramicznych do zastosowań biomedycznych.....................................................................................................................................................................................181 2012 International Conference on Quality, Reliability, Risk, Maintenance, and Safety Engineering, Chengdu, Sichuan, China.................189 II Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 W SKRÓCIE - ABSTRACTS Loska A. Remarks about modelling of maintenance processes with the use of scenario techniques. Eksploatacja i Niezawodnosc - Maintenance and Reliability 2012; 14 (2): 92-98. Loska A. Uwagi o modelowaniu procesów eksploatacyjnych z wykorzystaniem technik scenariuszowych. Eksploatacja i Niezawodnosc - Maintenance and Reliability 2012; 14 (2): 92-98. This article aims at presenting the results of research use of modelling of exploitation processes in practical applications (industrial). Firstly, there has been made an identification and of solutions in this area, particular attention was paid equally to the problems associated with processes modelling in the context of wider organizational and technical approach. Basing on the performed studies, of both basic (theoretical) and industrial character, the research problem has been defined. It concentrates on the difficulties of using exploitation process models in relation to the maintenance organization in an industrial enterprise, because of their "flat" nature. Based on the developed research problem, there has been proposed a way of its solution by applying scenario techniques. Secondly, following a review of the state of knowledge (literature) in the context of exploitation scenario building, there has been prepared and presented a detailed research procedure, that will result in the development of methodology for creation of exploitation scenarios, including the way of their use in practice. Celem artykułu jest zaprezentowanie wyników przeprowadzonych badań w zakresie sposobów wykorzystania metod modelowania procesów eksploatacyjnych w zastosowaniach praktycznych (przemysłowych). W pierwszej kolejności dokonano identyfikacji i klasyfikacji stosowanych w tym zakresie rozwiązań, zwrócono przy tym uwagę na problemy związane z modelowaniem procesów w aspekcie szerszego ujęcia organizacyjno-technicznego. W oparciu o przeprowadzone badania, które miały charakter zarówno podstawowy (teoretyczny), jak i przemysłowy, sformułowano problem badawczy. Jego istota sprowadza się do trudności wykorzystania modeli procesów eksploatacji w odniesieniu do funkcjonujących organizacji utrzymania ruchu przedsiębiorstw przemysłowych ze względu na ich „płaski” charakter. W oparciu o sformułowany problem badawczy zaproponowano sposób jego rozwiązania poprzez zastosowanie do modelowania procesów eksploatacji technik scenariuszowych. Następnie po dokonaniu przeglądu stanu wiedzy (literatury) w aspekcie budowy scenariuszy eksploatacyjnych, opracowano i przedstawiono szczegółowo procedurę prowadzenia badań, których efektem będzie opracowanie metodologii tworzenia scenariuszy eksploatacyjnych, z uwzględnieniem sposobu praktycznego ich wykorzystania. ZAHARIA S M, MARTINESCU I, MORARIU C O. Life time prediction using accelerated test data of the specimens from mechanical element. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 99–106. ZAHARIA S M, MARTINESCU I, MORARIU C O. Prognozowanie czasu pracy elementu mechanicznego z wykorzystaniem danych z badań przyspieszonych. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 99–106. In accelerated life testing, products are exposed to stress levels higher than those at normal use in order to obtain information in a short time. In this paper ,we expand the limits of performing accelerated life tests to an aerospace product (supple platinum). Specimens from the IAR 330 Puma helicopter structure, made of supple platinum, were subjected to accelerated life testing, and a significant reduction of the testing time was obtained. A simulation of the accelerated life testing data from the same case study was performed using the Monte Carlo method, with the purpose of comparing the data resulting from the experimental study (accelerated life tests) with the simulated data. W przyspieszonych badaniach cyklu życia, produkty poddaje się naprężeniom o wartościach wyższych niż występujące w normalnych warunkach użytkowania, w celu szybszego uzyskania informacji. W przedstawionym artykule poszerzamy granice zastosowań badań przyspieszonych o produkt przemysłu lotniczego (giętka platyna, ang. supple platinum). Próbki z giętkiej platyny pochodzące z elementu konstrukcji śmigłowca IAR 330 Puma poddano badaniom przyspieszonym otrzymując znaczne skrócenie czasu badania. Dla otrzymanych danych z badań przyspieszonych przeprowadzono symulację metodą Monte Carlo, w celu porównania danych eksperymentalnych (z badań przyspieszonych) z danymi symulacyjnymi. Dębski H, Koszałka G, Ferdynus M. Application of fem in the analysis of the structure of a trailer supporting frame with variable operation parameters. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 107–114. Dębski H, Koszałka G, Ferdynus M. Wykorzystanie mes w analizie struktury nośnej ramy naczepy o zmiennych parametrach eksploatacyjnych. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 107–114. This work presents a numerical analysis of the structure of a trailer frame of adjustable length and an increased load capacity designed for the transportation of oversize loads. The study was aimed at developing adequate numerical FEM models which would allow identification of the effort of the structure and the state of strain under operating loads. The analysis was carried out using ABAQUS/Standard, a numerical tool which enables calculations in the geometrically non-linear range with the use of the incremental–iterative Newton– Raphson method. As a result of the analysis, trouble areas in the frame were found in which dangerous stresses occurred. This enabled modification of the structure, leading to a reduction of the stresses to a safe level. W pracy zaprezentowano numeryczną analizę konstrukcji ramy naczepy o zmiennej długości i podwyższonej nośności przeznaczonej do transportu ładunków ponadnormatywnych. Prowadzone badania miały na celu opracowanie adekwatnych modeli numerycznych MES umożliwiających identyfikację wytężenia konstrukcji oraz stanu odkształcenia w warunkach obciążeń eksploatacyjnych. Zastosowanym do analizy narzędziem numerycznym był program Abaqus/Standard, umożliwiający prowadzenie obliczeń w zakresie geometrycznie nieliniowym z wykorzystaniem przyrostowo-iteracyjnej metody Newtona-Raphsona. W wyniku przeprowadzonych prac ustalono newralgiczne obszary ramy w których występowały niebezpieczne naprężenia. Umożliwiło to modyfikację konstrukcji pozwalającą na zmniejszenie naprężeń do bezpiecznego poziomu. ZHANG XL, HUANG HZ, WANG ZL, XIAO NC, LI YF. Uncertainty analysis method based on the combination of maximum entropy principle and point estimation method. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 115–120. ZHANG XL, HUANG HZ, WANG ZL, XIAO NC, LI YF. Metoda analizy niepewności oparta na połączeniu zasady maksymalnej entropii i metody oceny punktowej. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 115–120. Uncertainty is inevitable in product design processes. Therefore, to make reliable decisions, uncertainty analysis incorporating all kinds of uncertainty is needed. In engineering practice, due to the incomplete knowledge, the distribution of some design variables can not be determined. Furthermore, the performance function is highly nonlinear, therefore, the high order moments of the performance function are needed to calculate the probability of failure accurately. In this paper, an uncertainty analysis method combining the maximum entropy principle and the bootstrapping method is proposed. Firstly, the bootstrapping method is used to calculate the confidence intervals of the first four moments for mixed random variables and sample variables. Secondly, the high order moments of limit state functions are estimated using the reduced dimension method. Thirdly, to calculate the probability density function (PDF) and cumulative distribution function (CDF) of the limit state functions, an optimization model based on the maximum entropy principle is formulated. In the proposed method, the assumptions that the distribution of the random variables are known Niepewność jest nieodłącznym elementem procesów projektowania produktu. Dlatego też podejmowanie niezawodnych decyzji wymaga analizy niepewności, która uwzględniałaby wszystkie rodzaje niepewności. W praktyce inżynierskiej, z powodu niepełnej wiedzy, wyznaczenie rozkładu niektórych zmiennych projektowych nie jest możliwe. Co więcej, funkcja stanu granicznego jest wysoce nieliniowa, co sprawia, że do poprawnego obliczenia prawdopodobieństwa uszkodzenia potrzebna jest znajomość momentów wyższych rzędów tej funkcji. W niniejszej pracy zaproponowano metodę analizy niepewności łączącą zasadę maksymalnej entropii z metodą bootstrapową. W pierwszej części pracy wykorzystano metodę bootstrapową do obliczenia przedziałów ufności czterech pierwszych momentów dla zmiennych losowych typu mieszanego oraz zmiennych z próby. Następnie, wyznaczono momenty wyższych rzędów funkcji stanu granicznego przy użyciu metody redukcji wymiarów. Po trzecie, w celu obliczenia funkcji gęstości prawdopodobieństwa (PDF) oraz dystrybuanty (CDF) funkcji stanu granicznego, sformułowano model optymalizacji oparty na Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 III W SKRÓCIE - ABSTRACTS and the calculation of the sensitivity for limit state function with respect to the Most Probable Point (MPP) are avoided. Finally, comparisons of results from the proposed methods and the MCS method are presented and discussed with numerical examples. zasadzie maksymalnej entropii. Proponowana metoda nie wymaga założenia znajomości rozkładów zmiennych losowych ani obliczania wrażliwości dla funkcji stanu granicznego w odniesieniu do najbardziej prawdopodobnego punktu awarii. W końcowej części artykułu porównano na podstawie przykładów numerycznych wyniki otrzymane za pomocą proponowanej metody oraz symulacji Monte Carlo (MCS). ZHANG Z, ZHOU Y, SUN Y, MA L. Condition-based maintenance optimisation without a predetermined strategy structure for a twocomponent series system. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 121–130. ZHANG Z, ZHOU Y, SUN Y, MA L. Optymalizacja zależnego od stanu technicznego utrzymania urządzeń dla dwuskładnikowego systemu szeregowego nie wymagająca z góry ustalonej struktury strategii. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 121–130. Most existing research on maintenance optimisation for multi-component systems only considers the lifetime distribution of the components. When the condition-based maintenance (CBM) strategy is adopted for multi-component systems, the strategy structure becomes complex due to the large number of component states and their combinations. Consequently, some predetermined maintenance strategy structures are often assumed before the maintenance optimisation of a multi-component system in a CBM context. Developing these predetermined strategy structure needs expert experience and the optimality of these strategies is often not proofed. This paper proposed a maintenance optimisation method that does not require any predetermined strategy structure for a two-component series system. The proposed method is developed based on the semi-Markov decision process (SMDP). A simulation study shows that the proposed method can identify the optimal maintenance strategy adaptively for different maintenance costs and parameters of degradation processes. The optimal maintenance strategy structure is also investigated in the simulation study, which provides reference for further research in maintenance optimisation of multi-component systems. WU W, HUANG HZ, WANG ZL, LI YF, PANG Y. Reliability analysis of mechanical vibration component using fuzzy sets theory. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 131–134. The conventional reliability analysis of mechanical vibration component only considers the randomness of vibration but rarely for the fuzziness that may exist. It is therefore difficult to be consistent with the engineering practices. Based on the mechanical vibration theory, a novel fuzzy reliability approach by integrating the fuzzy comprehensive evaluation and fuzzy set theory is proposed in this paper. The fuzzy comprehensive evaluation is used to optimize the fuzzy factors of the reliability analysis of vibration component. With the aim of comparing the performance of the proposed approach with the conventional approach, two engineering examples are presented. The results demonstrate that the proposed approach is better than the conventional approach for its capability of covering fuzzy factors in the engineering problems. GALAR D, GUSTAFSON A, TORMOS B, BERGES L. Maintenance decision making based on different types of data fusion. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 135–144. Over the last decade, system integration is applied more as it allows organizations to streamline business processes. A recent development in the asset engineering management is to leverage the investment already made in process control systems. This allows the operations, maintenance, and process control teams to monitor and determine new alarm level based on the physical condition data of the critical machines. Condition-based maintenance (CBM) is a maintenance philosophy based on this massive data collection, wherein equipment repair or replacement decisions depend on the current and projected future health of the equipment. Since, past research has been dominated by condition monitoring techniques for specific applications; the maintenance community lacks a generic CBM implementation method based on data mining of such vast amount of collected data. The methodology would be relevant across different domains. It is necessary to integrate Condition Monitoring (CM) data with management data from CMMS (Computer Maintenance Management Systems) which contains information, such as: component failures, failure information related data, servicing or repairs, and inventory control and so on. These systems are the core of traditional scheduled maintenance practices and rely on bulk observations from historical data to make modifications to regulated maintenance actions. The most obvious obstacle in the integration of CMMS, process and CM data is the disparate nature of the data types involved, and there have benn several attempts to remedy this problem. Although, there have been many recent efforts to collect and maintain large repositories of these types of data, there have been IV Większość badań nad optymalizacją utrzymania systemów wieloskładnikowych bierze pod uwagę jedynie rozkład czasu życia elementów składowych. Kiedy przyjmie się dla systemów wieloskładnikowych strategię utrzymania urządzeń zależną od ich bieżącego stanu technicznego (condition-based maintenance, CBM), struktura strategii staje się złożona w związku z dużą liczbą stanów składowych oraz ich kombinacji. W konsekwencji, często przyjmuje się pewne z góry ustalone struktury strategii utrzymania przed optymalizacją utrzymania systemu wieloskładnikowego w kontekście CBM. Opracowanie takich z góry ustalonych struktur strategii wymaga jednak specjalistycznego doświadczenia, a i tak brak dowodów na optymalność tych strategii. W artykule zaproponowano metodę optymalizacji utrzymania szeregowego systemu dwuskładnikowego, która nie wymaga wcześniej ustalonej struktury strategii. Proponowaną metodę opracowano na podstawie semimarkowskiego procesu decyzyjnego (SMDP). Badanie symulacyjne pokazało, że za pomocą proponowanej metody można ustalać optymalną strategię utrzymania w sposób adaptacyjny dla różnych kosztów utrzymania oraz parametrów procesów degradacyjnych. Za pomocą symulacji badano także optymalną strukturę strategii utrzymania, jako punkt odniesienia dla przyszłych studiów nad optymalizacją systemów wieloskładnikowych. WU W, HUANG HZ, WANG ZL, LI YF, PANG Y. Analiza niezawodnościowa mechanicznego elementu wibracyjnego z wykorzystaniem teorii zbiorów rozmytych. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 131–134. Tradycyjna analiza niezawodnościowa wibracyjnego elementu mechanicznego bierze pod uwagę jedynie losowość drgań, rzadko zaś wyjaśnia mogącą występować rozmytość. Taka analiza nie odpowiada zatem praktyce inżynierskiej. Opierając się na teorii drgań mechanicznych, w niniejszym artykule zaproponowano nowatorskie podejście w ramach teorii rozmytej niezawodności, które łączy rozmytą ocenę kompleksową oraz teorię zbiorów rozmytych. Rozmytej oceny kompleksowej użyto do optymalizacji rozmytych czynników analizy niezawodnościowej elementu wibracyjnego. W celu porównania efektywności proponowanego podejścia z efektywnością podejścia tradycyjnego przedstawiono dwa przykłady z dziedziny inżynierii. Wyniki pokazują, że proponowane podejście jest lepsze od tradycyjnego ze względu na możliwość objęcia w problemach inżynierskich czynników rozmytych . GALAR D, GUSTAFSON A, TORMOS B, BERGES L. Podejmowanie decyzji eksploatacyjnych w oparciu o fuzję różnego typu danych. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 135–144. W ostatniej dekadzie coraz częściej stosuje się integrację systemów, która pozwala przedsiębiorstwom zwiększać wydajność procesów biznesowych. Nowością w zarządzaniu infrastrukturą techniczną jest zwiększanie efektywności już poczynionych inwestycji w systemy kontroli procesów. Pozwala to zespołom do spraw operacyjnych, utrzymania ruchu oraz kontroli procesów monitorować i ustalać nowe poziomy alarmowe na podstawie danych o stanie fizycznym maszyn krytycznych. Utrzymanie urządzeń zależne od ich bieżącego stanu technicznego (condition-based maintenance, CBM) to filozofia utrzymania ruchu opierająca się na tym masowym poborze danych, wedle której decyzje dotyczące naprawy lub wymiany sprzętu zależą od jego obecnego oraz przewidywanego przyszłego stanu technicznego. Ponieważ dotychczasowe badania były zdominowane przez problem technik monitorowania stanu dla konkretnych aplikacji, nie opracowano ogólnej metody wdrażania CBM opartej na eksploracji (data mining ) owych olbrzymich ilości zebranych danych, która miałaby zastosowanie w różnych domenach. Konieczna jest integracja danych z monitorowania stanu (condition monitoring, CM) z danymi dotyczącymi zarządzania pochodzącymi ze skomputeryzowanych systemów zarządzania utrzymaniem ruchu (CMMS), które zawierają informacje na temat uszkodzeń elementów składowych, dane związane z uszkodzeniami, a także informacje dotyczące obsługi lub napraw czy sterowania zapasami. Systemy te stanowią Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 W SKRÓCIE - ABSTRACTS relatively few studies to identify the ways these to datasets could be related. This paper attempts to fulfill that need by proposing a combined data miningbased methodology for CBM considering CM data and Historical Maintenance Management data. It shows a system integration of physical and management data that also supports business intelligence and data mining where data sets can be combined in non-traditional ways. podstawę tradycyjnych praktyk obsługi planowej, a zasadzają się na całościowych obserwacjach dokonywanych na podstawie danych eksploatacyjnych, które pozwalają modyfikować regulowane działania obsługowe. Najbardziej oczywistą przeszkodą w integracji danych CMMS, danych procesowych oraz danych z monitorowania stanu jest rozbieżność ich natury. Dotychczas podjęto jedynie kilka prób rozwiązania tego problemu. Chociaż ostatnio wiele wysiłku włożono w gromadzenie i utrzymanie dużych zasobów tego typu danych, istnieje stosunkowo niewiele badań na temat możliwych sposobów powiązania owych zestawów danych. W prezentowanej pracy poczyniono próbę wypełnienia tej luki proponując metodologię łączoną opartą na eksploracji danych dla celów CBM, która bierze pod uwagę dane z monitorowania stanu i eksploatacyjne dane z zarządzania ruchem. W pracy przedstawiono integrację systemową danych fizycznych i danych z zarządzania, która wspiera także analitykę biznesową (business intelligence) oraz eksplorację danych, gdzie zestawy danych można łączyć w sposób nietradycyjny. JEDLIŃSKI Ł. Multi-channel registered data denoising using wavelet transform. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 145–149. JEDLIŃSKI Ł. Odszumianie danych rejestrowanych wielokanałowo z użyciem transformaty falkowej. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 145–149. In order to obtain information regarding given phenomenon or object, it is usually necessary to register selected measurement signals obtained using sensors. Unfortunately, obtained signals, apart form desired information, contain disturbances caused by, amongst many other, properties of the measurement channel and processes associated with object operation. In many cases it is necessary to measure the same value in different places and/or directions. Thus, there is a demand for a tool improving signal to noise ration of the multi-channel registered signals.Wavelet transform is a relatively new method of data processing used in different fields (e.g. technique and physics). In case of signals it can be used for denoising, compression, trend detection or discontinuity detection. In this work it was used to denoise vibration signals registered by two three-axis sensors. Object of investigation was the bevel toothed gear. Signals denoising was to improve efficiency of the diagnosis of transmission gears teeth damage. W celu uzyskania informacji o interesującym nas zjawisku lub obiekcie najczęściej rejestrowane są wybrane sygnały pomiarowe otrzymane za pośrednictwem czujników. Niestety uzyskane sygnały oprócz pożądanej informacji zawierają również zakłócenia, które są spowodowane m.in. właściwościami toru pomiarowego i procesami towarzyszącymi działaniu obiektu. W wielu przypadkach zachodzi potrzeba pomiaru takiej samej wielkości w różnych miejscach obiektu i/lub kierunkach. Potrzebne są zatem narzędzia do poprawy stosunku sygnału do szumu sygnałów rejestrowanych wielokanałowo.Transformata falkowa jest stosunkowo nową metodą przetwarzania danych, która znalazła zastosowanie w różnych dziedzinach takich jak technika i fizyka. W odniesieniu do sygnałów może być używana do odszumiania, kompresji, wykrywaniu trendu czy nieciągłości sygnału. W pracy tej transformata falkowa została użyta od odszumiania sygnałów drgań zarejestrowanych z dwóch trójosiowych czujników. Obiektem badań była przekładnia zębata stożkowa. Odszumianie sygnałów miało na celu poprawę skuteczności diagnozy uszkodzenia kół zębatych przekładni. TOMPOROWSKI A. Stream of efficiency of rice grains multi-disc grinding. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 150–153. TOMPOROWSKI A. Strumień wydajności wielotarczowego rozdrabniania ziaren ryżu. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 150–153. A search for design solutions for grain grinding devices offering energy-saving production processes justify the research into the improvement of the theory and the design of grinders. The efficiency, functionality and performance of the rice grains grinding are significantly influenced by processes, difficult to describe, occurring within the working space of the grinder. Although there are some studies on the principles of the functionality and analysis of multi-disc grinders, an attempt has not been made so far at describing the influence of features and behaviour of ground material on efficiency, performance and usefulness of the biomaterial comminution/grinding process. The basis for the improvement of the functionality of a grinding device is an analysis of the potential of existing solutions and a determination of the scope of effective design features of the working unit within the permissible area. The criteria may be fulfilled by objective-oriented control of the design features of the multi-disc unit. In order for these actions to be effective, relevant relationships need to be explored and a mathematical description needs to be developed for the flow of ground grains through the working space of a multi-disc grinder, as a result variable of the structure and the operation of a working unit. Poszukiwania rozwiązań konstrukcyjnych zespołów rozdrabniających ziarna zbóż, prowadzące do energooszczędnych procesów produkcyjnych uzasadniają podjęcie badań nad doskonaleniem teorii i konstrukcji rozdrabniaczy. Istotny wpływ na wydajność, funkcjonalność i sprawność procesu rozdrabniania ziarna ryżu mają trudne do opisania zjawiska zachodzące w przestrzeni roboczej rozdrabniacza. Pomimo, że dostępne są opracowania na temat podstaw funkcjonalności i badań rozdrabniaczy wielotarczowych, jak dotychczas nie podejmowano próby opisu wpływu cech i zachowań rozdrabnianego materiału na wydajność, sprawność i użyteczność procesu rozdrabniania biomateriałów. Podstawą do poprawy funkcjonalności działania maszyny rozdrabniającej jest przeprowadzone rozpoznanie możliwości istniejących rozwiązań oraz określenie zakresu, skutecznych w przetwórstwie, cech konstrukcyjnych zespołu roboczego z obszaru dopuszczalnego. Spełnienie kryteriów może być osiągnięte między innymi na drodze celowego sterowania cechami konstrukcyjnymi zespołu wielotarczowego. Aby jednak działania te przyniosły planowane korzyści, konieczne staje się poznanie zależności oraz opracowanie opisu matematycznego przepływu rozdrabnianego ziarna przez przestrzeń roboczą rozdrabniacza wielotarczowego, jako zmiennej wynikowej konstrukcji i działania zespołu roboczego. LINGAITIS L P, MJAMLIN S, BARANOVSKY D, JASTREMSKAS V. Prediction methodology of durability of locomotives diesel engines. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 154–159. LINGAITIS L P, MJAMLIN S, BARANOVSKY D, JASTREMSKAS V. Metodologia prognozowania trwałości silników diesla w lokomotywach. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 154–159. The article testifies that technical maintenance and repair terms as well as durability can be predicted accordingly to the maintenance parameters of the diesel locomotives. It is determined that fixing fuel consumption and capacity of diesel locomotives and taking in comparison with limit values allows to set a performance date for technical maintenance. Introduced suggested aspects of interrepair resource of the diesels dependent on comparable fuel consumption and evaluating their operating probability without failure for the durability prediction of diesel locomotives. Currently, the most common are three strategies: until failure, scheduled – premonitory and adaptive (diagnostic). When the quantity of necessary technical maintenances is known for the specific kind of diesel locomotives it is possible to determine interrepair resource depending W artykule wykazano, że częstotliwość przeglądów technicznych į remontów spalinowozów z silnikami diesla można prognozować analizując parametry eksploatacyjne. Obecnie najbardziej rozpowszechnione są trzy strategie: do awarii, planowo-wyprzedzająca i adaptacyjna (diagnostyczna). W konkretnych warunkach zarządzania gospodarczego na pierwsze miejsce wysuwa się specyfika użytkowania spalinowozów, o zaraz po niej – normatywne wymagania techniczne, reglamentujące eksploatację spalinowozów. Ustalono, że poprzez odnotowywanie zużycia paliwa oraz mocy spalinowozów można określić czas eksploatacji, po upływie którego konieczne będzie przeprowadzenie przeglądu technicznego. Przedłożona została teoretyczna zależność okresu międzyremontowego diesli od porównawczego zużycia paliwa, wykorzystywana do oceny Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 V W SKRÓCIE - ABSTRACTS on comparative and required per hour consumption of fuel and predict their durability taking into account probability of operating without failures. prawdopodobieństwa bezawaryjnej pracy oraz zaproponowana metodyka prognozowania trwałości spalinowozów z silnikami diesla. Znając właściwą dla danej marki spalinowozu liczbę przeglądów technicznych, można oszacować okres międzyremontowy w zależności od godzinnego i porównawczego zużycia paliwa oraz uwzględniając ich prawdopodobieństwo bezawaryjnej pracy, w ten sposób prognozując ich trwałość. Chłopek Z. Testing of hazards to the environment caused by particulate matter during use of vehicles. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 160–170 Chłopek Z. Badania zagrożenia środowiska cząstkami stałymi podczas eksploatacji pojazdów samochodowych. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 160–170. The study presents results of tests on emissions of fractions of PM10, PM2.5 and PM1 dusts. For modeling of emissions of fractions of PM2.5 and PM1 particles, results of empirical tests were used as carried out in air quality supervision stations located in the agglomeration of the city of Brno. The results of modeling of emissions of fractions of PM2.5 and PM1 particles did not make it possible to make unequivocal conclusions, which proves that the discussed problem has to be treated statistically. However, a significant relation between models of emissions of fractions of particulate matter and sources of emissions of dusts and conditions for distribution of the same were observed. W pracy przedstawiono wyniki badań imisji frakcji pyłów PM10, PM2.5 i PM1. Do modelowania imisji frakcji cząstek stałych PM2.5 i PM1 wykorzystano wyniki badań empirycznych, przeprowadzonych na stacjach nadzorowania jakości powietrza w aglomeracji czeskiego miasta Brna. Wyniki modelowania imisji frakcji cząstek stałych PM2.5 i PM1 nie umożliwiły sformułowania jednoznacznych wniosków, co dowodzi konieczności statystycznego potraktowania badanego problemu. Stwierdzono jednak istotną zależność modeli imisji frakcji cząstek stałych od źródeł emisji pyłów i warunków ich rozprzestrzenia. HARDYGÓRA M, KOMANDER H, BŁAŻEJ R, JURDZIAK L. Method of predicting the fatigue strength in multiplies splices of belt conveyors. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 171–175. HARDYGÓRA M, KOMANDER H, BŁAŻEJ R, JURDZIAK L. Metoda prognozowania trwałości zmęczeniowej złączy wieloprzekładkowych taśm przenośnikowych. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 171–175. Method of testing the stress distribution in overlap adhesive joints of multiply conveyor belts is presented The impact of strength properties of belts and adhesive rubber on the stress scale and fatigue life of splices were defined. The relations enabling to calculate the strength magnitude and predicting the durability of conveyor’s belts splices were given. Przedstawiono metodę badań rozkładu naprężeń w spoinie klejowej zakładkowych złączy taśm przenośnikowych wieloprzekładkowych. Określono wpływ właściwości wytrzymałościowych taśm i gumy klejowej na wielkość naprężeń i na trwałość zmęczeniową złączy. Podano zależności pozwalające na obliczenie wielkości naprężeń i prognozowanie trwałości złączy taśm przenośnikowych. PARCZEWSKI K, WNĘK H. Make use of the friction coefficient during braking the vehicle. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 176–180. PARCZEWSKI K, WNĘK H. Wykorzystanie przyczepności podczas hamowania pojazdu. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 176–180. In this publication is presented use the tyre-road friction during vehicle braking. Results presented in this publication are based on the road tests of the vehicle equipped in the anti-lock brake system (ABS). Two kinds of tests applied were carried out - the road tests of vehicle making the manoeuvre of braking on the straight section of the road and on the curve of the road. The braking forces and the friction coefficients for the individual wheels of the vehicle were defined on the basis of road tests, including the border values of the friction coefficient. W publikacji przedstawiono zagadnienie wykorzystania przyczepności opony do nawierzchni jezdni podczas hamowania. Wyniki prezentowane w publikacji oparto na badaniach drogowych pojazdu osobowego wyposażonego w układ zapobiegający blokowaniu kół. Przeprowadzono dwa rodzaje prób stosowanych do badania wykorzystania przyczepności - badania pojazdu wykonującego manewr hamowania na prostoliniowym odcinku drogi oraz na łuku drogi. Na podstawie badań określono siły hamowania oraz współczynniki przyczepności dla poszczególnych kół pojazdu, w tym wartości graniczne współczynnika przyczepności. PIENIAK D, NIEWCZAS A, KORDOS P. Influence of thermal fatigue and ageing on the microhardness of polymer-ceramic composites for bio-medical applications. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 181–188. PIENIAK D, NIEWCZAS A, KORDOS P. Wpływ zmęczenia cieplnego oraz starzenia na mikrotwardość kompozytów polimerowo – ceramicznych do zastosowań biomedycznych. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 181–188. Studies presented in this paper, concern polymer-ceramic composites applied in the conservative dentistry. The aim of the study was to evaluate a long-term impact of the humid environment and cyclic thermal loads on the microhardness of new silorane-based composites and two methacrylate-based composites. The composite samples were subjected to normal saline environment with cyclically variable temperatures (5°C and 65°C), using a special thermal shock simulator. Microhardness was measured with Vicker’s method before the fatigue test and after a series of 4000 thermal cycles. It is known that microhardness of siloranebased composite in opposite to methacrylate-based composites not decrease under the influence of cyclic thermal loads. It was found slight increase of microhardness under conditions of conducted tests. The ageing studies were also conducted consisting in microhardness evaluation of the composite samples in 6 months period. During that time the samples were kept in normal saline. The studies of hardness were carried out after each month of the exposure time. No long-term impact of normal saline environment with constant temperature on the microhardness of the studied materials has been noticed. Badania prezentowane w niniejszej publikacji dotyczyły kompozytów polimerowo-ceramicznych stosowanych w stomatologii zachowawczej. Celem pracy była porównawcza ocena wpływu długotrwałego oddziaływania wilgotnego środowiska oraz cyklicznych obciążeń cieplnych na mikrotwardość nowego kompozytu bazującego na siloranach oraz dwóch tradycyjnych kompozytów bazujących na związkach metakrylanu. Próbki z kompozytów poddano oddziaływaniu środowiska soli fizjologicznej o cyklicznie zmiennych temperaturach (5°C i 65°C) wykorzystując specjalny symulator szoków termicznych. Wykonywano pomiary mikrotwardości metodą Vickersa przed rozpoczęciem testu zmęczenia cieplnego oraz po serii 4000 cykli termicznych. Wykazano, że w przeciwieństwie do tradycyjnych kompozytów stomatologicznych mikrotwardość kompozytu bazującego na siloranach nie zmniejsza się pod wpływem cyklicznego oddziaływania szoków termicznych odpowiadających warunkom fizjologicznym jamy ustnej. W warunkach przeprowadzonych badań stwierdzono nieznaczny wzrost tej mikrotwardości. Przeprowadzono również badania starzeniowe polegające na ocenie mikrotwardości próbek kompozytów przez okres 6 miesięcy. W tym okresie czasu próbki przechowywano w soli fizjologicznej. Pomiary mikrotwardości wykonywano po każdym miesiącu ekspozycji. Wykazano, że długotrwałe oddziaływanie środowiska soli fizjologicznej w warunkach stałej temperatury nie zmienia mikrotwardości żadnego z badanych materiałów. VI Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Article citation info: LOSKA A. Remarks about modelling of maintenance processes with the use of scenario techniques. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 92–98. Andrzej LOSKA Remarks about modelling of maintenance processes with the use of scenario techniques Uwagi o modelowaniu procesów eksploatacyjnych z wykorzystaniem technik scenariuszowych* This article aims at presenting the results of research use of modelling of exploitation processes in practical applications (industrial). Firstly, there has been made an identification and of solutions in this area, particular attention was paid equally to the problems associated with processes modelling in the context of wider organizational and technical approach. Basing on the performed studies, of both basic (theoretical) and industrial character, the research problem has been defined. It concentrates on the difficulties of using exploitation process models in relation to the maintenance organization in an industrial enterprise, because of their "flat" nature. Based on the developed research problem, there has been proposed a way of its solution by applying scenario techniques. Secondly, following a review of the state of knowledge (literature) in the context of exploitation scenario building, there has been prepared and presented a detailed research procedure, that will result in the development of methodology for creation of exploitation scenarios, including the way of their use in practice. The article is funded by National Science Centre in Poland under the project no. 5636/B/T02/2011/40 titled: “The use of scenario methods in exploitation processes modelling”. Keywords: exploitation, maintenance, maintenance scenarios, exploitation processes, modelling. Celem artykułu jest zaprezentowanie wyników przeprowadzonych badań w zakresie sposobów wykorzystania metod modelowania procesów eksploatacyjnych w zastosowaniach praktycznych (przemysłowych). W pierwszej kolejności dokonano identyfikacji i klasyfikacji stosowanych w tym zakresie rozwiązań, zwrócono przy tym uwagę na problemy związane z modelowaniem procesów w aspekcie szerszego ujęcia organizacyjno-technicznego. W oparciu o przeprowadzone badania, które miały charakter zarówno podstawowy (teoretyczny), jak i przemysłowy, sformułowano problem badawczy. Jego istota sprowadza się do trudności wykorzystania modeli procesów eksploatacji w odniesieniu do funkcjonujących organizacji utrzymania ruchu przedsiębiorstw przemysłowych ze względu na ich „płaski” charakter. W oparciu o sformułowany problem badawczy zaproponowano sposób jego rozwiązania poprzez zastosowanie do modelowania procesów eksploatacji technik scenariuszowych. Następnie po dokonaniu przeglądu stanu wiedzy (literatury) w aspekcie budowy scenariuszy eksploatacyjnych, opracowano i przedstawiono szczegółowo procedurę prowadzenia badań, których efektem będzie opracowanie metodologii tworzenia scenariuszy eksploatacyjnych, z uwzględnieniem sposobu praktycznego ich wykorzystania. Artykuł jest wynikiem realizacji części badań w ramach projektu badawczego, finansowanego ze środków Narodowego Centrum Nauki nr 5636/B/T02/2011/40 pt.: Wykorzystanie metod scenariuszowych w modelowaniu procesów eksploatacyjnych. Słowa kluczowe: eksploatacja, utrzymanie ruchu, scenariusze eksploatacyjne, procesy eksploatacyjne, modelowanie. 1. Introduction Both in the theory of exploitation of technical systems, as well as in its practical application, much attention is paid to the problem of modelling. Current tasks of technical departments of most of industrial enterprises exceed traditional framework for planning, execution and settlement of maintenance works, particularly in terms of rationalization and optimization of decision-making processes both in short and long term. In terms of rapid development of strategies and methods of maintenance management, as well as IT tools used for acquisition, collection and processing large amounts of data describing variety of technical facilities and conditions of their operation, (*) Tekst artykułu w polskiej wersji językowej dostępny w elektronicznym wydaniu kwartalnika na stronie www.ein.org.pl 92 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology it seems to be possible and necessary to carry out the exploitation research related primarily to the development of the way of making optimal strategic decisions concerning, among others, issues technically, organizationally and commercially reasonable terms and conditions for maintenance tasks, withdrawal of objects from exploitation, long-term practices within exploitation philosophies or scopes of works. This requires the construction of appropriate models of both technical objects, events, and exploitation processes carried out in an enterprise. These models should accurately reflect selected aspects of exploitation "reality" and must be based on foundations of modelling theory. This article represents the next stage of research conducted on the possibility of support of exploitation decision making process in the industrial enterprises by the author at the Institute of Production Engineering of Silesian University of Technology. The research leads to develop a method of processes modelling, both their technical aspects as well as non-technical considerations of maintenance tasks (such as organizational, economic), which will use such models in practice. 2. Practical considerations of classical methods of exploitation processes modelling It is assumed that the exploitation process is a set of structured operations, performed by technical objects overt their lifetime. Course of individual operations is determined by individual characteristics of objects and their organizational and technical environment, resulting in uniqueness of individual exploitation processes. In order to identify and define the research problems considered in the article, it is necessary to identify and classify the exploitation processes precisely. The basis for such classification may be a way of describing (modelling), which determines the realization of particular tasks related to this process and regarding decision making as well. According to such criteria of classification, there are two basic ways to describe (models) exploitation processes [7]: a) the exploitation process model as a sequence of events, by which certain tasks are undertaken and realized on the basis of information about specific events occurring at different moments of time (fig. 1), Fig. 1. The exploitation process model as a sequence of events [7]. Zij - exploitation events, Ti - time interval between events, ti - moments in which events occur b) the exploitation process model as a series of states, by which maintenance tasks are undertaken and realized basin on momentory conditions of the object identified in equal intervals (fig. 2). Considering the exploitation process model as a sequence of events, one can distinguish two types of possible events: intended and unintended ones. Each type of event determines consequently the procedure, reflecting different principles of Fig. 2. The exploitation process model as a series of states [7]. St(i) identified technical conditions, Δt - time interval between the moments of identification of the technical condition, ti - moments of identification of the technical condition maintenance of facilities, comprised the maintenance strategy. Using the process model as a series of states to identify decision making in maintenance requires periodic (constant in time) control of technical conditions of the object and taking into account different procedures. Exploitation practice shows that description of exploitation process related to specific item of technical objects requires most frequently simultaneous use of models discussed above. For example, technical objects, for which technical condition monitoring activities are carried out to (related to the exploitation process model as a sequence of states), may be subject to breakdown (events associated with exploitation process model as a sequence of events). The aforementioned way of modelling exploitation processes in industrial practice results in development and implementation of general and dedicated solutions taking forms of operational strategies. In particular, basing on the relationship between exploitation processes and the way of decision-making in organizational and technical exploitation systems, it can be stated that most common decisions result from use of one of the following maintenance strategies: 1. breakdown maintenance strategy (BM), based on the exploitation process model as a sequence of unintended events, 2. preventive maintenance strategy, based on the exploitation process model as a sequence of unintended intended events (PM), 3. predictive maintenance strategy, which is based on the exploitation process model as a series of states (PM). In industrial practice by linking different strategies with decision-making process regarding specific technical objects (and not necessarily the entire technological setting), all three strategies create maintenance policy of the company, where any of them is a dominant ona (is the basis for decision-making in relation to the majority or the most important technical objects), and the rest is supplementing it. 3. Identification of research problems and proposals for solutions Exploitation process models, presented in the previous section and described in the literature, allow to solve important practical problems concerning use and maintenance of machinery and equipment. However, the above mentioned capabilities are subject to certain constraints, resulting from the character of these methods. The main constraint of this type is number of characteristics (parameters) possible to include in the model, that describe condition of the object/diagnostic signal or event. Such restriction significantly influences the accuracy of identification of a specific condition or event. It is also difficult to comprise qualitative features and non-technical aspects (organ- Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 93 Science and Technology izational and economic), which significantly affect the implementation of various maintenance tasks. The most important constraints of the described ways characterizing the exploitation processes include: •significant simplification of the models of processes in relation to organization, using technical objects, which are the subject of modelling, •flat nature of the models showing the cause-effect relationships in a unidimensional way without taking into account the interaction with the environment in the form of additional non-technical aspects (eg. organizational or economic), •difficulties with including qualitative characteristics and non-technical aspects (organizational and economic) in this type of models, which significantly affect the implementation of particular maintenance tasks (including maintenance management area). The restrictions mentioned in the article hinder satisfaction of the needs related to effective use of typical models for the needs of decision making optimization in relation to the operating of technical objects by maintenance organizations of industrial enterprises. Proceeding from the existing state of knowledge in the area and the conditions mentioned above, there can be formulated the main research problem of this article: currently used models and methods of modelling have important restrictions which do not allow for effective use in optimization of decision-making related with operating of the technical objects and thus they are not an effective solution for evaluation and improvement of efficiency of maintenance management. Solution to the problem of optimizing the decision-making process may be the use of scenario methods, using a multithreaded picture of reality at simultaneously possibility to look ahead of current time. Scenario techniques belong to the forecasting methods that have been used in economic sciences for forecasting and strategic management. In technical sciences They have yet to be recognized. Among several methods to create scenarios, in this case, the best seems to be the method proposed by H. Kahn, who called it as scenariowriting. This method consists of description of events and indication their logical and coherent consequences in order to determine the way of development of an object or situation. The exemplary point of reference can be past or the current technical condition. The main emphasis is put on these events and situations, that could form the basis of future condition or sequence of events. In this way one gets a set of possible events or, in other words, images of the future. Beside many ways of creating scenarios with the use of various methods, one can distinguish a limited number of types or variants in the described methodology. In this particular methodology, there are several types or variants. For considered area, possible, appropriate and reasonable to use are two of such types, which correspond to the needs and possibilities of exploitation process modelling [23]. Basing on specific hypotheses one determines possible events being an effect of current state or situation. This is called research scenario. This type of scenario corresponds to maintenance planning problems which are directly related to defining the maintenance strategy. In spe- 94 cific situations it is possible to use anticipatory scenario (fig. 3b), in which identified relations allow to specify the effects of the situation. This type of scenario provides the basis of analysis of past events, in particular failure analysis, where the most important element is to determine the reasons of the situation or failures in maintenance work. a) b) Fig. 3. Types of scenarios: a) research scenario, b) anticipatory scenario [23] It can be noticed that, the method of construction of scenarios has much in common with the forecasting techniques. Starting from the current situation (a set of features describing exploitation processes), one can determine the factors influencing the future situations that are known. Relating these factors to the nearest future causes a slight change, while in longer periods, these changes may be significant. This is reflected both in changes in particular factors as well as in the influence of changes to individual objects or systems. Through the use of scenario methods for modelling of exploitation processes, one can show the way and scope of achieving possible alternative scenarios describing the effects of taken decisions. As a part of further research it is proposed to use the events and processes scenarios, which provide qualitatively new approach in the area of modelling exploitation processes for the needs of effective decision making/maintenance management. Key elements of the scenario methodology and features of exploitation area determine the necessary set of activities (detailed objectives), which implementation allows to solve the determined research problem. These activities include: • defining the need and basis for scenario/collection of possible scenarios generation based on specific exploitation models, which may result from the reliability criteria, in this case (eg. a set of quantitative exploitation indicators), • determining the internal and formal structure of the description of the scenario (identifying a set of parameters, the quantitative elements and features, defined as qualitative components of the situation/event), • filling in scenarios for the object as such by mapping its environment (that is, by analogy, the author of the article proposes to describe the "scenery" in which scenario is “going on” - in addition to the same scenario), • solution to the problem of practical use of scenarios in maintenance works, and optimization of decision-making processes relating to operating technical systems, taking into account multivariant issue of possible events Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology and behavioral simulation of objects in shorter and longer term. 4. The importance of exploitation process modelling with the use of scenario techniques The problem of use of scenario techniques in exploitation occurs incidentally in the literature. It is mainly related to individual descriptions of case studies of scenarios in selected events (eg. in [24]). Detailed analysis of the literature, both domestic and foreign has shown that studies on the use of scenarios for the complex strategic aspects of the exploitation of technical objects are conducted primarily at the Institute of Production Engineering of the Silesian University of Technology. This is reflected in publications [8, 13, 16, 15, 18]. In terms of the issues discussed in the article, there are separate publications describing various elements associated with this topic, in particular: 1. modelling of selected elements of exploitation, 2. scenario methods and techniques, 3. computer aided maintenance management with strategic approach. The first area is described quite extensively in terms of modelling the various elements of exploitation processes. Valuable in this area is [13], where the classification and review of the major exploitation models was carried out. Other items the literature describe a more or less detailed the various groups of exploitation models (event models eg. [7, 27], diagnostic models eg. [1, 2, 27, 28], technical objects and assets models [13, 19, 7], or exploitation process models [7, 13, 27, 22, 26]. The second area is widely described in the literature mainly from the viewpoints of economics, particularly in terms of strategic management [3, 4, 5, 6, 23]. The third area is described mainly in trade magazines, particularly in [17]. The results of the review of existing knowledge can be summarized in two aspects: 1. From the perspective of national and international publications and other studies, there are references to methods and issues which are parts of the proposed solution. 2. In complex terms, scenario techniques used for exploitation processes modelling can be considered a new issue not published both domestically and abroad. 5. The concept of research on the development of methods for the use of scenarios techniques in exploitation processes modelling The research initiated and carried out by the author to develop the use of scenarios in exploitation processes modelling includes a wide range of work, both fundamental (theoretical) as well as industrial and developmental. On the one hand, this research is aimed at developing a methodology for construction of exploitation scenarios, as the basis and method of modelling and optimization of exploitation processes, and, on the other hand, at preparation of practical tools to use developed methodology in industrial environments. Diagram of concept of carrying out such research is shown in fig. 4. Its individual components are described in the following paragraphs. Fig. 4. Concept diagram of research 5.1. Stage 1: Identification and acquisition of information and knowledge for the needs of modelling of objects, events and exploitation processes The first step of research is to acquire and collect an appropriate repository of information and knowledge about objects and all the technical and nontechnical "circumstances" of their operation, which form potential subject and scope of exploitation scenarios. The quality of the scenario, and thus the accuracy of forecasts results from the possibly large diversity and level of detail of information, therefore searching and acquiring should be orderly. The basis of such arrangement are the reasons based on the three levels related to the operations of selected objects: technical, organizational, decision-making and information exchange. 1. Technical level includes a set of features of the objects and performed with use of them. These activities, which aim at getting information of this type concern: •identification of complexity of technical objects which are the subject to modelling, •arrangement of maintenance procedures carried out on previously identified technical objects, •inventory of maintenance resources (labor, parts, tools, external services). 2. Organizational and decision-making level includes factors resulting from the way of functioning of the typi- Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 95 Science and Technology cal maintenance department, both internally as well as for external environment in wider context. The basis for this can be one of the models: a model of maintenance management [7] or the BCM model (Business Centered Maintenance) [9]. In this case, there are identified typical characteristics of standard maintenance department, allowing to gather a large amount of information and create knowledge repository for development of appropriate models, and exploitation scenarios. 3. The information exchange level includes criteria related both to the way of information flow within a typical maintenance organization, and to/from units of the external environment. Special attention is needed to identify the structure of information flow, which should begin with identification and distinction of the object that is maintenance unit (fig. 5). Fig.6. Hierarchical structure of exploitation information [7, 12] Fig.5. Diagram of standard maintenance organization in the relationship with the environment Such an object is a generalized and simplified model of information flow between the considered organizational unit and external entities. External units are both ones that are included in the organizational structure of the company (technical system), as well as ones which are outside of these structures. Inventory of the information flow should include identification: •input and output of information, •structure of information flow within the maintenance organization unit. Research under this phase must be related to real existing enterprises, to take into account features that result from different industries. It is assumed that the information and knowledge will be collected simultaneously with performing the analyses starting from the exploitation activities and functions, through technical objects identifying and modelling, formulating strategic and prevention plans, specifying the organizational aspects, by defining elements of control in the end. The result of the stage 1 should be a set of information and knowledge related to the functioning of a typical standard maintenance organization unit including computer tools (eg. CMMS/ EAM systems or process modelling tool - ARIS Toolset). 5.2. Stage 2: Building and testing of objects, events and processes models The objective of research carried out within this stage is to build models of objects, events and exploitation processes. These studies must be carried out based on the results of stage 1 – repository of information and knowledge about objects and exploitation processes. Due to the high diversity of models used in the area of exploitation, they will be built and organized on the basis of a diagram illustrating relationship between selected exploitation characteristics (fig. 6). 96 Each of the levels may include a set of models, which is characteristic for aspects of the exploitation of technical systems, discussed here. Due to the hierarchy of exploitation aspects arrangement, causing dependencies between superior and inferior elements, building and ordering the models will be done starting from the lowest level. According to the layout, it can be assumed that: •in the modelling of technical objects (object characteristics) it is necessary to use methods of structural and functional modelling, as well as methods of group and individual modelling, •for exploitation events, it is necessary to model intended and unintended events on the basis of the existing tools in this area (event cards, event trees, logic trees, etc.) [2, 12] taking into account specificity of decision making process as a result or effect of a particular situation, •exploitation processes will be the subject to modelling with the use of process maps and computer tools (eg. Aris Toolset), which allows to perform a simulation in relation to the prospective behaviour of technical objects. The result of implementation of research in this stage will be technical and organizational set of models, as the basis of exploitation scenarios and system supporting operational events and exploitation processes. 5.3. Stage 3: Assessment of the impact of events and processes on a exploitation condition of typical industrial enterprise Research in this stage will be carried out with the assumption, that the exploitation condition of industrial companies (from a technical point of view) determined by rank indicators divided into specified classes) will allow to determine the effects of particular events and processes in the enterprise. The subject of research will cover primarily identification of the classes that are basis for identifying the indicators of rank, using proposed set of measures allowing for quantitative evaluation of operating technical objects and functioning of servicing and maintenance department. Because of wide variety of this type of measurements, the aim of research in this area is assessment of their suitability, and then selection of these that will reflect the efficiency, quality and reliability of operation of analyzed technical objects best. Then, based on Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology models developed in stage 2, for each identified class there will be performed identification of particular events and processes and their impact on the state of and enterprises. This requires an extensive work, that will classify technical and non-technical (organizational, economic) factors of events and processes, influencing the company condition. That outlined study will be carried out in three ways: •by introducing a "trial" object failure and simulating possible effects, •by detailed analysis of cause and effect analysis of past events and activities realized as a result of this (exploitation processes), organized with including models that arise from Root Cause Failure Analysis (RCFA) methodology, •by detailed analysis of current events and their effects from which the information will complement the research material (occurring along with the research). After arranging the results of research and developing of cause-effect lists, events and processes will be assigned to the appropriate rank indicator. 5.4. Stage 4: Building and testing of scenario models of events and exploitation processes During this stage, starting from the results of research done within the previous stages, there will be built exploitation process models, by introducing some elements of scenario methodology. It will be then possible to include multi-aspect nature of exploitation processes management and optimal decision-making. Due to specificity of exploitation, there will be taken into account two types of scenarios [4, 23]: scenarios of possible events and simulation scenarios. This stage will be implemented in two steps: 1. Based on the selected way of scenarios creating, there will be designed a set exploitation scenarios for selected technical and organizational area. Selection of the area will be determined by the results of research carried out in the previous stages, in particular the developed objects, events, processes models (stage 2) and research on their influence on particular events and processes in the enterprise (stage 3). 2. Based on a set of developed scenarios of events and processes, there will be developed guidelines for use of scenario techniques as a basis for modelling exploitation events and processes. This will allow to make strategic decisions about This will allow to make strategic decisions about exploiting technical objects and functioning of maintenance department. The result of the research in this stage will be a set of exploitation processes models built using scenario methods along with the guidelines that define their use in medium-and longterm planning and implementation of maintenance tasks. 5.5. Stage 5: Building a system supporting operational events and exploitation processes The aim of the research in this stage will comprise development of principles for building a system supporting operational events and exploitation processes and building such a system, basingd on the results of previous stages. As a starting point for this research it is assumed that the system will consist of two layers: • data and knowledge acquisition layer about the events and exploitation processes, in the form of components ensuring communication with data and knowledge sources in the enterprise information system, • exploitation decision support layer, in the form of subsystem of control of flow related to events and processes, including exploitation scenarios, as a set of components of a system supporting operational events and exploitation processes. The result of the research in this stage will be a prototype system supporting operational events and exploitation processes, which action will confirm practical feasibility of the assumptions in this research. 6. Summary Taking into account the multi-aspect and multivariant nature of maintenance management process, reflected by the possibility of optimization seems to be more realistic in terms of scenario than in classical terms. The shape of the model and a set of criteria must take into account a wide range of technical and nontechnical aspects with as well as their connections “inside” modelled processes and with their closer and farther environment. It may be possible provided that an appropriate scenario building method is used to describe existing exploitation situation. The limit of complexity of the scenario description seems to be not in the same method but in the capabilities of the tools - particularly computer tools - used for practical applications of this method. The presented method of modelling exploitation processes is a key element of the author’s research carried out in the Institute of Production Engineering of the Silesian University of Technology. Justification for undertaking this subject can be specified in two fields: 1. in the exploitation theory, 2. in the area of methods and supporting tools in industrial practice. The first area, according to the author, due to unreasonable lack of greater interest in the use of scenario techniques in the field of engineering sciences, despite the fact that these techniques have been known for many years and applied in the field of economic sciences. According to the author, the research discussed here can create added value for the area of machinery and technical object exploitation. The second area emerges from practical and long-term author’s experience in functioning of maintenance departments in a wide variety of industries (food, automotive, paper, chemical, water and sewage, etc.). Currently, the most important exploitation problems of engineers relate to long term planning, including moments of the replacement of worn and damaged components and whole objects, in terms of the need to continually costs reducing. These problems do not occur in technical isolation, but there must be also taken into account organizational, economic, and normative-legal aspects of the enterprise. Until now, these problems were solved in two distinct ways: •technical, that is based on indicators of reliability and supplemented in some cases by the results of diagnostic tests, •organizational and economic, that is based on indicators of cost of the enterprise’s technical activities. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 97 Science and Technology Use of scenario techniques in modelling of exploitation processes can be an effective tool for solving problems of planning and implementation of maintenance tasks in many industrial enterprises and therefore will enable implementation of specific actions for: •implementation of modern maintenance management philosophy (TPM, RCM, WCM), • building of intelligent system supporting operational events and exploitation processes, • developing exploitation procedures in relation to the enterprises being at the design stage, • performing different analyses (eg. failure analysis), and above all •optimizing exploitation decision-making process. The article is funded by National Science Centre in Poland under the project no. 5636/B/T02/2011/40 titled: “The use of scenario methods in exploitation processes modelling”. Bibliography 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. Cempel C. Diagnostyka wibroakustyczna maszyn. Warszawa: WNT, 1989. Cholewa W, Kaźmierczak J. Data Processing and Reasoning in Technical Diagnostics. Warszawa: WNT, 1995. Ducot C, Lubben G J. A Typology of Scenarios. Futures 1980: No. 1. Gierszewska G, Romanowska M. Analiza strategiczna przedsiębiorstwa. Warszawa: PWE, 2003. Heijden K. Planowanie scenariuszowe w zarządzaniu strategicznym. Kraków: Dom Wydawniczy ABC, 2000. Kahn H, Wiener A J. The Year 2000. A Framework for Speculation on the Next Thirty Three Years. Macmillan, New York 1967. Kaźmierczak J. Eksploatacja systemów technicznych. Gliwice: Wydawnictwa Politechniki Śląskiej, 2000. Kaźmierczak J, Loska A. Scenariusze eksploatacyjne. Służby Utrzymania Ruchu 2007; 3: 64–69. Kelly A. Strategic Maintenance Planning. Butterworth-Heinemann, Oxford 2006. Legutko S. Development trends in machines operation maintenance. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2009; 2(49): 8–16. Levitt J. The Handbook of Maintenance Management. New York: Industrial Press Inc., 1997. Loska A. Bazy danych w zarządzaniu eksploatacją maszyn i urządzeń. Rozprawa doktorska. Gliwice: Politechnika Śląska, 2002. Loska A. Koncepcja wykorzystania scenariuszy w modelowaniu procesów eksploatacyjnych. Ekonomika i Organizacja Przedsiębiorstwa ORGMASZ 2005; 4: 106. Loska A. Modelowy sposób funkcjonowania służb utrzymania ruchu - strategie eksploatacyjne, Służby Utrzymania Ruchu; 2010: 3(21). Loska A. Pozyskiwanie informacji dla potrzeb budowy scenariuszy eksploatacyjnych. Świnoujście-Kopenhaga 2006: IV Międzynarodowa Konferencja Naukowo-Techniczna EXPLO-SHIP, 197–206. Loska A. Przegląd metod modelowania jako podstawa budowy scenariuszy eksploatacyjnych. Opole: Oficyna Wydawnicza Polskiego Towarzystwa Zarządzania Produkcją, 2010, t. 2, 152–161. Loska A. Sposób wspomagania zarządzania utrzymaniem ruchu z wykorzystaniem wybranych narzędzi informatycznych, Służby Utrzymania Ruchu 2008–2009. Loska A. Zastosowanie metodologii Business Centered Maintenance w pozyskiwaniu i porządkowaniu informacji dla potrzeb konstruowania scenariuszy eksploatacyjnych. Ekonomika i Organizacja Przedsiębiorstwa ORGMASZ 2008; 6: 77–78. Loska A., Senczyna S., Kaźmierczak J. Modelowanie struktury technicznej przedsiębiorstwa dla potrzeb komputerowego wspomagania zarządzania eksploatacją. Ekonomika i Organizacja Przedsiębiorstwa ORGMASZ 2007; 7: 81–82. Moubray J. RCM II – Reliability Centered Maintenance. Oxford: Elsevier Ltd., 2007. Niebel W B. Engineering Maintenance Management. New York: Marcel Dekker Inc., 1994. Percy D F., Kobbacy K. A. H. Preventive maintenance modelling. A Bayesian perspective, Journal of Quality in Maintenance Engineering 1996; vol. 2, nr 1: 15–24. Praca zbiorowa po red. Cieślak M. Prognozowanie gospodarcze. Metody i Zastosowanie. Warszawa: PWN, 2001. Praca zbiorowa pod red. Koradecka D. Bezpieczeństwo pracy i ergonomia. Warszawa: Centralny Instytut Ochrony Pracy, 1999. Skotnicka-Zasadzień B, Biały W. An analysis of possibilities to use Pareto chart for evaluating mining machines’ failure frequency. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2011; 3(51): 51–55. Staniszewski R. Sterowanie procesem eksploatacji. Warszawa: WNT, 1990. Żółtowski B, Niziński S. Modelowanie procesów eksploatacji maszyn. Bydgoszcz, Sulejówek: MARCAR, 2002. Żółtowski B.: Podstawy diagnostyki maszyn. Bydgoszcz: Akademia Techniczno-Rolnicza, 1996. Andrzej Loska, PhD (Eng) Institute of Production Engineeiring Silesian University of Technology ul. Roosevelta nr 26, 41-800 Zabrze, Poland E-mail: [email protected] 98 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Article citation info: ZAHARIA S M, MARTINESCU I, MORARIU C O. Life time prediction using accelerated test data of the specimens from mechanical element. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 99–106. Sebastian Marian Zaharia Ionel Martinescu Cristin Olimpiu Morariu Life time prediction using accelerated test data of the specimens from mechanical elemenT Prognozowanie czasu pracy elementu mechanicznego z wykorzystaniem danych z badań przyspieszonych In accelerated life testing, products are exposed to stress levels higher than those at normal use in order to obtain information in a short time. In this paper ,we expand the limits of performing accelerated life tests to an aerospace product (supple platinum). Specimens from the IAR 330 Puma helicopter structure, made of supple platinum, were subjected to accelerated life testing, and a significant reduction of the testing time was obtained. A simulation of the accelerated life testing data from the same case study was performed using the Monte Carlo method, with the purpose of comparing the data resulting from the experimental study (accelerated life tests) with the simulated data. Keywords: reliability, lifetime, accelerated life tests, acceleration model, Monte Carlo simulation. W przyspieszonych badaniach cyklu życia, produkty poddaje się naprężeniom o wartościach wyższych niż występujące w normalnych warunkach użytkowania, w celu szybszego uzyskania informacji. W przedstawionym artykule poszerzamy granice zastosowań badań przyspieszonych o produkt przemysłu lotniczego (giętka platyna, ang. supple platinum). Próbki z giętkiej platyny pochodzące z elementu konstrukcji śmigłowca IAR 330 Puma poddano badaniom przyspieszonym otrzymując znaczne skrócenie czasu badania. Dla otrzymanych danych z badań przyspieszonych przeprowadzono symulację metodą Monte Carlo, w celu porównania danych eksperymentalnych (z badań przyspieszonych) z danymi symulacyjnymi. Słowa kluczowe: : niezawodność, cykl życia, badania przyspieszone, model przyspieszenia, symulacja Monte Carlo. 1. Introduction The obtaining information regarding the reliability of mechanical products is usually done either by following the behaviour of the products during operation or during the reliability tests. In the real time operation we carefully record all the phenomena that occur during the product’s use. A study based on this information represents only a “historical” study, its value being only that of collection experimental data or screening factors that can lead to a low reliability of the products. The information gathered from the operation usually refers to products or equipments suffering from the wear process, so that at the moment of the conclusions they might not be that important for the improvement of some aspects regarding the design and manufacturing of products, as a requirement of reliability [1, 10]. Considering the main requirement of reliability - the increase of performance of the industrial products, in direct connection with the scope of the study - it is necessary to extend a significant importance to reliability tests. A reliability test is represented by an experiment performed to determine the parameters of reliability for a well-defined product. The main reliability parameters during reliability tests is mean time to failure (MTTF), knowing that, based on the existing relations between reliability parameters, they can be easily deduced from one in other. It is required to specify that in order to apply the statistical methods of estimation of the reliability parameters, is recommended to know the low of distribution for the mean time to failure [21]. Many of the mechanical products produced today for complex technical systems have very high reliability under normal use conditions. The questions then arise of how to make the optimal choice between several types or designs of a device and how to collect information about the corresponding life distributions under normal use conditions. A common way of tackling these problems is to expose the products to sufficient overstress to bring the mean time to failure down to an acceptable level. Thereafter, one tries to “extrapolate” from the information obtained under over stress to normal use conditions. This approach is called Accelerated Life Testing (ALT) or overstress testing [1, 5, 6, 12, 13, 15, 19, 21]. In these tests, reliability practitioners may force the product to fail more quickly than it would under normal use conditions. Accelerated failure time modelling is one part of the quantitative accelerated life testing. The interest of this theory is to know the influence of any given stress on the life duration of an item. The technique of accelerated life testing enjoys both intuitive appeal and objective support from the physics of failure, and can be implemented in various ways; experimental aims are, however, usually phrased in statistical terms. For instance, the practitioner may wish to Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 99 Science and Technology estimate a specified percentile of life or the probability of failure within a given warranty period of an item operating under design stress. The primary purpose of an ALT is to estimate the life distribution and quantities of interest at a use condition. This estimation involves extrapolation from higher stress levels by using an acceleration model, and thus includes the model error and statistical uncertainty. Sometimes, the model error outweighs the statistical one. The model error may be reduced or eliminated only by better understanding the failure mechanisms and using a more accurate model, whereas the statistical uncertainty can be reduced by carefully selecting a good test plan. A typical test plan is characterized by the stress levels, the number of test units allocated to each level, and their censoring times. The most commonly used ALT in modern manufacturing industry is the constant-stress ALT (CSALT) where stress applied to a sample of units is constant. A typical parametric model of CSALT consists of two components: (1) a lifetime distribution that models the time-to-failure at a constant-stress level; and (2) a stress-life model that quantifies the manner in which the lifetime distribution changes a cross different stress levels. There are different types of ALT plans in use, which include subjective, traditional, best traditional, statistically optimum and compromise plans [4, 6, 8, 9, 17, 19, 20]. Pursuing the previously stated main purpose of the accelerated life tests, we need a model that relates life to accelerating stress, such as temperature, humidity, and voltage. Such models, usually called acceleration models, can be classified into the following three types: physical models; quasiphysical models; empirical models. Among the most important acceleration models (life-stress) we can mention the following: Arrhenius, Eyring, Inverse Power Law; Life - Thermal Cycling, Life - Voltage, Life - Vibration, Life - Humidity, Life - Temperature - Humidity [2, 5, 8, 13, 15, 21]. Accelerated life testing [7,11, 13, 16, 21] is used in electronics, (resistors, lasers, liquid crystal displays, electronic bounds, switches, circuit breakers, relays, cells and batteries) in the study of metals and composite materials, but also for certain components and mechanical assemblies (automobile parts, hydraulic components, tools, bearings). In aerospace industry, the accelerated life testing is used to test certain components: the engine, the oil pumps, the landing gear, onboard electronic equipment and stiffening components (strips, spear). As a result of the critical study above on the accelerated life tests, the scope of the paper becomes evident, which is: the theoretical and experimental study regarding the management of accelerated life tests for a product from the field of aviation (supple platinum), in order to reduce the testing time and obviously of the material costs related to this kinds of tests. 2. Case study In this paper, accelerated life tests will be implemented and performed on a vital component from the structure of the IAR 330 Puma helicopter, following all the stages that are necessary to the execution of such tests. Starting from the constructive and functional analysis of the tested product and of the experimental bench, then proceeding with the optimization and the design of the accelerated life test plans and finalizing with the 100 statistical processing of the experimental data using software that is specific to accelerated life tests. For a validation of the accelerated life tests we used the Monte Carlo simulation method for the data in accelerated conditions. The purpose of using this method is to make a comparison between the data from accelerated conditions that resulted from the Monte Carlo simulation and the experimental data. 2.1. The necessity of implementation of accelerated life tests in the mechanical product’s testing If a mechanical product requires, for example, 108 cycles to cause a failure from fatigue in normal use condition, by using an accelerated life test we can obtain the same result in 105 cycles. The design of the accelerated life test and the interpretation of data require the understanding of the relation, in the course of the destructive process considered, between the level of stress and the failure rate. The result of these accelerated life tests is the reliability estimated by statistical and mathematical means, using specific software. Thus, for the design of the accelerated life tests plans and for the processing of statistical data obtained from the quantitative accelerated life test in this paper we used the ALTA 7 (Accelerated Life Testing Data Analysis) software. In the case study of this paper, we applied a cyclical mechanical stress. During the cyclical stress the most frequently used are the metallic components and systems and the phenomenon is fatigue. The degradation by fatigue involves a variety of aspects regarding to: the type of stress; the shape of the part; the quality of the part that will be processed; the environment of the part’s operation. The accelerated life tests where the main degradation phenomenon is fatigue are very important in the industrial field and especially in the aerospace industry [7, 11, 14, 18], with application on the components in the structure of airplanes and helicopters. The introduction of accelerated life tests in the aviation is necessary, because the fatigue tests can have the operating time of millions of cycles until the failing of components occurs (the helicopter blade, the supple platinum, command rods, the wing and the landing gear). That is the reason why the accelerated life tests represent a method by which the testing time of components from the aerospace field is shortened and thus leading to optimizing of the testing system. 2.2. Constructive and functional of supple platinum The french helicopter constructors introduced the system for reducing the vibrations like supple platinum. Supple platinum (Fig. 1) is made of a titanium alloy and is placed between the inferior part of the main transmission box and the mechanical board at the Puma IAR 330 helicopter. Supple platinum is a vital part and has the role to reduce the vertical vibrations produced by the main rotor’s blade. 2.3. The structure of the test bench Due to its configuration, supple platinum cannot be subjected to stress tests. That’s why the tests are being made on specimens, with standardized shapes and dimensions, made of the same material as supple platinum (a titanium alloy known in the aviation as TA 10). Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Fig. 1. Supple platinum The tests of the specimen made of supple platinum consists of the repeated stress using a bending force, on a test bench that includes: a fixation device for the specimen - the specimen is embedded at one end, and at the other end we apply a displacement (arrow) on vertical; the device for the tensioning of the specimen - the main function of the device is to introduce a displacement at the free end of the specimen; the energy installation - that is composed of an engine which produces the mechanical energy for production at displacement at the free end of the specimen; the command installation - has the role to turn off and on the engine during the testing of the specimen; the electronic installation - is composed of an apparatus that reads frequencies and a cycle counter. The test bench used for the fatigue testing of specimens is presented in figure 2. Poor accelerated life test plans waste time, effort and money and may not even yield the desired information. Before starting an accelerated life test, it is advisable to have a plan that helps in accurately estimating reliability at operating conditions while minimizing test time and costs. To design the plan for the accelerated life tests of the specimen made out of supple platinum it is necessary to establish the following parameters: A)The acceleration model: for accelerated life tests where the failure mechanism is a mechanical one (the fatigue in the case of accelerated life tests for the specimens made out of supple platinum), the most adequate one is the IPL model; The inverse power law (IPL) model is commonly used for non-thermal accelerated stresses and is given by: 1 (1) L(V) = KV n where L represents a quantifiable life measure, such as mean life, characteristic life, median life, B(x) life, etc; V represents the stress level; K is one of the model parameters to be determined, (K > 0); n is another model parameter to be determined. The failure rate under normal operating conditions is: Fig. 2. The test bench for the fatigue testing of specimens The data acquisition system is composed of: Hottinger measuring device with 6 channels, Hottinger LC 11 strain gauges, apparatus that show the frequency of stresses and a cycle counter. 2.4. Plan and accelerated test conditions of test specimens The experimental research has been performed on specimens made out of supple platinum, specifically from removed portions that result from the technological processing. Figure 3.a shows the constructive model of the specimen used in accelerated life tests and, also, the placement of strain gages. On the specimens are mounted and connected in half-bridge 2 strain gages used to measure the specific strain. The placement of this strain gages is described in figure 3b. λ(t, V) = βKVn ∙ (KVnt)β-1 (2) B)The number of specimens subjected to accelerated life tests: for accelerated life testing we used 20 specimens made from supple platinum; C)The distribution law of the number of cycles until failure used in accelerated life testing: the Weibull distribution was chosen to test the specimens; D)The stress under normal condition and in accelerated condition: the bending force in normal testing conditions is 50 daN and the maximum bending force is 70 daN (according to the specimens’ test chart). E) The accelerated life test plan: for the accelerated life testing of the specimens made from supple platinum we chose 3 levels best compromise plan. The test plan was realised using the ALTA software, introducing the aforementioned parameters. The ALTA 7 software generates an optimum testing report, where the levels of accelerated life testing and the number of specimens tested at every accelerated stress level are specified. The testing parameters resulted from the design of the accelerated life testing plan for the specimens made from supple platinum are as follows: 3 levels of testing: 60, 64, 70 daN; Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 101 Science and Technology Fig. 3. The specimen made of supple platinum: a) the specimen’s constructive model, b) the experimental model the number of tested specimens corresponding to the level of acceleration: 9, 5 and 6 specimens. 2.5. The statistical processing of experimental data For the results obtained from the accelerated life testing of the specimens for the 3 levels of accelerated stress we verified the hypothesis that the distribution law of the number of cycles until failure is Weibull (the Kolmogorov - Smirnov test was used). For this verification, we measured the spacing between empirical distribution function of the sample and the cumulative distribution function of the reference distribution. We compared these results with a level of confidence of the Kolmogorov – Smirnov test. Following the statistical processing of the experimental data for the 3 levels of stress the Weibull distribution was accepted. For the determination of the mean number of cycles to failure and of the reliability parameters under normal testing conditions (F=50 daN) for the specimens made from supple platinum, the experimental data resulted from accelerated conditions have been processed using the ALTA7 software. We calculated the three parameters corresponding to the IPL-Weibull model using the maximum likelihood estimation method for the data from the accelerated regime introduced in the ALTA7 software. The following values of parameters resulted: β=7.283; k=4.04267E18; n=6.788. We determined the acceleration factor corresponding to the IPL model. This is determined for every level of acceleration of the bending force. The accelerated levels are: La1=60 daN, La2=64 daN, La3=70 daN, and the normal testing level is Lu=50 daN. By calculating the product of the calculated values of the acceleration factors and the number of cycles to Tab. 1. The determination of the number of cycles to failure in normal conditions Num. The number of cycles to failure in accelerated conditions Acceleration factor The number of cycles to failure in normal conditions 1 141453 487711 2 157243 542153 3 169634 584875 3.447 4 182723 5 201634 695207 6 206256 711143 7 219634 757269 8 231238 797278 9 247234 852430 10 112787 388874 11 119462 12 123300 13 139498 745421 14 152087 812692 15 50231 268414 16 56922 17 67037 5.343 630005 411889 658866 304168 9.818 658200 18 73008 716826 19 78025 766085 20 85341 837917 102 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Table 2. The dependence between the numbers of cycles in normal conditions – Reliability - Unreliability – Failure Rate The number of cycles to failure in normal conditions Reliability R(t) Unreliability F(t) Failure Rate λ(t)·10-6 268414 0.999 0.001 0.020 304168 0.998 0.002 0.044 388874 0.989 0.011 0.204 411889 0.984 0.016 0.293 487711 0.945 0.055 0.847 542153 0.885 0.115 1.648 584875 0.808 0.192 2.654 630005 0.693 0.307 4.233 658200 0.604 0.396 5.573 658866 0.602 0.398 5.609 695207 0.472 0.528 7.859 711143 0.413 0.587 9.062 716826 0.392 0.608 9.527 745421 0.287 0.713 12.181 757269 0.247 0.753 13.450 766085 0.218 0.782 14.465 797278 0.131 0.869 18.587 812692 0.096 0.904 20.964 837917 0.054 0.946 25.403 852430 0.036 0.964 28.297 failure in accelerated conditions we determined the number of cycles to failure in normal testing conditions (Table 1) for the specimens made from supple platinum from the structure of the IAR 330 Puma helicopter. We determined the reliability parameters (the reliability function, unreliability and the rate of failure) depending on the number of cycles to failure in normal testing condition (Table 2). Using the calculated values (the number of cycles in normal testing conditions), the reliability function 3D (Fig. 4a) and the failure rate 3D (Fig. 4b) were plotted. The main purpose of accelerated quantitative tests is to determine the life time in normal testing conditions. Using the data resulted from accelerated life tests we can determine the mean number of cycles to failure of the specimens made from supple a) b) Fig. 4. Reliability parameters: a) reliability function, b) failure rate Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 103 Science and Technology Fig. 5. The determination of the number of cycles to failure of the specimens in normal testing conditionst platinum in normal testing conditions. The mean number of cycles to failure is determined using the relation of the mean time to failure from table 1. The mean number of cycles to failure for the tested specimens was of 677882. Life versus stress plots is the most important plot type in accelerated life testing analysis. Life versus stress plots are widely used for estimating the parameters of life-stress relationships. Any life measure can be plotted versus stress in the life vs. stress plots available in ALTA 7. In figure 5, by drawing a line through the mean number of cycles to failure for the 3 levels of acceleration (60 daN, 64 daN and 70 daN) and marking the point of intersection of this line with the vertical line at the normal level of stress of 50 daN, we found out the mean number of cycles to failure in normal testing conditions. The mean number of cycles to failure of the specimens in normal testing conditions given by the manufacturer at the approval of the vital element supple platinum is 700000 cycles, whereas the mean number of cycles to failure obtained from the accelerated life tests was of 677882 cycles. By adding the number of cycles to failure resulted from the accelerated life tests, this is of 2814747. The total number of cycles to failure in normal testing conditions is 12627432. The main purpose of the accelerated life tests on the specimens made from supple platinum from the structure of the IAR 330 Puma helicopter, to reduce the number of cycles using the accelerated life tests, was validated. Using the accelerated life tests on the specimens made from supple platinum from the structure of the IAR 330 Puma helicopter, the number of cycles to failure has been reduced by 4.5 times, making this result responsible for significant reductions of the material costs. 2.6. The application of the Monte Carlo method for the simulation of data in accelerated conditions (for supple platinum - a component of the IAR 330 Puma helicopter) Fig. 6. The calculation of the mean number of cycles to failure in normal testing conditions for the specimens of supple platinum with the Monte Carlo method using QCP 104 Using the Monte Carlo method we simulated N stages of a product with the help of an acceleration model (Inverse Power Law) and statistical distribution (Weibull) which are suited to the analyzed case study. Using the previously determined parameters (β=7.283; k=4.04267E-18; n=6.788) and the three accelerated levels (60, 64 and 70 daN), we simulated with the Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology help of ALTA7 software the values for the number of cycles to failure in accelerated conditions. Following the simulation of accelerated data using the Monte Carlo method for the specimens from supple platinum from the structure of the IAR 330 Puma helicopter, we obtained the value of 694789, which represents the mean number of cycles until failure (Fig. 6), which is close to the value of 677882, which represents the mean number of cycles to failure of the specimens tested in accelerated conditions in the previously presented abovementioned experiment. The Quick Calculation Pad (QCP) provides you with a quick and accurate way of gaining access to some of the most frequently requested reliability results. 3. Conclusion At some industrial products (from the aviation, nuclear and electronic fields), for which a high reliability is estimated, the determination of the life time and of the reliability parameters, under normal stress conditions, implies a long testing period. For this reason we opted for the accelerated life testing methods. These are tests being performed at more intense stress conditions, compared to the normal stress conditions, with the purpose of intensifying the degradation processes and, as an economic result, the shortening of the period and costs related to the testing, while preserving the same failure modes and mechanisms. The implementation of accelerated life tests on the products from the aviation field has produced a significant reduction of the testing time. For the case study under analysis, figure 7 shows the number of cycles to failure from: accelerated life tests, the simulation with the Monte Carlo method and the data from the normal testing conditions. We can observe that, by using the accelerated life tests, the testing time has been reduced by 4.5 times. Given the fierce competition existing on the aviation industry market and considering the reduction in testing time and therefore the optimizing of the products’ life cycle, many companies will implement and develop various methods of obtaining data as fast as possible regarding the reliability and quality of the products. Fig. 7. Significant reduction of time using accelerated life tests Acknowledgement: This paper is supported by the Sectoral Operational Programme Human Resources Development (SOP HRD), financed from the European Social Fund and by the Romanian Government under the project number POSDRU/89/1.5/S/59323. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. Bagdonavičius V, Nikulin M. Accelerated Life Models: Modeling and Statistical Analysis. Boca Raton: Chapman & Hall/CRC Press, 2002. Bai D S, Chung S W. An accelerated life test model with the inverse power law. Reliability Engineering and System Safety 1989; 24(3): 223–230. Couallier V, Denis L, Bagdonavicius V, Nikulin M, Gerville–Reache L. Statistical inference through AFT model for biotechnical systems. Journal of Statistical Planning and Inference 2009; 139(5): 1649–1656. Elsayed E A, Zhang H. Design of PH-based accelerated life testing plans under multiple-stress-type. Reliability Engineering and System Safety 2007; 92(3): 286 – 292. Escobar L A, Meeker W Q. A Review of Accelerated Test Models. Statistical Science 2006; 21(4): 552–577. Fard N, Li C. Optimal simple step stress accelerated life test design for reliability prediction. Journal of Statistical Planning and Inference 2009; 139(5): 1799–1808. Khatibi G, Lederer M, Weiss B, Licht T, Bernardi J, Danninger H. Accelerated mechanical fatigue testing and lifetime of interconnects in microelectronics. Procedia Engineering 2010; 2(1): 511–519. Ling L, Xua W, Li M. Parametric inference for progressive Type-I hybrid censored data on a simple step-stress accelerated life test model. Mathematics and Computers in Simulation 2009; 79(10): 3110–3121. Liu X, Tang L C. Planning sequential constant-stress accelerated life tests with stepwise loaded auxiliary acceleration factor. Journal of Statistical Planning and Inference 2010; 140(7): 1968–1985. Meeker W Q, Escobar L A. Statistical Methods for Reliability Data. New York: Wiley, 1998. Miyano Y, Nakada M, Nishigaki K. Prediction of long-term fatigue life of quasi-isotropic CFRP laminates for aircraft use. International Journal of Fatigue 2006; 28(10): 1217–1225. Mohammadian S H, Aït-Kadi D. Design stage confirmation of lifetime improvement for newly modified products through accelerated life testing. Reliability Engineering and System Safety 2010; 95(8): 897–905. Nelson W B. Accelerated Testing: Statistical Models, Test Plans, and Data Analysis. New Jersey: Wiley, 2004. Nelson W B. Fitting of Fatigue Curves with Nonconstant Standard Deviation to Data with Runouts. Journal of Testing and Evaluation 1984; 12(2): 69–77. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 105 Science and Technology 15. Nikulin M, Gerville-Reache L. On statistical modelling in accelerated life testing. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2006; 2(30): 48 – 52. 16. Özsoy S, Çelik M, Suat Kadıoğlu F. An accelerated life test approach for aerospace structural components. Engineering Failure Analysis 2008; 15(7): 946 – 957. 17. Pascual F. Accelerated life test planning with independent lognormal competing risks. Journal of Statistical Planning and Inference 2010; 140(4): 1089–1100. 18. Poulose P K, Jones D L. Prediction of constant amplitude fatigue lives of precracked specimens from accelerated fatigue data. International Journal of Fatigue 1983; 5(2): 99–103. 19. Seo J H, Jung M, Kim C M. Design of accelerated life test sampling plans with a nonconstant shape parameter. European Journal of Operational Research 2009; 197 (2): 659–666. 20. Wang B X. Testing for the validity of the assumptions in the exponential step-stress accelerated life-testing model. Computational Statistics and Data Analysis 2009; 53(7): 2702 – 2709. 21. Yang G B. Life Cycle Reliability Engineering. New Jersey: Wiley, 2007. Sebastian Marian Zaharia, PhD Prof. Ionel Martinescu, PhD Associate Professor Cristin Olimpiu Morariu, PhD Department of Advanced Manufacturing Technologies and Systems Transilvania University of Brasov B-dul Eroilor nr. 29, 500036, Brasov, Romania E-mail: [email protected] 106 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Article citation info: Dębski H, Koszałka G, Ferdynus M. Application of fem in the analysis of the structure of a trailer supporting frame with variable operation parameters. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 107–114. Hubert Dębski Grzegorz Koszałka Mirosław Ferdynus APPLICATION OF FEM IN THE ANALYSIS OF THE STRUCTURE OF A TRAILER SUPPORTING FRAME WITH VARIABLE OPERATION PARAMETERS WYKORZYSTANIE MES W ANALIZIE STRUKTURY NOŚNEJ RAMY NACZEPY O ZMIENNYCH PARAMETRACH EKSPLOATACYJNYCH* This work presents a numerical analysis of the structure of a trailer frame of adjustable length and an increased load capacity designed for the transportation of oversize loads. The study was aimed at developing adequate numerical FEM models which would allow identification of the effort of the structure and the state of strain under operating loads. The analysis was carried out using ABAQUS/Standard, a numerical tool which enables calculations in the geometrically nonlinear range with the use of the incremental–iterative Newton–Raphson method. As a result of the analysis, trouble areas in the frame were found in which dangerous stresses occurred. This enabled modification of the structure, leading to a reduction of the stresses to a safe level. Keywords: finite element method, supporting frame, strength calculations, trailer, oversized transport. W pracy zaprezentowano numeryczną analizę konstrukcji ramy naczepy o zmiennej długości i podwyższonej nośności przeznaczonej do transportu ładunków ponadnormatywnych. Prowadzone badania miały na celu opracowanie adekwatnych modeli numerycznych MES umożliwiających identyfikację wytężenia konstrukcji oraz stanu odkształcenia w warunkach obciążeń eksploatacyjnych. Zastosowanym do analizy narzędziem numerycznym był program Abaqus/Standard, umożliwiający prowadzenie obliczeń w zakresie geometrycznie nieliniowym z wykorzystaniem przyrostowo-iteracyjnej metody Newtona-Raphsona. W wyniku przeprowadzonych prac ustalono newralgiczne obszary ramy w których występowały niebezpieczne naprężenia. Umożliwiło to modyfikację konstrukcji pozwalającą na zmniejszenie naprężeń do bezpiecznego poziomu. Słowa kluczowe: metoda elementów skończonych, rama nośna, obliczenia wytrzymałościowe, naczepa, transport nienormatywny. 1. Introduction Design of modern structures is a complex task that requires taking account of numerous material and mechanical features when optimizing the geometry of the structure being designed. This concerns, in particular, critical structural elements, whose load capacity determines the strength of the entire supporting system. The process of designing such elements necessitates the use of modern tools which make it possible to search for the best design solutions [10–12]. A contemporary tool that provides a wide range of possibilities of analyzing the strength parameters of a structure being designed is CAE numerical software, which uses the finite element method [3–6, 8]. Such software currently finds broad application in many branches of industry, in particular the aerospace, aircraft and automotive industries. A group of supporting structures that have to meet high strength and stiffness requirements are the frames used in the modern means of transport. Among design solutions that are subjected to the action of particularly high operating loads are frames of trailers for the transportation of oversize or overweight loads. Designing of this type of structures requires taking account of the various operating configurations of the trailer which make it possible, when necessary, to extend the length and width of the load deck [9–10]. A change of the configuration of the structure during its operation considerably affects the change in the character of its loading, which requires additional consideration in the process of designing and construction of the structure. This study presents a conception of a central frame whose length can be changed over a wide range depending on the operating needs. A solution based on two mating thin-wall box beams was adopted in which frame length was adjusted by mutual sliding of the load-bearing members in and out of each other. In the calculations, structural details were taken into account, such as process holes, which could have a significant effect on the assessment of the stress-strain state of the structure of the frame. The calculations were done us- (*) Tekst artykułu w polskiej wersji językowej dostępny w elektronicznym wydaniu kwartalnika na stronie www.ein.org.pl Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 107 Science and Technology ing the Abaqus/Standard software, a modern numerical tool that uses the finite element method in the geometrically and physically non-linear range [1, 2, 6, 8]. 2. Goal and scope of the study The aim of the study was to assess the strength and stiffness of the design solution developed for a trailer support frame intended for the transportation of long loads of up to 45 tons. In the adopted concept of the frame, consisting of two independent parts – front and rear, length could be easily changed by sliding the front part of the frame in and out of its rear part. This enabled adjustment of the length of the frame to the transported load. The new structural solution required precise analysis of the state of strain and effort, which was conducted using the finite element method. Prior to FEM calculations, the bending moments acting on the trailer frame were determined analytically for different combinations of its length and distribution of the transported load. The analysis covered only those cases which were deemed probable and permissible (e.g., loading of a maximally extended frame with a point force, of a value corresponding to the maximum permissible load weight, applied in the middle of its length was considered unlikely and was also listed in the trailer user's manual as impermissible). On the basis of the above considerations, cases for FEM analyses were chosen (Fig. 1). Because the analysis of the structure was carried out in the range of operating loads, thus not allowing permanent mined by the vehicle suspension system, should simultaneously provide support for the front part, the movable element of the frame. The length of the frame could be changed by sliding the front part in and out of the rear part structure of the frame. A general view of the numerical model of the frame is shown in Fig. 2 The loading of the numerical model consisted of the structure’s own weight as well as point mass loading deriving from the weight of the load of 45 tons, additionally taking into account 20% of dynamic surplus. A schematic diagram of the distribution of the external load in the form of point masses for the two frame configurations is shown in Fig. 2. The structure of the spatial discrete model of the frame was based on shell-type elements having six degrees of freedom per element node [1]. In the joining area between the front and rear part of the frame, a solution was adopted which consisted in using flexible cushions which enlarged the mating area, and, by the same token, eliminated adverse local edge pressures in the zone of the joint. The mutual interactions between the front and the rear part of the frame were reproduced by defining contact interactions between the mating elements. 4. Boundary conditions of the discrete model The principal task in developing a discrete model of the structure of the frame was to reproduce the operating conditions of the air suspension, which ensures identical reactions for each trailer wheel axle during operation of the vehicle. It was decided that the control arm would be reproduced by means of beam elements of specific stiffness for which different support conditions were defined in the front and rear node of the model of the control arm. The connection between the control arm and the frame in the front node ensured rigid transfer of load from the suspension onto the structure of the frame, allowing the arm only to rotate relative to the wheel axles. The Fig. 1. Schematic diagrams of loading of a retracted (a) and an extended (b) frame with a load rear node of the control arm was connected with the frame via an spring reproducing the air bag of the air suspension. The stiffdeformations to develop in structure members, a decision ness of the element modelling the air bag was selected so as was taken to adopt a linear elastic material model. Taking to ensure similar reactions in the axles of all trailer wheels. into consideration the contact interactions applied to the The support of an individual wheel was realized by deprivnumerical model and the possibility of occurrence of large ing the middle node of the control arm (in the connection deformations, it was decided that the calculations should be node between beam elements) of the possibility of displacecarried out in the geometrically non-linear range using the ment in the direction of the X axis – Fig. 3a. incremental–iterative Newton–Raphson method [2, 6, 8]. For the front part of the frame, support of the kingpin was defined by disabling displacement of kingpin nodes in 3. Development of a numerical FEM model the direction of the Y and Z axes – Fig. 3b. Moreover, in the direction of the X axis (vertical movements of the kingpin), Two configurations of frame operation were considered: support of nodes was introduced in the form of elastic rea retracted frame configuration and a maximally extended sponse of an appropriately specified stiffness. Additionally, frame configuration. In both variants, the same design and for the entire model, axial symmetry conditions were defined material parameters were used. The structure of the frame (along the axis of the frame – the Y axis) by depriving the was designed so that its rear part, being the basic element nodes located on the model’s symmetry axis of the possibilof the structure for which the boundary conditions are deter- 108 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology a) b) Fig. 2. Discrete models of the frame indicating the manner in which external loading was modelled: a) retracted configuration, (b) extended configuration ity of displacement in the direction of the Z axis and rotation relative to the X and Y axes. Elimination of these degrees of freedom ensured stable work of the model without unwanted displacements of the structure in the direction crosswise to its axis (sideways) – Fig. 3b. 5. Results of numerical analysis The numerical calculations conducted in this study made possible evaluation and comparison of the state of strain and effort of the structure of the frame for both of the considered operating variants – the retracted and maximally extended frame. The obtained maps of distribution of H-M-H reduced stresses in the elements of the structure pointed to the areas in which those stresses considerably exceeded the adopted yield point of the material Re = 360 MPa. The most adverse stress gradients occurred in the sliding part of the frame and in the joining zone between the two load-bearing members, both in the case of the retracted and the extended configuration of the frame – Fig. 4. The very high levels of reduced stress occurring in some areas required elimination, which called for a better design solution. Therefore, modifications of structural details were introduced in the current solution. They primarily involved a change in the location of the process holes in the side walls of the frame, which were a source of dangerous stress gradients. The openings were removed and replaced with one hole in the lower beam wall. Moreover, flat bars for the reinforcement of the side walls of the frame were introduced in the risk zone and reinforcing ribs in the gooseneck were added – Fig. 5. As a result of such activities, successive variants of the structure were being developed, which were then subjected to FEM analysis. The newly introduced modifications made possible elimination of the trouble zones in the supporting system. Fig. 6. shows the results of FEM calculations for the final variant of the structure. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 109 Science and Technology a) b) Fig. 3. Boundary conditions of the FEM model: a) model of the wheel suspension, b) support of the frame’s kingpin and the model’s axial symmetry conditions 6. Conclusions The numerical analysis of the structure of the support frame of a trailer conducted in this study enables evaluation of the adequacy of the newly developed design solution, allowing identification of the trouble areas determining the strength of the entire structure. This is extremely important in cases where, in the process of looking for new design solutions, too many unknown design parameters occur at the stage of designing complex load-bearing members. The knowledge of stress distribution in the critical elements is then an issue of primary importance, and the application of the finite element method allows analysis of the effort of a structure as early as the design stage [6, 8, 9]. Use of numerical FEM calculations in the design process enables fast and effective introduction of indispensable 110 modifications of structural details, leading to the creation of successive, ever more adequate design variants which, ultimately, make it possible to develop an optimum solution. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology a) b) Fig. 4. H-M-H stress distribution in the model of the frame: a) retracted frame – general view, b) extended frame – detail of the joining zone Fig. 5. Cover plates and ribs reinforcing the front part of the frame Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 111 Science and Technology a) b) Fig. 6. H-M-H stress distribution in the modified model of the frame: a) general view b) detail of the front part of the frame References 1. 2. 3. 4. 5. 6. 7. 8. 9. Abaqus HTML Documentation, 2010. Bathe K J. Finite element procedures. Prentice Hall, 1996. Dacko M, Borkowski W, Dobrociński S, Niezgoda T, Wieczorek M. Metoda elementów skończonych w mechanice konstrukcji. Arkady, Warszawa 1994. Grądzki R. Wprowadzenie do metody elementów skończonych. Politechnika Łódzka, Łódź 2002. Koszałka G., Niewczas A., Dębski H., Golec M., Kaczor M., Taratuta L. FEM analysis in design of extendable central beam for a semi-trailer. Machine Design, 2012(3);1: 47–50. Osiński J. Obliczenia wytrzymałościowe elementów maszyn z zastosowaniem metody elementów skończonych. Oficyna Wydawnicza PW, Warszawa 1997. Rakowski G, Kacprzyk Z. Metoda Elementów Skończonych w mechanice konstrukcji. Oficyna Wydawnicza PW, Warszawa 2005. Rudawska A., Dębski H. Experimental and numerical analysis of adhesively bonded aluminum alloy sheets joints. Eksploatacja i Niezawodnosc – Maintenance and Reliability, 2011; 1: 4–10. Rusiński E, Czmochowski J, Smolnicki T. Zaawansowana metoda elementów skończonych w konstrukcjach nośnych. Oficyna Wydawnicza Politechniki Wrocławskiej, Wrocław 2000. 112 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology 10. 11. 12. 13. Rusiński E, Kowalczyk M. Zasady przedłużania ram samochodów ciężarowych. Samochody Specjalne, 2003; 6: 78–81. Rusiński E, Porada D. Komputerowe wspomaganie projektowania ram samochodów ciężarowych. Samochody Specjalne, 2000; 6: 72–73. Tarnowski W. Podstawy projektowania technicznego. WNT, Warszawa 1997. Winkler T. Komputerowy zapis konstrukcji. WNT, Warszawa 1989. Hubert Dębski, PhD (Eng.) Grzegorz Koszałka, PhD (Eng.) Mirosław Ferdynus, PhD (Eng.) Faculty of Mechanical Engineering Lublin University of Technology Nadbystrzycka 36, 20-816 Lublin, Poland E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 113 Article citation info: ZHANG XL, HUANG HZ, WANG ZL, XIAO NC, LI YF. Uncertainty analysis method based on the combination of maximum entropy principle and point estimation method. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 114–119. Xiao-Ling ZHANG Hong-Zhong HUANG Zhong-Lai WANG Ning-Cong XIAO Yan-Feng LI UNCERTAINTY ANALYSIS METHOD BASED ON A COMBINATION OF THE MAXIMUM ENTROPY PRINCIPLE AND THE POINT ESTIMATION METHOD Metoda analizy niepewności oparta na połączeniu zasady maksymalnej entropii i metody oceny punktowej Uncertainty is inevitable in product design processes. Therefore, to make reliable decisions, uncertainty analysis incorporating all kinds of uncertainty is needed. In engineering practice, due to the incomplete knowledge, the distribution of some design variables can not be determined. Furthermore, the performance function is highly nonlinear, therefore, the high order moments of the performance function are needed to calculate the probability of failure accurately. In this paper, an uncertainty analysis method combining the maximum entropy principle and the bootstrapping method is proposed. Firstly, the bootstrapping method is used to calculate the confidence intervals of the first four moments for mixed random variables and sample variables. Secondly, the high order moments of limit state functions are estimated using the reduced dimension method. Thirdly, to calculate the probability density function (PDF) and cumulative distribution function (CDF) of the limit state functions, an optimization model based on the maximum entropy principle is formulated. In the proposed method, the assumptions that the distribution of the random variables are known and the calculation of the sensitivity for limit state function with respect to the Most Probable Point (MPP) are avoided. Finally, comparisons of results from the proposed methods and the MCS method are presented and discussed with numerical examples. Keywords: uncertainty analysis, bootstrapping , moments, maximum entropy principle. Niepewność jest nieodłącznym elementem procesów projektowania produktu. Dlatego też podejmowanie niezawodnych decyzji wymaga analizy niepewności, która uwzględniałaby wszystkie rodzaje niepewności. W praktyce inżynierskiej, z powodu niepełnej wiedzy, wyznaczenie rozkładu niektórych zmiennych projektowych nie jest możliwe. Co więcej, funkcja stanu granicznego jest wysoce nieliniowa, co sprawia, że do poprawnego obliczenia prawdopodobieństwa uszkodzenia potrzebna jest znajomość momentów wyższych rzędów tej funkcji. W niniejszej pracy zaproponowano metodę analizy niepewności łączącą zasadę maksymalnej entropii z metodą bootstrapową. W pierwszej części pracy wykorzystano metodę bootstrapową do obliczenia przedziałów ufności czterech pierwszych momentów dla zmiennych losowych typu mieszanego oraz zmiennych z próby. Następnie, wyznaczono momenty wyższych rzędów funkcji stanu granicznego przy użyciu metody redukcji wymiarów. Po trzecie, w celu obliczenia funkcji gęstości prawdopodobieństwa (PDF) oraz dystrybuanty (CDF) funkcji stanu granicznego, sformułowano model optymalizacji oparty na zasadzie maksymalnej entropii. Proponowana metoda nie wymaga założenia znajomości rozkładów zmiennych losowych ani obliczania wrażliwości dla funkcji stanu granicznego w odniesieniu do najbardziej prawdopodobnego punktu awarii. W końcowej części artykułu porównano na podstawie przykładów numerycznych wyniki otrzymane za pomocą proponowanej metody oraz symulacji Monte Carlo (MCS). Słowa kluczowe: analiza niepewności, bootstrapping, momenty, zasada maksymalnej entropii. 1. Introduction Uncertainty exists in the whole life-cycle of a product. Therefore, to make reliabile decisions, the representation, quantification, and propagation of uncertainty are needed in design processes, which have been widely studied in many advanced research fields. Uncertainty analysis is to evaluate the cumulative distribution function (CDF), probability density function (PDF) of a 114 performance function formulated by mutually independent random varaibles. The CDF of the performance function can be evaluated with a multidimensional integral. However, in practice it is very difficult or even impossible to obtain an analytical solution to the probability integration. Many research have been develpoed for approximating the probability integration. Mainly, there are three approximation approaches for uncertainty analysis including (1) simulation method, (2) agent models method, and (3) analytical method. The most direct re- Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology liability analysis method is Monte Carlo simulation (MCS) [5, 18, 19]. However, the efficiency of MCS is very low for high dimension problems or when the requirement of design accuracy is high. The main merit of the agent methods [7, 11, 12] is easy to solve. However, the accuracy of the agent methods usually does not meet engineering requirements. Analytical methods focus on simplifying the multi-dimensional integral calculation. The first order reliability method (FORM) and second order reliability method (SORM) [3, 6, 8, 9, 22] were widely used by first order or second order Taylor expansion of the performance function at the most probable failure point (MPP). In the MPP based analysis methods, the random variables were needed to be transformed into standard normal distribution, and the sensitivity analysis was required in both the FORM and SORM. Further, the MPP search was an iterative optimization process, which might be trapped into local optimum. The accuracy of the two methods was determined by the non-linearity of the performance function. When the performance function is highly non-linear, the results calculated with the two methods may cause huge errors. Another altenative analytical method [10, 15, 20] for uncertainty analysis have been developed with dimension reduced method combined numerical integration methods. Rahman and Xu [20] proposed a univariate dimension reduction method for multi-dimensional integration using moment based quadrature rule. Huang and Du [10] presented an uncertainty analysis method based on the combination of dimension reduction integration and saddlepoint approximation. In their method, all the random variable should be transformed into standard normal distribution, and the Gauss-Hermite integration was used to calculate the moments of the limit state functions. Lee and Choi, et al [15] developed an inverse analysis method using MPP based dimension reduction for reliability based design optimization. In their method, the MPP calculation was needed and all the random variables were transformed into standard normal distribution. Maximum entropy principle as a measure of uncertainty has beed developed rencently for uncertainty analysis and reliability based design optimization. As the performance function is highly nonliear or the MPP is not unique, the high order moments of the performace function are needed for estimating the CDF of the performance function accurately. Kang and Kwak [14] applied the maximum entropy principle to reliability based design optimization with the improved moment based quadrature rule. Li and Zhang [16] presented the combined reliability analysis approach with dimension reduction method and maximum entropy principle. The moment based quadrature rule was used to calculate the moments of the performance function. Sung and Kwark [21] proposed reliability bound analysis method based on maximum entropy method with respect to the first truncated moment. Ching and Hsieh [4] developed an estimation method to calculate the confidence interval of the probabilty of failure for the performance function with maximum entropy principle. Volpe and Bagan [23] analyzed the Maximum entropy PDFs and the moment problem of random variables under near-Gauss distribution. A constrained optimization problem is needed to solve in the maximum entropy principle based uncertainty analysis methods. Abramov [1-2] proposed BFGS methods to solve this nonliear optimization problem. In the above related work, the distribution of the random variables were assumed to be known, and were needed to transform into standard normal distribution. In this paper, an uncertainty analysis method combined maximum entropy principle and bootstrapping method is proposed. When the distribution of some random variables can not be exactly determined, the high order moments of limit state functions are estimated by bootstrapping method. Confidence intervals of the probability density function (PDF), and cumulative distribution function (CDF) of performance functions are calculated based on maximum entropy principle. The structure of this paper is as follows. In the second section of this paper, the bootstrapping method to estimate distribution information of uncertainty variables is introduced. The process to calculate moments of limit state functions are provided in the third section. An optimization model based on maximum entropy principle is formulated in the forth section. Numerical examples are analyzed using the proposed method in the fifth section. Finally the conclusions and disscussion are given in the sixth section. 2. Bootstrapping method to estimate the distribution of the uncertainty variables A general uncertainty analysis problem as in Eq. (1) is considered in this work. Performance function y = g (x) which is also referred to limit-state function is modeled as the output of mutually independent random variables x = [ x1, x2 ,, xn ] . { } F ( y) = P y ≤ y a = ∫ g ( x)≤ y a ∫ f (x)dx (1) where F ( y ) is the CDF of the limit state function, y a denotes a upper bound of the performance function, f (x) is the joint probability density funtion of x . Bootstrapping method is a statistical method for estimating the sampling distribution of a random by sampling with replacement from the original samples. The steps of bootstrapping method are analyzed as follows. Given the m sample points xi ,1, xi , 2, ,, xi , m for a random variable xi : Step (1) Construct an empirical probability distribution function f xi from the samples by placing a probability of 1 m for each point xi ,1, xi , 2, ,, xi , m of the samples. Step (2) from the empirical distribution function f xi , draw a random sample of size m with replacement. Step (3) calculate the statistic of the resample points Txi , k . Step (4) repeat step 2 and step 3 k times, where k equals to 1000. Step (5) construct the relative frequency histogram from the k number of Txi by placing a probability of 1 k at each point. Txi ,1 , Txi , 2 ,, Txi ,1000 . Txi ,(1) , Txi ,( 2) ,, Txi ,(1000) denote the bootstrap values by ranking Txi ,1 , Txi , 2 ,, Txi ,1000 from bottom to top. Then the bootstrap percentile confidence interval at 95% level of confidence would be [Txi ,( 25) , Txi ,(975) ] . xi lies within the centered bootstrap 95% percentile confidence interval [2 xi − Txi ,(975) , 2 xi − Txi ,( 25) ] . Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 115 Science and Technology 2.1. Calculation of moments for sample variables Given n samples of a random variable xi , the first four moments µ , σ , µ3 , µ4 of a random variable can be calculated by Eq. (2): E ( xi ) = µ E ( xi − µ ) 2 = E ( x i − µ )3 = E ( xi − µ ) 4 = σ2 n 1 2 n 1 n3 µ3 µ4 + 3(n − 1) n3 (2) σ4 n ∑ xi, j where xi = j =1 n . The centered bootstrap 95% percentile confidence interval of a random varaible xi is calculated by [2 xi − Txi ,(975) , 2 xi − Txi ,( 25) ] . 3. Moments estimation for the limit state functions In engineering practices, the limit state function g(x) is a nonliear function of large input variables x = [ x1, x2 ,, xn ] . The mean of the limit state funtion can be calculated by point estimation method using mn points. The computational burden is extremely large if n becomes large. In order to reduce the computational burden, a dimension reduced method [24] is introduced to approximate the limit state function which is expressed in Eq. (3): g '(x) = ∑ ( gi − g µ ) + g µ (3) where g µ = g ( µ1, µ2 ,, µn ) is the performance function value with all input variables taking the mean values. gi = g ( µ1, µ2 ,, xi ,, µn ) denotes the response value with all input variables taking the mean except the i th input variable. From Eq. (3), the computational burden is reduced largely and the number of the function calls is reached m × n . Since xi is mutually independent, gi is also mutually independent. The first four moments of the limit state g(x) can be calculated by Eq. (4): n µ g = ∑ ( µi − g µ ) + g µ i =1 n σ g2 = ∑ σ i2 µ3 gσ g3 = ∑ µ3iσ i3 116 4. Maximum entropy principle for calculation of CDF and PDF Entropy has been widely studied for uncertainty analysis and reliability design optimization since entropy was analyzed by Jaynes [13] as a measure of uncertainty . Maximum entropy method is developed to estimate the probability distribution of a random variable by maximizing the entropy subject to constraints supplied by the moments of the random variable. Generally, Eq. (5) and Eq. (6) are used to calculate the entropy for both the discrete and continuous variables respectively: n H ( x) = − ∑ pi ln pi (5) i =1 H ( x) = H ( p ( x)) = − ∫ p ( x)ln p ( x) (6) x where pi is the probability of the discrete variable xi, and P(x) is the PDF of the continuous variable xi. 4.1. Optimization formulation to calculate PDF and CDF Maximum entropy formulation of a function can be expressed by Eq. (7): max : H = − ∫ f ( g (x))ln f ( g (x))dg (x) R n i =1 where µi , σ i , µ3i , µ4i are the first four moments of gi which can be calculated with the point estimation method of the single variable. Considering incomplete knowledge of some random variables, the confidence interval of µi [2 µi − Tµi ,975 , 2 µi − Tµi , 25 ] can be calculated by bootstrapping method. i =1 n i =1 n n −1 n i =1 i =1 j >i µ4 gσ g4 = ∑ µ4iσ i4 + 6 ∑ ∑ σ i2σ 2j (4a) (4b) (4c) (4d) s.t. ∫R ∫R g (x) f ( g (x))dg (x) = µ g f ( g (x))dg (x) = 1 ∫R ( g (x) − µ g ) r (7) f ( g (x))dg (x) = µ gr where R is the integral domain, µ g is the mean value of g(x) , th and µ gr is the r central moment for the limit state function g(x) . Lagrange method can be used to solve problem in Eq. (7) and the Lagrange multipliers are denoted as ( λ0 , λ1,, λn ) , and the maximum entropy formulation for the PDF can be expressed in Eq. (8) , which is the optimal solution to Eq. (7): n f ( g (x)) = exp(λ0 + ∑ λi ( g (x) − µ g ) r ) i =1 (8) 4.2. Calculation of the probability of failure for limit state function The steps to calculate probability of failure for limit state functions based on maximum entropy approach can be summered as follows. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology The distributions information of the variables are given in Table 1. The samples of design variable f and s are displayed as follows: f=[0.9598 0.8596 0.8850 0.9389 0.9304 0.9346 0.9751 0.9649 0.9474 0.9725 0.8936 0.9767 0.9730 0.9796 0.9728 0.9511 0.9638 0.8676]. Point estimation method combined bootstrapping method to calculate the moments of uncertainty variables First setp Approximate limit state function s=[215560 215800 224130 215880 218690 219900 220500 226770 226890 212110 211250 219830 217280 214640 214710 222540 218570 216010]. Estimate moments using point estimation According to the method proposed in Section 2, the confidence interval of moments for the limit state function are shown in Table 2. Calculate PDF using the maximum entropy approach Second step Table 2. Confidence interval of moments for the limit state function Calculate CDF First moment Second moment Third moment Fourth moment Calculate probability of failure [2.1317, 2.88]×10-5 [5.376, 7.2734]× ×10-10 [1.3634,1.8446]× ×10-14 [3.4772,4.7044]× ×10-19 Fig. 1. Flowchart of the proposed method (1) The first four moments of random variables are calculated by point estimation method combined bootstrapping method. (2) Estimate moments of the limit state functions where only one random variable is involved, shown in Eq. (2). (3) Estimate moments of the limit state function where n random variables are involved, shown in Eq. (4). (4) Estimate PDF of the limit state functions according to Eq. (7) and Eq. (8). (5) Calculate CDF and probability of failure. The flowchart of the calculation process is shown in Figure 1. The PDF of the limit state function at the lower bound and upper bound can be expressed as Eq. (10) and Eq. (11) according to the maximum entropy approach: flower ( g ) = exp(0.5153 − 1.26 × 106 × g − 3.79 × 107 × g 2 +8.76 × 1015 × g 3 -2.39 × 1020 × g 4 ) fupper ( g ) = exp(8.8779-4.46 × 105 × g − 1.0036 × 109 × g 2 +1.032 × 1016 × g 3 -2.1549 × 1020 × g 4 ) (10) (11) The comparisons for the PDF and CDF of the limit state function from the proposed method and MCS are displayed in Fig. 2 and Fig. 3, respectively. 5. Numerical examples 5.1. Disk edge design The disk edge design problem used in [20] is expressed as in Eq. (9): y = M b = g ( x) = fs 3 × 385.82δ ( N 2π ) 2 ( R 3 60 − R03 )( R − R0 ) (9) where x = [ f , s, δ , N , R, R0 ]T ; f is the material utilization; s is the tensile strength limit; δ is the density; N is the rotor speed; R is the outer radius; and R0 is the inner radius. Table 1. Distributions of random variables Variable Distribution type Parameter 1 Parameter 2 f Sample - - s Sample -lb/in δ Normal 0.28 lb/in3 0.30 lb/in3 N Normal 21,000 rpm 1,000 rpm R Normal 24 in 0.5 in R0 Normal 8 in 0.3 in 2 Fig. 2. PDF of the limit state function for disk edge design 5.2. Fortini’s clutch problem The second example is the over running clutch assembly known as Fortini’s clutch [17]. The contact angle y in is de- Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 117 Science and Technology Table 3. Distribution information for variables Variable Distribution type Mean value[mm] Deviation[mm] Parameters q=r=5.0 x1 Beta 55.29 0.0793 x2 Normal 22.86 0.0043 x3 Normal 22.86 x4 0.0043 Sample The samples of design variable are listed as follow. x 4 = [154.4042,107.4187 115.6844 145.8643 156.3655 156.9087 109.7149 193.6139 158.2305 212.9646 205.6109 383.8824 231.2218 130.8089 110.0401]. termined by the independent random variable, x1, x2 , x3 , x4 as shown in Eq. (12). The distribution of design variables is displayed in table 3. y = arccos [ 2 x1 + ( x2 + x3 ) 2 x4 − ( x2 + x3 )] Fig. 3. CDF of the limit state function for disk edge design (12) The confidence interval of the first four moments for the limit state function are given in Table 4. And the PDF of the limit state function at the bounds are expressed by Eq. (13) and Eq. (14). The comparisons for the PDF and CDF of the limit state function from the proposed method and MCS are displayed in Fig. 4 and Fig. 5, respectively. Fig. 5. CDF of the limit state function for clutch 6. Conclusions In this paper, an uncertainty analysis method with bootsrapping method combined maximum entropy method is proposed. The exact distribution functions of some random variables are not determined using a limited mumber of observations. Therefore, the bootstrapping method is used to estimate the confidence intervals for the stochastic moments of the random variables. Further, the confidence interval of PDF and CDF for the limit state functions are calculated using maximum entropy approach. In the proposed method, neither derivative nor the MPP search are needed. And the random variables are not needed to be transformed into standard normal distribution. The comparison of results form the proposed method with MC method presents the accuracy of the proposed method. Fig. 4. PDF of the limit state function for clutch This research is partially supported by the National Natural Science Foundation of China under the contract number 51075061, and the Research Fund for the Doctoral Program of Higher Education of China (New Faculty) under the contract number 20100185120029. 118 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. Abramov R V. An improved algorithm for the multidimensional momen-constrained maximum entropy problem. Journal of Coputational Physics 2007; 226: 621–644. Abramov R V. The multidimensional moment-constrained maximum entropy problem: a BFGS algorithm with costraint scaling. Journal of Computational Physics 2009; 228: 96–108. Breitung K. Asymptotic approximations for multinomial integrals. Journal of Engineering Mechanics 1984; 110(3): 357–367. Ching J, Hsieh Y H. Local estimation of failure probability function and its confidence interval with maximum entropy principle. Probabilistic Engineering Mechanics 2007; 22: 39–49. Dey A, Mahadevan S. Ductile structural system reliability analysis using adaptive importance sampling. Structural Safety 1998; 20(2): 137–154. Du X P, Chen W. Sequential optimization and reliability assessment method for efficient probabilistic design. Journal of Mechanical Design 2004; 126: 225–233. Haldarand A, Mahadevan S. Reliability Assessment Using Stochastic Finite Element Analysis. New York: John Wiley and Sons, 2000. Hasofer A M, Lind N C. Exact and invariant second-moment code format. Journal of the Engineering Mechanics Division 1974; 100(EM1): 111–121. Hohenbichler M, Gollwitzer S, Kruse W, Rackwitz R. New light on first- and second-order reliability methods. Structural Safety 1987; 4(4): 267–284. Huang B Q, Du X. Uncertainty analysis by dimension reduction integration and saddlepoint approximations. Journal of Mechanical Design 2006; 128: 26–33. Isukapalli S S, Georgopoulos P G. Stochastic response surface methods (SRSMs) for uncertainty propagation: application to environmental and biological system. Risk analysis 1998; 18(3): 351–363. Jin R, Du X, Chen W. The use of metamodeling techniques for design under uncertainty. Structural and Multidisciplinary Optimization 2003; 25(2): 99–116. Jaynes E T. Information theory and statistical mechanics. Physical Review. 1957; 106: 620–630. Kang H Y, Kwak B M. Application of maximum entropy principle for reliability-based design optimization. Structural and Multidisciplinary Optimization 2009; 38: 331–346. Lee I, Choi K K, Du L, Gorsich D. Inverse analysis method using MPP-based dimension reduction for reliability-based design optimization of nonlinear and multi-dimensional systems. Computer Methods in Applied Mechanics 2008; 198: 14–27. Li G, Zhang K. A combined reliability analysis approach with dimension redunction method and maximum entropy method. Structural and Multidisciplinary Optimization 2011; 43: 121–134. Lee S H, Kwak B M. Response surface augmented moment method for efficient reliability analysis. Structure and Safety 2006; 28: 261–272. Moarefzadeh M R, Melchers R E. Directional importance sampling for ill-proportioned spaces. Structural Safety 1999; 21(1): 1–22. Papadrakakis M, Lagaros N D. Reliability-based structural optimization using neural networks and Monte Carlo simulation. Computer Methods in Applied Mechanics and Engineering 2002; 191(32): 3491–3507. Rahman S, Xu H. A univariate dimension-reduction method for multi-dimensional integration in stochastic mechanics. Probabilistic Engineering Mechanics 2004; 19: 393–408. Sung Y H, Kwark B M. Reliability bound based on the maximum entropy principle with respect to the first truncated moment. Journal of Mechanical Science and Technology 2010; 24(9): 1891–1900. Tu J, Choi K K, Park Y H. A new study on reliability based design optimization. Journal of Mechanical Design 1999; 121: 557–564. Volpe E V, Baganoff D. Maximum entropy pdfs and the moment problem under near-Gaussian condiitons. Probabilistic Engineering Mechanics 2003; 18: 17–29. Zhao Y G, Alfredo H S, Ang H M. System reliability assessment by method of moments. Journal of Structural Engineering 2003; 129(10): 1341–1349. Xiao-Ling ZHANG, Ph.D. candidate Prof. Hong-Zhong HUANG, Ph.D. Associate Professor Zhong-Lai WANG, Ph.D. Ning-Cong XIAO, Ph.D. candidate Yan-Feng LI, Ph.D. candidate School of Mechanical, Electronic, and Industrial Engineering University of Electronic Science and Technology of China Chengdu, Sichuan, 611731, P. R. China E-mail: [email protected] Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 119 Article citation info: ZHANG Z, ZHOU Y, SUN Y, MA L. Condition-based maintenance optimisation without a predetermined strategy structure for a two-component series system. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 120–129. Zhisheng ZHANG Yifan ZHOU Yong SUN Lin MA Condition-based Maintenance Optimisation without a Predetermined Strategy Structure for a Two-component Series System Optymalizacja zależnego od stanu technicznego utrzymania urządzeń dla dwuskładnikowego systemu szeregowego nie wymagająca z góry ustalonej struktury strategii Most existing research on maintenance optimisation for multi-component systems only considers the lifetime distribution of the components. When the condition-based maintenance (CBM) strategy is adopted for multi-component systems, the strategy structure becomes complex due to the large number of component states and their combinations. Consequently, some predetermined maintenance strategy structures are often assumed before the maintenance optimisation of a multicomponent system in a CBM context. Developing these predetermined strategy structure needs expert experience and the optimality of these strategies is often not proofed. This paper proposed a maintenance optimisation method that does not require any predetermined strategy structure for a two-component series system. The proposed method is developed based on the semi-Markov decision process (SMDP). A simulation study shows that the proposed method can identify the optimal maintenance strategy adaptively for different maintenance costs and parameters of degradation processes. The optimal maintenance strategy structure is also investigated in the simulation study, which provides reference for further research in maintenance optimisation of multi-component systems. Keywords: semi-Markov decision process, condition-based maintenance, multi-component system. Większość badań nad optymalizacją utrzymania systemów wieloskładnikowych bierze pod uwagę jedynie rozkład czasu życia elementów składowych. Kiedy przyjmie się dla systemów wieloskładnikowych strategię utrzymania urządzeń zależną od ich bieżącego stanu technicznego (condition-based maintenance, CBM), struktura strategii staje się złożona w związku z dużą liczbą stanów składowych oraz ich kombinacji. W konsekwencji, często przyjmuje się pewne z góry ustalone struktury strategii utrzymania przed optymalizacją utrzymania systemu wieloskładnikowego w kontekście CBM. Opracowanie takich z góry ustalonych struktur strategii wymaga jednak specjalistycznego doświadczenia, a i tak brak dowodów na optymalność tych strategii. W artykule zaproponowano metodę optymalizacji utrzymania szeregowego systemu dwuskładnikowego, która nie wymaga wcześniej ustalonej struktury strategii. Proponowaną metodę opracowano na podstawie semimarkowskiego procesu decyzyjnego (SMDP). Badanie symulacyjne pokazało, że za pomocą proponowanej metody można ustalać optymalną strategię utrzymania w sposób adaptacyjny dla różnych kosztów utrzymania oraz parametrów procesów degradacyjnych. Za pomocą symulacji badano także optymalną strukturę strategii utrzymania, jako punkt odniesienia dla przyszłych studiów nad optymalizacją systemów wieloskładnikowych. Słowa kluczowe: semimarkowski proces decyzyjny, condition-based maintenance, system wieloskładnikowy. 1. Introduction Most practical engineering assets are multi-component systems, i.e., they have more than one component. During optimising the maintenance of these multi-component systems, one needs to consider three interactions among components: economic dependence, stochastic dependence, and structural dependence. Economic dependence means that the cost of grouping maintenance can be different from the sum of individual maintenance costs. Stochastic dependence implies that degra- 120 dation processes of different components influent each other. Structural dependence means that a certain group of components are connected together and should be replaced together. The three interactions make the maintenance strategy optimisation of a multi-component system much more complex than that of a mono-component system. Various approaches have been developed to optimise the maintenance strategy of multi-component systems [9]. However, most of these approaches were based on the lifetime distribution of system components [5, 10-12]. Only few papers Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology discussed the maintenance optimisation for multi-component systems in the context of CBM. Van Der Duyn Schouten proposed two types of maintenance strategies for multi-component systems [15]. In that paper, an essential condition of a whole system replacement was that the number of components in the doubtful state exceeded a threshold. Gürler further optimised the threshold of the doubtful state based on the research of Van Der Duyn Schouten [3]. Castanier developed a more flexible maintenance strategy; the state dependent inspection interval was adopted in the research [1]. Based on the research by Castanier, Naini considered both preventive replacement and imperfect preventive maintenance to optimise the maintenance strategy of a two-component system [8]. In that paper, the inspection interval was simplified as state independent. These existing approaches to optimising the CBM strategy of multi-component systems largely predetermined maintenance strategy structures. The optimality of these predetermined structures have not been proofed or discussed. Furthermore, identifying an appropriate predetermined maintenance strategy structure also requires expert knowledge and experience that is not always available in reality. Therefore, a maintenance optimisation method that does not require a predetermined strategy structure is more applicable in reality and can be more cost-effective. This paper proposes a maintenance optimisation approach for multi-component systems without a predetermined strategy structure using the semi-Markov decision process (SMDP). When maintenance strategy optimisation is carried out based on the Markov decision process (MDP) or the SMDP, the optimal maintenance structure can be identified simultaneously with the optimal strategy. Therefore, the MDP and the SMDP are widely used in the maintenance strategy optimisation and the optimal strategy structure investigation of mono-component systems [2, 7, 14, 18]. However, the application of SMDP to multi-component systems is still inadequate. A critical reason is that the health state of a multi-component system is difficult to be expressed, which makes the construction of the relative cost functions for SMDP become challenging. This paper divides the degradation process of a multi-component system into three stages, i.e., normal, partially failed, and completely failed. A SMDP is then developed for the maintenance optimisation of a two-component system. In addition, the optimal maintenance strategy structure of the two-component system under various situations is also investigated. The body of this paper is organised as follows: Section 2 introduces the formulations of the degradation process of a two-component system and the costs of related maintenance activities. After that, a SMDP for the two-component system is developed in Section 3. The performance of the proposed maintenance optimisation method is investigated by simulation studies in Section 4. Section 4 also investigates the structure property of the optimal maintenance strategy for the two-component system. 2. Description of the System 2.1. The Degradation Model A two-component system is investigated in this paper. The degradation processes of both components are assumed to follow the stationary Gamma process that are formulated as . λ u ( t + ∆t ) − λ u ( t ) ~ Ga ( au ⋅ ∆t , ξu ) u = 1, 2 (1) Here, λ u ( t ) denotes a degradation indicator of Component u at time t , and Ga ( au ⋅ ∆t , ξu ) presents the Gamma distribution with the shape parameter au ⋅ ∆t and the scale param- eter ξu . When the process λ u ( t ) exceeds a failure threshold Lu , Component u fails. Component u is in a perfect health state when λ u ( t ) = 0 . The Gamma process is monotonically increasing, which is consisted with the irreversible degradation process of most engineering assets. Therefore, the Gamma process is widely used in degradation modelling [16, 17]. The degradation processes of the two components are assumed to be independent from each other, i.e. the stochastic dependence is not considered in this paper. The two components are assumed to be connected in series, and the whole system suffers from a failure when one of the two components is failed. The failure of the system cannot be detected immediately. However, operating the system in a failure condition will cause an additional cost, and the normal component still degrades even if the system is operating in a failure condition. A practical example of this scenario is a production line that consists of two machines, and each machine produces a certain part of a product. If one machine fails to produce qualified parts, the final product cannot meet the specifications and the production line is considered as failed. However, the failure of the production line may be not detected until an inspection is conducted on the two machines or final products. 2.2. Maintenance Related Costs and Durations In this paper, three types of maintenance activities are considered, i.e., inspection, preventive replacement, and corrective replacement. The inspection is assumed to be able to completely reveal the state of the two components. Each inspection entails a cost Ci . Inspections are scheduled according to the health state of the two components to avoid unnecessary inspections. A preventive replacement action for Component u is conducted at a cost C pu , while the cost of corrective replacement for Component u is Ccu . The preventive replacement cost is lower than the corrective replacement cost, i.e., C pu < Ccu . In this paper, both the preventive and corrective replacement can bring a component to an “as good as new” state ( λ u ( t ) = 0 ). Any preventive replacement or corrective replacement activity brings about a system set-up cost Cs . The set-up cost is caused by the dismantling and the reassembly of the system, or production losses during the system maintenance. The set-up cost is incurred only once for a group of replacement actions performed simultaneously. For example, correctively replacing the whole system costs Cs + Cc1 + Cc 2 . Subsequently, economic dependence exists between the two components if Cs > 0 . Besides the cost of maintenance activities, running the system in a failure state will cause an additional cost cd per unit time. The cost rate cd is assumed to be significant and therefore leaving the system failure after an inspection is not optimal. In this paper, the expected cost incurred by failure and maintenance activities per unit time is adopted as the criterion of maintenance optimisation. The durations of replacement and inspections can be ignored compared to the life time of components. Resources to carrying out inspections and replacement activities are assumed to be always adequate. The minimum reliability and availability constrains are not considered in this research as well. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 121 Science and Technology 3. The Semi-Markov Decision Process Approach 3.1. The Representation of System States and Transitions Different from a mono-component system, the failure of the two-component system can be caused by the failure of one component or the failures of both the two components. The optimal maintenance action and relative costs in the SMDP under the two situations may be different. Consequently, the states of the two-component system are divided into three types, i.e., normal, partially failed, and completely failed. The normal system state implies that both the two components are running in a normal state. The partially failed system state means that one component is failed, while the other component is still in a normal state. In the completely failed situation, both the two components are in a failure state. To apply the SMDP, the continuous degradation process of Component u is discretised into M u different states. The state of Component u at time epoch t is then represented by xtu = 1, 2,, M u u = 1, 2 , where the state xtu = 1 denotes the u “as good as new” state and the state xt = M u stands for the failure state. By combining component states, the system state at time is given by: ( ) ⋅ ( M 2 − 1) + M − 1 M − 1 + x2 xt = ( 1 ) ( 2 ) t M1 ( M 2 − 1) + x1t M1M 2 x1t − 1 xt2 x1t < M1, xt2 < M2 x1t = M1, xt2 < M 2 x1t < M1, xt2 = M 2 (2) x1t = M1, xt2 = M 2 Equation (2) divides the system states into four subsets: when x1t < M1, xt2 < M 2 , the system is in a normal state; when x1t = M1, xt2 < M 2 , the system is partially failed, and the failed 1 2 component is Component one; when xt < M1, xt = M 2 , the system is failed, and the failed component is Component two; 1 2 when xt = M1, xt = M 2 , the system is completely failed. To facilitate the formulation of the SMDP, the state of an individual component given the system state is presented as: xtu = gu ( xt ) u = 1, 2 (3) u After discretisation, the degradation process xt becomes a continuous time discrete state Markov Chain, and the transition matrix during an interval ∆t can be approximated as: (4) u u ( Pu ( ∆t ))ij = piu, j ( ∆t ) = Pr LLuj ≤ λ u (t + ∆t ) ≤ ULuj λ u (t ) = LLi +2 ULi , ULui LLui where, and denote the upper limit and the lower limit of the i th state of Component u , respectively. The degrada- tion indicator before discretisation, i.e., λ u ( t ) , follows the Gamma process as in Equation (1). Consequently, Equation (4) can be calculated according to the property of the Gamma process. Because the two components degrade independently, the transition matrix for the system is obtained as: 122 ( P ( ∆t ) )ij = p1g1 (i ), g1 ( j ) ( ∆t ) ⋅ pg22 (i ), g2 ( j ) ( ∆t ) (5) Similarly, the reliability of the system after ∆t given that the current system state is i can be calculated as: 2 LLu + ULui R ( ∆t i ) = ∏ Pr λ u ( t + ∆t ) ≤ Lu λ u ( t ) = i 2 u =1 ,(6) which can be calculated according to the property of the Gamma process [16]. The expected survival time of the system starting at state i during a time interval ∆t can be then derived as: ∆t (7) τ ( ∆t i ) = ∫ R ( s i ) ds 0 3.2. The Relative Cost Functions The relative cost function that formulates the relative cost of a single step in the long-run decision process is a crucial part of constructing and solving the SMDP [6]. In this paper, the relative cost function is a function of the current system state xt . When the system is in a normal state, i.e., x1t < M1 2 and xt < M 2 , four alternative maintenance activities are available. One is performing an inspection after a certain period of time. The waiting duration till the next inspection depends on the current state of the two system components. The others are preventively replacing Component one, preventively replacing Component two, and conducting a complete system replacement. The relative cost function for a normal system state can be then written as: (8) V ( xt ) = min {VIN ( xt , nI ∆ ID ) ,VPR1 ( xt ) ,VPR 2 ( xt ) ,VPRAll ; nI = 1, 2,, N I } Here, VIN ( xt , nI ∆ ID ) denotes the relative cost of performing an inspection after a period nI ∆ ID when the current system sate is xt, and N I ∆ ID is the maximum waiting time for the next inspection. The notation ∆ ID can be regarded as the minimum time unit of inspection intervals considered in a maintenance strategy. Theoretically, reducing ∆ ID can enhance the accuracy of the optimal strategy. However, in reality, the value of ∆ ID should be selected based on the application. An unpractical short ∆ ID is not beneficial and makes the strategy difficult to implement. For example, when the maintenance strategy of the engine in a locomotive is investigated, ∆ ID can be a week instead of an hour. The function VPRu ( xt ) u = 1, 2 is the relative cost when only Component u is preventively replaced. The variable VPRAll is the relative cost of a complete preventive system replacement. u i ≠u When the system is partially failed ( xt = M u , xt < M i ), there are also two optional strategies. One is replacing the failed component only, and the other is a complete system replacement. The corresponding relative cost function is given by: V ( xt ) = min {VPRu ( xt ) ,VPRAll } − C pu + Ccu u = 1, 2 (9) Because a corrective replacement is performed to Component u , the difference between the costs of a corrective re- Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology placement activity and a preventive replacement activity should be added to Equation (9). 1 2 When the system is failed completely ( xt = M1, xt = M 2 ), the only possible maintenance activity is complete corrective system replacement, and the relative cost function is as follows: V ( xt ) = Cs + Cc1 + Cc 2 + V (1) (10) Here, V (1) denotes the relative cost function starting at the 1 2 “as good as new” system state, i.e., xt = 1 , xt = 1 . In Equation (8), the relative cost of conducting an inspection after a given time interval ∆t starting at system state xt = i is calculated as: VIN ( xt = i, ∆t ) = Ci + M1 M 2 ∑ j =i ( (11) V ( j ) pi , j ( ∆t ) − γ ⋅ ∆t + cd ⋅ ∆t − τ ( ∆t i ) ) , where, γ is the expected cost incurred by failure and maintenance activities per unit time and pi , j ( ∆t ) is an element in the system transition matrix during the time interval ∆t . The relative cost of preventively replacing Component one and two given that the current system state xt is i are given by: VPR1 ( xt = i ) = Cs + C p1 + V ( g 2 ( i ) ) (12) 3.3. The Policy Iteration After the relative cost functions are constructed, the policy iteration is used to find the optimal maintenance policy that minimises the expected cost per unit time. A policy is denoted as δ ( A ) = B , where A = 1, 2,, M1 ⋅ M 2 is a certain discretised system state derived by Equation (2) and B is the corresponding maintenance action. For a normal system state, the maintenance action can be chosen from: B ∈ {( IN , nI ∆ ID ) , PR1, PR2 , PRall ; nI = 1, 2,, N I } . The first candidate maintenance activity ( IN , nI ∆ ID ) implies performing an inspection after a duration nI ∆ ID . The other optional maintenance actions PR1 , PR2 , and PRall denote preventively replacing Component one, preventively replacing Component two, and complete system preventive replacement, respectively. When only Component u is failed, the maintenance action space becomes B ∈ {CRu , CRu + PRi ≠ u } . Here, the maintenance action CRu denotes correctively replacing Component u , while the CRu + PRi ≠ u represents correctively replacing Component u and preventively replacing the other component at the same time. For complete failure, the determinate maintenance action is complete system corrective replacement, i.e., CR1 + CR2 . The general process of the policy iteration is as shown in Table 1. For a more detailed introduction of the policy iteration, readers can refer to [7, 13]. and VPR 2 ( xt = i ) = Cs + C p 2 + V (( g1 (i ) − 1) ⋅ ( M 2 − 1) + 1) (13) respectively. The relative cost for a complete system preventive replacement which is state independent can be calculated as: VPRAll = Cs + C p1 + C p 2 + V (1) . (14) Table 1: The process of policy iteration Step 1: Set an initial policy function The initial policy function is selected by the rule of thumb, and any policy satisfies the conditions discussed at the beginning of Section 3.3 can be adopted as the initial policy. Step 2: Calculate the relative costs {V ( A ) ; A = 2, 3,, M1M 2 − 1} and the expected cost per unit time γ by solving the following system of linear equations that is constructed according to the current maintenance policy δ k ( ⋅) : Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 123 Science and Technology I ( IN , n ∆ ) (δ k ( A ) ) ⋅ VIN ( A, nI ∆ ID ) I ID + I PR (δ k ( A ) ) ⋅ VPR1 ( A ) 1 + I PR (δ k ( A ) ) ⋅ VPR 2 ( A ) + I PR (δ k ( A ) ) ⋅ VPRAll 2 all V ( A ) = I CR (δ k ( A ) ) ⋅ VPR1 ( A ) 1 + I CR1 + PR2 (δ k ( A ) ) ⋅ VPRAll − C p1 + Cc1 I A δ A ⋅V CR2 ( k ( ) ) PR 2 ( ) + I PR1 + CR2 (δ k ( A ) ) ⋅ VPRAll − C p 2 + Cc 2 , 2 ≤ A ≤ ( M1 − 1) ( M 2 − 1) , ( M1 − 1) ( M 2 − 1) < A ≤ M1 ( M 2 − 1) , M1 ( M 2 − 1) < A < M1M 2 , where the formulations of , VPR1 ( A ) , VPR2 ( A ) and are given by Equations (11), (12), (13), and (14) respectively, and is the indicator function given by: 0, x ≠ B (15) IB ( x) = 1, x = B The relative cost functions when the system is brand new and completely failed are determinate, i.e., and V ( M1M 2 ) = Cs + Cc1 + Cc 2 . Step 3: Calculate the relative costs under different maintenance actions: VIN ( A, nI ∆ ID ) , A = 2, 3,, ( M1 − 1) ( M 2 − 1) , nI = 1, 2,, N I , VPR1 ( A ) , A = 2, 3,, M1M 2 − 1 and VPR2 ( A ) , A = 2, 3,, M1M 2 − 1 given by Equations (11), (12), and (13) using the values of {V ( A ) ; A = 2, 3,, M1M 2 − 1} and γ obtained in Step 2. Step 4: Obtain the improved policy function δ k +1 ( ⋅) using the relative costs calculated in Step 3. The δ k +1 ( ⋅) is identified piecewisely as: When the system is in a normal state, i.e., 1 ≤ A ≤ ( M1 − 1) ( M 2 − 1) , the policy function is: ( IN , l ⋅ ∆ ID ) , PR1, δ k +1 ( A ) = PR2 , PR , all VIN ( A, l ⋅ ∆ ID ) = min {VIN ( A, nI ∆ ID ) ,VPR1 ( A ) ,VPR 2 ( A ) ,VPRAll ; nI = 1,, N I } VPR1 ( A ) = min {VIN ( A, nI ∆ ID ) ,VPR1 ( A ) ,VPR 2 ( A ) ,VPRAll ; nI = 1,, N I } VPR 2 ( A ) = min {VIN ( A, nI ∆ ID ) ,VPR1 ( A ) ,VPR 2 ( A ) ,VPRAll ; nI = 1,, N I } VPRAll = min {VIN ( A, nI ∆ ID ) ,VPR1 ( A ) ,VPR 2 ( A ) ,VPRAll ; nI = 1,, N I } . When only Component One is failed, i.e. ( M1 − 1) ( M 2 − 1) < A ≤ M1 ( M 2 − 1) , the policy function is: CR , VPR1 ( A ) < VPRAll δ k +1 ( A ) = 1 CR1 + PR2 , VPR1 ( A ) > VPRAll . When only Component Two is failed, i.e. M1 ( M 2 − 1) < A < M1M 2 , the policy function is: VPR 2 ( A ) < VPRAll CR , δ k +1 ( A ) = 2 CR PR V + , 1 PR 2 ( A ) > VPRAll 2 . When both the two components are failed, i.e. A = M1M 2 , the whole system should be replaced, and the policy function is therefore predetermined as δ k +1 ( A ) = CR1 + CR2 . Step 5: If δ k +1 ( ⋅) = δ k ( ⋅) , the optimal maintenance policy δ * ( ⋅) is obtained as δ k ( ⋅) . Otherwise, go to Step 2 and start a new iteration. 124 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology The most time-consuming part of the policy iteration algorithm in Table 1 is Step 2 that entails solving a system of linear equations with M1M 2 − 1 variables. When the numbers of discretised states (i.e. M1 and M 2 ) are large, some iterative methods (e.g. the Jacobi method and the Gauss–Seidel method) are required to solve the system of linear equations. Fortunately, according to the stimulation study in Section 4.2, the policy iteration can obtain a satisfactory approximate optimal maintenance strategy when the resolution of component state discretisation is moderate. Consequently, the system of linear equations in Step 2 is simply solved based on the LU decomposition. Another potential factor relates to the efficiency of the policy iteration is the number of possible inspection intervals N I . A large N I can reduce the efficiency of Step 3 and Step 4 in Table1. However, the value of does not change the number of variables in the system of linear equations in Step 2 which is the bottle-neck of the whole algorithm. Consequently, the number of optional inspection intervals does not increase the computing time of the policy iteration significantly. 4. The Simulation Study of the Proposed Approach 4.1. Investigation of the Optimal Maintenance Strategy Structures Markov decision process (MDP) has been adopted to explore the maintenance strategy structure property of mono-component systems and multi-component systems based on lifetime distribution (i.e. two-state assumption) [4, 6]. However, these strategy structure properties cannot be simply extended to the CBM of multi-component systems. The structure of the CBM strategy of multi-component systems is much more complex due to the large number of component states and their combinations. To address this research gap, this study investigates structure properties of the CBM strategy of a continuous-state two-component system. The results can provide guidelines for approximate maintenance optimisation algorithms of multicomponent systems in a CBM context. In addition, investigating the strategy structure can also validate the effectiveness of the proposed SMDP approach. Maintenance Strategy Structures for Different Set-up Costs The set-up cost is an important element in the maintenance optimisation of multi-component systems. When the set-up cost covers a considerable proportion of the maintenance cost, significant economic dependence among components exists, and the group maintenance should be adopted. Subsequently, the influence of different set-up costs on maintenance strategy structures were studied first. In this part of simulation study, parameters of system degradation processes and maintenance costs were selected without particular physical meaning, and were for illustrative purpose only. The parameters of the system degradation processes were set as follows: a1 = a2 = 1 , ξ1 = ξ 2 = 1 3 , and L1 = L2 = 2 . The inspection cost and the failure cost per unit time were assumed as Ci = 1 and cd = 10 . The shortest inspection interval was ∆ ID =0.2, and the corresponding N I was selected as 15. As discussed in Section 3.2, the selection of ∆ ID is application-dependent in reality, and an unpractical short ∆ ID is not preferred. The value of N I is initially selected by the rule of thumb, and may be modified according to the maintenance optimisation results. When the longest inspection interval in the obtained optimal strategy is equal to N I ⋅ ∆ ID , a larger N I should be used so that the policy iteration can access a potential optimal policy with a longer inspection interval. First of all, a small set-up cost ( Cs = 1 ) was considered, and costs for preventive and corrective replacement were selected as: C p1 = C p 2 = 39 and Cc1 = Cc 2 = 99 , respectively. After the policy iteration, a minimum average cost per unit time γ = 19.4894 was derived. The result of the policy iteration is presented as the matrix in Figure 1. Each colour standards for a particular type of maintenance action; the numbers in white rectangles are the waiting durations till the next inspection. Because the degradation processes and the maintenance costs of the two components are the same, the policy matrix is symmetrical about the diagonal line. Figure 1 also shows that the optimal maintenance action for a component is not monotonic in State 9. Preventive replacement for the component in State 9 is required, when the state of the other component is below state 6. On the other hand, a further inspection is optimal when the other component is in State 7 and State 8. A complete system replacement is required when both the components are in or above state 9. This unexpected optimal maintenance structure is caused by the economic dependence: When the other component degrades to a state near the preventive replacement threshold, a more economical way is leaving the component in State 9 along and performing complete system replacement later. To demonstrate the effects of this non-monotonic structure, a monotonic strategy in Figure 2 was also adopted, and the average cost per unit time was γ = 19.5157 . Therefore, the non-monotonic strategy in Figure 1 was more cost-effective. Then a significant set-up cost ( Cs = 20 ) was adopted. To maintain the replacement costs for an individual component (i.e., Cs + C pu and Cs + Ccu ) unchanged, the costs for preventive and corrective replacement were selected as C p1 = C p 2 = 20 and Cc1 = Cc 2 = 80 , respectively. After the policy iteration, the minimum average cost per unit time was calculated as γ = 17.3396 and the optimal strategy is presented in Figure 3. Finally a more significant set-up cost ( Cs = 30 ) was used, and the costs for replacement were C p1 = C p 2 = 10 and Cc1 = Cc 2 = 70 . Using the policy iteration, the minimum average cost per unit time was obtained as γ = 15.4537 and the optimal maintenance strategy is shown in Figure 4. Some conclusions can be drawn from the maintenance optimisation results for the three different set-up costs. Firstly, the cost reduction by introducing opportunistic maintenance is more significant when the set-up cost covers a larger proportion of the total replacement cost. Secondly, the non-monotonic part of the strategy and the threshold for opportunistic replacement is near the “as good as new” state for a large set-up cost. Finally, besides opportunistic replacement, complete system replacement is required when the two components are both near but still below the preventive replacement thresholds. More costeffective maintenance strategy structures are expected after the non-monotonic properties that are derived by this simulation study are described appropriately. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 125 Science and Technology Fig. 1. The optimal maintenance strategy when a1 = a2 = 1 , ξ1 = ξ 2 = 1 3 , Cs = 1 , C p1 = C p 2 = 39 , and Cc1 = Cc 2 = 99 Fig. 3. The optimal maintenance strategy when a1 = a2 = 1 , ξ1 = ξ 2 = 1 3 , Cs = 20 , C p1 = C p 2 = 20 , and Cc1 = Cc 2 = 80 126 Fig. 2. The monotonic maintenance strategy when a1 = a2 = 1 , ξ1 = ξ 2 = 1 3 , Cs = 1 , C p1 = C p 2 = 39 , and Cc1 = Cc 2 = 99 Fig. 4. The optimal maintenance strategy when a1 = a2 = 1 , ξ1 = ξ 2 = 1 3 , Cs = 30 , C p1 = C p 2 = 10 , and Cc1 = Cc 2 = 70 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Fig. 5. The optimal maintenance strategy when a1 = a2 = 1 , ξ1 = 0.5 , ξ 2 = 1 6 , Cs = 20 , C p1 = C p 2 = 30 , and Cc1 = Cc 2 = 80 Maintenance Strategy Structures for Different Degradation Process Parameters Different from the methods developed in [3, 15], the approach proposed in this paper can process a system whose components follow different degradation processes. This part of simulation study explores the influence of process parameters on maintenance strategy structures. The parameters for the degradation processes of the two components were a1 = a2 = 1 and ξ1 = 0.5 , ξ 2 = 1 6 . A larger scale parameter ξu u = 1, 2 indicates a faster degradation process. Therefore, Component one degrades more quickly. Two sets of maintenance costs were used: Cs = 20 , C p1 = C p 2 = 30 , Cc1 = Cc 2 = 80 and Cs = 47 , C p1 = C p 2 = 3 , Cc1 = Cc 2 = 53 . The minimum average cost per unit time for the two situations were γ = 21.4651 and γ = 18.5801 , respectively. The corresponding maintenance strategies are showed in Figure 5 and Figure 6. Figure 5 and Figure 6 show that lower preventive and opportunistic thresholds are set for Component one due to the faster degradation process of that component. Consequently, the strategy structures become unsymmetrical about the diagonal line. The difference between Figure 5 and Figure 6 shows that the proposed SMDP can adaptively identify the maintenance strategy structure according to different maintenance costs and degradation process parameters. 4.2. Influence of the Number of Discretised Intervals The system state space is discretised to perform the SMDP, which can introduce errors into the estimate of average cost per unit time. The discretised system state space also leads thresholds for preventive and corrective replacement to be less accurate. Increasing the number of states can reduce the errors that are brought in by discretisation. However, the consumed memory and elapsed time increase quickly with the resolution Fig. 6. The optimal maintenance strategy when a1 = a2 = 1 , ξ1 = 0.5 , ξ 2 = 1 6 , Cs = 47 , C p1 = C p 2 = 3 , and Cc1 = Cc 2 = 53 of the system state space. Therefore, it is necessary to find a balance between the accuracy of a maintenance strategy and the length of computing time. In this part of simulation study, different numbers of discretised component states were trailed to investigate the relationship between the effectiveness of the maintenance strategy and the elapsed time of the policy iteration. The effectiveness of the maintenance strategy was evaluated through the average cost per unit time of a simulated degradation process. The parameters of the degradation processes were selected as and ; the costs of maintenance actions were a1 = a2 = 1 , ξ1 = ξ 2 = 1 3 , and Cs = 30 . To explore the effects of the resolution of the component states, four different numbers of discretised component states were adopted, i.e., M1 = M 2 = 7 , M1 = M 2 = 12 , M1 = M 2 = 22 , and M1 = M 2 = 32 . For the three different resolutions, the policy iteration was carried out and elapsed durations were recorded. The derived maintenance strategies were applied to a simulated degradation process of 106 unit time length. The simulated average costs per unit time were calculated to compare with the approximated results derived by the policy iteration. The results are demonstrated in Figure 7. Figure 7 shows that the approximated average costs are lower than the simulated average costs, and the difference between the two costs reduces with the growth of the number of discretised states. The increase of the approximated average costs is caused by the reduction of errors in the policy iteration, and the decrease of the simulated average cost is due to more accurate thresholds in maintenance strategies. Figure 7 also shows that when , adopting a finer resolution of component states cannot save the simulated average cost significantly, while the elapsed time is considerably longer. The simulated average cost per unit time when M1 = M 2 = 22 and M1 = M 2 = 32 were 15.6524 and 15.6398, respectively. The corresponding elapsed Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 127 Science and Technology Fig. 7. The simulated and approximated average cost per unit time and the elapsed time durations were 106.8 seconds and 597.3 seconds. This small variance between the two simulated average costs shows that the proposed approach is able to identify an approximate global optimal strategy for a continuous state two-component system without a predetermined strategy structure. 5. Conclusions This paper has developed a SMDP approach to optimise the maintenance strategy of a multi-component system without a predetermined strategy structure. The state of the multi-compo- nent system has been divided into three different types: normal, partially failed, and completely failed to construct the relative cost function and perform the policy iteration. Compared with other existing approaches, the proposed SMDP do not need to predetermine the maintenance structure and the number of inspection intervals. Therefore, the SMDP developed in this paper is more adaptive and applicable in reality. Furthermore the SMDP divides the long-term degradation process into single time steps. Consequently, the SMDP approach is easier to be carried out in more complex practical situations, e.g., imperfect maintenance, state-dependent maintenance cost, and statedependent maintenance durations. In addition, the SMDP uses the transition matrix to express the system degradation process. Therefore, the stochastic dependence and the structure dependence can be also processed by the proposed approach when the transition matrix of system states is established. This research has also explored the structure property of the optimal CBM strategy for a two-component system through a simulation study. The results can provide a guideline to develop an approximate optimal maintenance strategy for multi-component systems. The simulation study also shows that the proposed approach using the SMDP can provide satisfactory optimisation results for a continuous state two-component system. For a more complex multi-component system with intractable number of component state combinations, approximate solving methods for the SMDP can be adopted. Acknowledgement: The research work is supported by the Foundation of the Key Laboratory of the Measurement and Control of Complex Engineering Systems, Ministry of Education of China, under contract number 2010B001, and Research Fund for the Doctoral Program of Higher Education of China, under contract number 20110092120007. References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Castanier B, Grall A, Bérenguer C. A condition-based maintenance policy with non-periodic inspections for a two-unit series system. Reliability Engineering & System Safety 2005; 87(1): 109–120. Chen D, Trivedi K S. Optimization for condition-based maintenance with semi-markov decision process. Reliability Engineering & System Safety 2005; 90(1): 25–29. Gürler , Kaya A. A maintenance policy for a system with multi-state components: An approximate solution. Reliability Engineering & System Safety 2002; 76(2): 117–127. Jia Q. A structural property of optimal policies for multi-component maintenance problems. Automation Science and Engineering, IEEE Transactions on 2010; 7(3): 677–680. Laggoune R, Chateauneuf A, Aissani D. Impact of few failure data on the opportunistic replacement policy for multi-component systems. Reliability Engineering & System Safety 2010; 95(2): 108–119. Maillart L M. Maintenance policies for systems with condition monitoring and obvious failures. IIE Transactions 2006; 38: 463–475. Moustafa M S, Maksoud E Y A, Sadek S. Optimal major and minimal maintenance policies for deteriorating systems. Reliability Engineering & System Safety 2004; 83(3): 363–368. Naini S G J, Aryanezhad M B, Jabbarzadeh A, Babaei H. Condition based maintenance for two-component systems with reliability and cost considerations. International Journal of Industrial Engineering & Production Research 2009; 20(3): 107–116. Nicolai R P, Dekker R. Complex system maintenance handbook. London: Springer London, 2008. Pham H, Wang H. Optimal (τ, t) opportunistic maintenance of a k-out-of-n:G system with imperfect pm and partial failure. Naval Research Logistics (NRL) 2000; 47(3): 223–239. Sandve K, Aven T. Cost optimal replacement of monotone, repairable systems. European Journal of Operational Research 1999; 116(2): 235–248. Taghipour S, Banjevic D, Jardine A K S. Periodic inspection optimization model for a complex repairable system. Reliability Engineering & System Safety 2010; 95(9): 944–952 Tijms H C, Van Der Duyn Schouten F A. A markov decision algorithm for optimal inspections and revisions in a maintenance system with partial information. European Journal of Operational Research 1985; 21(2): 245–253. Tomasevicz C L, Asgarpoor S. Optimum maintenance policy using semi-markov decision processes. in Power Symposium, 2006. NAPS 2006. 38th North American. 2006. 128 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology 15 16 17 18 Van Der Duyn Schouten F A, Vanneste S G. Two simple control policies for a multicomponent maintenance system. Operations Research 1993; 41(6): 1125–1136. Van Noortwijk J M. A survey of the application of gamma processes in maintenance. Reliability Engineering & System Safety 2009; 94(1): 2–21. Zhou Y, Ma L, Mathew J, Sun Y, Wolff R. Asset life prediction using multiple degradation indicators and failure events: A continuous state space model approach. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2009; 4: 72–81. Zhou Y, Ma L, Mathew J, Sun Y, Wolff R. Maintenance strategy optimization using a continuous-state partially observable semimarkov decision process. Microelectronics Reliability 2011; 51(2): 300–309. Prof. Zhisheng Zhang Dr. Yifan Zhou School of Mechanical Engineering Southeast University Nanjing, China, 211189 E-mail: [email protected] E-mail: [email protected] Dr. Yong Sun Prof. Lin Ma CRC of Integrated Engineering Asset Management (CIEAM), School of Engineering Systems, Faculty of Built Environment and Engineering, Queensland University of Technology, Brisbane, Australia E-mail: [email protected] E-mail: [email protected] Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 129 Article citation info: WU W, HUANG HZ, WANG ZL, LI YF, PANG Y. Reliability analysis of mechanical vibration component using fuzzy sets theory. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 130–134. Wenjie WU Hong-Zhong HUANG Zhong-Lai WANG Yan-Feng LI Yu PANG RELIABILITY ANALYSIS OF MECHANICAL VIBRATION COMPONENT USING FUZZY SETS THEORY Analiza niezawodnościowa mechanicznego elementu wibracyjnego z wykorzystaniem teorii zbiorów rozmytych The conventional reliability analysis of mechanical vibration component only considers the randomness of vibration but rarely for the fuzziness that may exist. It is therefore difficult to be consistent with the engineering practices. Based on the mechanical vibration theory, a novel fuzzy reliability approach by integrating the fuzzy comprehensive evaluation and fuzzy set theory is proposed in this paper. The fuzzy comprehensive evaluation is used to optimize the fuzzy factors of the reliability analysis of vibration component. With the aim of comparing the performance of the proposed approach with the conventional approach, two engineering examples are presented. The results demonstrate that the proposed approach is better than the conventional approach for its capability of covering fuzzy factors in the engineering problems. Keywords: reliability analysis, mechanical vibration, fuzzy reliability, fuzzy comprehensive evaluation. Tradycyjna analiza niezawodnościowa wibracyjnego elementu mechanicznego bierze pod uwagę jedynie losowość drgań, rzadko zaś wyjaśnia mogącą występować rozmytość. Taka analiza nie odpowiada zatem praktyce inżynierskiej. Opierając się na teorii drgań mechanicznych, w niniejszym artykule zaproponowano nowatorskie podejście w ramach teorii rozmytej niezawodności, które łączy rozmytą ocenę kompleksową oraz teorię zbiorów rozmytych. Rozmytej oceny kompleksowej użyto do optymalizacji rozmytych czynników analizy niezawodnościowej elementu wibracyjnego. W celu porównania efektywności proponowanego podejścia z efektywnością podejścia tradycyjnego przedstawiono dwa przykłady z dziedziny inżynierii. Wyniki pokazują, że proponowane podejście jest lepsze od tradycyjnego ze względu na możliwość objęcia w problemach inżynierskich czynników rozmytych . Słowa kluczowe: analiza niezawodnościowa, drgania mechaniczne, niezawodność rozmyta, rozmyta ocena kompleksowa. 1. Introduction Many component failures of engineering systems are related to vibration [12]. The conventional reliability analysis approach is purely based upon the probabilitistic reliability theory and mechanical vibration theory [3-6]. However, it is assumed that components or systems only have two states, either perfect working or completely failed in the conventional reliability theory [1]. The assumption implies that the state of components or systems can be exactly identified and furthermore there are no intermediate states between these two states. Nevertheless, it is widely observed in the engineering practices that the performance of systems may degrade during their lifetime [8, 14]. On the other hand, it is very difficult or even impossible to collect accurate and sufficient failure data in some real systems when quantifying the reliability characteristics, especially for those systems which consist of new components or components with extremely low failure rates [9, 11]. Many uncertainty factors in mechanical vibration could not be covered only with probability theory. To address the issues, a novel fuzzy reliability analy- 130 sis method by integrating the fuzzy comprehensive evaluation [7] and fuzzy set theory is developed to analyze the reliability of vibration component. The organization of this paper is as follows. In Section 2, the analysis of mechannical vibration is briefly reviewed. The fuzzy comprehensive evaluation is introduced in Section 3. In Section 4, fuzzy reliability analysis of vibration components is presented. Two engineering examples are followed to illustrate the proposed method in Section 5. Conclusions are provided in Section 6. 2. Analysis of mechanical vibration The speed when the resonance occurs is called the critical speed. Critical speed analysis is very important and may be quite complex. Due to the randomness of the load and geometrical shape of components, the critical speed of the vibration component could usually not be expressed by a constant, but a special region with a given probability. Hence, the critical Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology speed has discreteness, which reflects in the special performance. For example, the vibration component has no certain accurate frequency. Amplitude, frequency and phase angle of vibration is not deterministic but random at a given time. In mechanical vibration theory, the machinery has many critical speeds with different orders in nature. When the running speed is close to the first order critical speed, the state is the most dangerous. Hence we often consider the first order critical speed because the effect of the higher orders could be ignorable. It is required that the running speed does not fall into the resonant region which is determined by the experimental data and the natural frequency of component during the process of calculating vibration of mechanical component. Let nc and n denote the critical speed and the running speed respectively. Thus the range of running speed has the following properties [2]. If n < nc , then n < nc (1 − δ1 ) and 0 < δ1 < 0.3 If n > nc , then n > nc (1 + δ 2 ) and 0 < δ 2 < 0.3 where δ1 and δ 2 are fuzzy factors affecting the reliability of vibration component. Assume that the critical speed nc is a random variable with normal distribution, and running speed n is a constant. From the above formulas, in the conventional reliability analysis of mechanical vibration component, when the random variable nc is over n / (1 − δ1 ) or below n / (1 + δ 2 ) , the failure will not occur. When the value of nc falls into the range from n / (1 + δ 2 ) to n / (1 − δ1 ) , the failure will occur. This relationship is illustrated in Fig. 1. In Fig. 1, the dash line represents the characteristic function. CCl λ((nncc)) n1 n2 nc The reliability of mechanical vibration component is the probability that no resonance occurs. In other words, the running speed should be not close to the critical speed. The reliability can be computed by: n1 −∞ R=∫ +∞ n2 f ( x ) dx + ∫ f ( x ) dx (1) where f ( x) is the probability density function of the critical speed nc . When the critical speed follows a normal distribution with the mean value µ and the standard deviation σ , the reliability is expressed by: R = 1 − [Φ ((n2 − µ ) / σ ) − Φ ((n1 − µ ) / σ )] 3. Fuzzy comprehensive evaluation The selection of values of δ1 and δ 2 involves some different kinds of information of systems. However, when there is no available information, it is usually assumed δ1 = δ 2 =0.15 [10]. In other cases, especially involving much information about influence factors, fuzzy comprehensive evaluation can be used to determine the values of δ1 and δ 2 via fuzzy transformation principle. The procedure of evaluation is summarized as follows. 3.1. Determine factor set and evaluation set Let U={u1,u2,...,un} be a set consisting of m influence factors, which represent the attributes of a system. The set is called a factor set. It should be noted that the factors in the factor set often possess fuzziness. Let V={v1,v2,...,vn} be a set consisting of n remarks, which could be obtained by accounting for the lower and upper boundaries of V, and it is called the evaluation set. V could be determined by the boundaries of V and the number of steps. The aim of fuzzy comprehensive evaluation is to select the optimal result from the evaluation set based on the factor set. Obviously, for each vi, there are only two options: belonging to this remark or not. Therefore, the evaluation set is a classic set. 3.2. Constructing fuzzy evaluation matrix Fig. 1. Characteristic function where [Φ ((n2 − µ ) / σ ) − Φ ((n1 − µ ) / σ )] denotes the failure probability of vibration component. From the above analysis, it can be concluded that the values of δ1 and δ 2 are important to reliability analysis of mechanical vibration component because these values directly determine the failure region. Many factors would impact on the selection of the values of δ1 and δ 2 . Hence, factors associated with the selection of the values of δ1 and δ 2 must be considered comprehensively. In this paper, the fuzzy comprehensive evaluation is developed to determine the value of δ1 and δ 2 by considering the fuzziness. (2) First, let Ri be a single-factor fuzzy evaluation set and be expressed as Ri = (ri1, ri 2 ,, rin ) , where rij is the membership degree with respect to the remark vj in terms of the factor ui. Then, let R be a fuzzy evaluation matrix and be expressed as: R1 r11 r12 ... r1n ... r2 n r r R R = 2 = 21 22 ... R m rm1 rm 2 ... rmn where Ri = (ri1, ri 2 ,, rin ) , the ith row in the matrix R , is the single-factor evaluation of the ith factor µi , which is a fuzzy subset on V. (U ,V , R ) is called a comprehensive evaluation model. 3.3. Comprehensive evaluation Let W = ( w1, w2 ,, wm ) be a weight set, which can be determined with the experience of experts or designers. The set denotes the different influence on the evaluation of every factor Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 131 Science and Technology where wi represents the weight of the influence factor ui and n ∑ wi = 1 ( wi ≥ 0 ). where Φ ( β1, β 2 ) = i =1 Based on the single-factor evaluation matrix, the comprehensive effect of every factor with respect to the remark v j can m be denoted by R j = ∑ rij . However, this approach does not In the fuzzy reliability analysis, other types of membership function could also be used to replace the membership function used here. i =1 consider the weight of every factor. Then, when W and R are known, fuzzy comprehensive evaluation set B can be obtained by the fuzzy transformation B = W R = (b1, b2 ,, bm ) . The n operator ‘ ’ is defined by the equation b j = ∑ wi rij here. i =1 However, different models could be obtained depending on other various operation compositions. v = m m j =1 j =1 ∑ b jv j ∑ b j can be used to denote an evaluation result. Then the values of δ1 and δ 2 are equal to the evaluation result. 4. Fuzzy reliability analysis of vibration component The resonant region has a jump in the conventional reliability analysis. However, from the point of practice, there should exist a transition region which can be represented by the fuzzy set. The fuzzy reliability considers the transition region that the conventional approach ignores. Let [n1′ ,n1 ] and [n2 ,n2′ ] denote the transition region of resonance. As shown in Fig. 2, the dash line denotes the membership function of the fuzzy failure. From Fig. 2, the fuzzy reliability of mechanical vibration component is provided by: +∞ µ ( x) f −∞ 1 R = P (nc ≥ n2′ ) + P (nc ≤ n1′ ) = ∫ +∞ µ ( x) f −∞ 2 ( x)dx + ∫ ( x)dx mμl λ((nncc)) 5. Case study 5.1. Fuzzy reliability analysis of a shaft Here a shaft is taken as an example. The manufacturing level and working condition are normal, material quality is good, and importance degree is high. Its running speed is 3000rpm. Then the first order critical speed follows a normal distribution and the mean value and standard deviation are 2640rpm and 145.14rpm respectively. The proposed method will be used to compute the reliability of vibration of the shaft. Let U ={manufacture level, material quality, work condition, importance degree} and V ={0.1,0.11,0.12,0.13,0.14,0.15}. Then, the evaluation matrix can be constructed as follows: R1 0.0 0.5 0.8 1.0 0.5 0.0 R2 = 0.0 0.2 0.4 0.8 1.0 0.5 R = R3 1.0 0.9 0.6 0.4 0.1 0.0 R4 0.0 0.0 0.5 0.6 0.8 1.0 Moreover, the weight set W = (0.26, 0.24, 0.30, 0.20) and B = W R = (b1 , b2 , , b6 ) = (0.3, 0.488, 0.584, 0.692, 0.56, 0.3 32) Therefore, the values of δ1 and δ 2 equal to 0.1257 ( v = 0.1257 by using v = n1 n2 n2' j =1 j =1 ). From Eq. (1), the 3000 / (1 − 0.12) − 2640 3000 / (1 + 0.12) − 2640 ]} ] − Φ[ 145.14 145.14 = 1 − [Φ (5.299) − Φ (0.266)] = 0.606 R = 1 − {Φ[ where µ1 ( x) is the membership function when critical speed (3) nc is greater than n2′ and µ2 ( x) is the membership function when critical speed nc is less than n1′ . Let β1 = n1′ − µ , σ n1 − µ n −µ n′ − µ , β1′ = 2 and β 2′ = 2 , the fuzzy reliσ σ σ ability is rewritten as: R = 1 − [Φ ( β1′ , β 2′ ) − Φ ( β1, β 2 )] R = 1 − (0.9998 − 0.5755) = 0.5755 5.2. Fuzzy reliability analysis of a suspension system nc Fig. 2. Membership function 132 m From Eq. (3), if the values of δ1 and δ 2 is considered, the fuzzy reliability should be given by: n1' m ∑ b jv j ∑ b j conventional reliability can be computed by: β2 = β *β β *β 1 1 {[ β 2Φ ( β 2 ) − β1Φ ( β1 )] + [exp(− 2 2 ) − exp(− 1 1 )]} 2 β 2 − β1 2 2π . (4) Suspension system is a very important part for cars and trains [13]. One function of suspension system is to transform force and moment while the other one is to reduce the vibration from the rude road. Therefore, fuzzy reliability analysis for the suspension system in a train with the proposed method is conducted. The frenquency from the actuator is 1.50 Hz by accounting for the road condition. The first order inherent frequency is assumed to be normally distributed with the mean value 0.98 Hz and standard deviation 0.12 Hz. The proposed method is employed to compute the reliability of a suspension system under vibration. Let U ={design level, manufacture level, installation level, work condition} and V ={0.11,0.12,0.13,0.14,0.15,0.16}. The evaluation matrix could be provided as: Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology R1 0.0 0.0 R R = 2 = R3 0.0 R4 1.0 0.3 0.2 0.0 0.8 0.6 0.4 0.4 0.5 0.8 0. 8 0.7 0.2 1.0 1.0 0.9 0.0 0.8 0.5 1.0 0.0 The weight set is given by W = (0.32, 0.24, 0.22, 0.22) and B = W R = (b1, b2 ,, b6 ) = (0.22, 0.32, 0.486, 0.646, 0.758, 0..596) With v = m m j =1 j =1 ∑ b jv j ∑ b j , we can get δ1 = δ 2 = 0.1405 . The conventional reliability could be obtained with Eq. (2): 1.5 / (1 − 0.12) − 0.98 1.5 / (1 + 0.12) − 0.98 R = 1 − {Φ[ ] − Φ[ ]} 0.12 0.12 = 1 − [Φ (6.0375) − Φ (2.994)] = 0.9986 By comparing the results from the conventional method and the proposed method, a conlusion that the reliability becomes lower with consideration of fuzziness is arrived. 6. Conclusion In this paper, a fuzzy reliability analysis approach of mechanical vibration component is developed. By comparing the proposed approach and the conventional approach, a conclusion that the proposed approach is capable of considering the fuzziness and the reliability of vibration component can be computed in a more comprehensive sense. It should be noted that the proposed approach incorporates the fuzzy comprehensive evaluation into fuzzy reliability theory. From the two engineering examples, it is shown that the proposed approach is more suitable for the complicated reliability analysis because it considers not only aleatory uncertainty but also fuzziness in an integrating framework. If the fuzziness is considered, the fuzzy reliability of the suspension system by using δ1 = δ 2 = 0.1405 could be represented as: R = 0.9974 Acknowledgement: This research is partially supported by the National Natural Science Foundation of China under the contract number 51075061 and Specialized Research Fund for the Doctoral Program of Higher Education of China under the contract number 20090185110019. 7. References 1. An Z W, Huang H Z, Liu Y. A discrete stress-strength interference model based on universal generating function. Reliability Engineering and System Safety 2008; 93(10): 1485–1490. 2. Crandall S H, Mark W D. Random Vibration in Mechanical Systems. Academic Press: New York, 1963. 3. de la Fuent E, Sanz M A, Hernando J L. Exact reliability calculation with quadratic failure indices in random vibration. Engineering Structures 2010; 32(3): 793–799. 4. Hosseini S A A, Khadem S E. Vibration and reliability of a rotating beam with random properties under random excitation. International Journal of Mechanical Sciences 2007; 49(12): 1377–1388. 5. Huang H Z, An Z W. A discrete stress-strength interference model with stress dependent strength. IEEE Transactions on Reliability 2009; 58(1): 118–122. 6. Huang H Z, Qu J, Zuo M J. Genetic-algorithm-based optimal apportionment of reliability and redundancy under multiple objectives. IIE Transactions 2009; 41(4): 287–298. 7. Huang H Z, Li Y H, Xue L H. A comprehensive evaluation model for assessments of grinding machining quality. Key Engineering Materials 2005; 291–292: 157–162. 8. Liu Y, Huang H Z, Levitin G. Reliability and performance assessment for fuzzy multi-state elements. Journal of Risk and Reliability 2008; 222(4): 675–686. 9. Liu Y, Li Y F, Huang H Z, Zuo M J, Sun Z Q. Optimal preventive maintenance policy under fuzzy Bayesian reliability assessment environments. IIE Transactions 2010; 42(10): 734–745. 10. Soong T T, Grigoriu M. Random Vibration of Mechanical and Structural Systems. Prentice Hall: New York, 1993. 11. Wang Z L, Huang H Z, Du L. Reliability analysis on competitive failure processes under fuzzy degradation data. Applied Soft Computing 2011; 11(1): 2964–2973. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 133 Science and Technology 12. Williams T, Ribadeneira X, Billington S, Kurfess T. Rolling element bearing diagnostics in run-to-failure lifetime testing. Mechanical Systems and Signal Processing 2001; 15(5): 979-993. 13. Zhang L M, Dong T J. Analysis of Eigen value sensitivity for a railway vehicle suspension system. Journal of vibration and Shock 2009; 28(1): 28-31. 14. Zuo M J, Jiang R Y, Yam R CM. Approaches for reliability modeling of continuous-state devices. IEEE Transaction on Reliability 1999; 48(1): 9-18. Wenjie WU, Ph.D. candidate Prof. Hong-Zhong HUANG, Ph.D. Associate Prof. Zhong-Lai WANG, Ph.D. Yan-Feng LI, Ph.D. candidate Yu PANG, Ph.D. candidate School of Mechanical, Electronic, and Industrial Engineering University of Electronic Science and Technology of China No. 2006, Xiyuan Avenue, West Hi-Tech Zone, Chengdu, Sichuan, 611731 P. R. China E-mail: [email protected] 134 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Article citation info: GALAR D, GUSTAFSON A, TORMOS B, BERGES L. Maintenance Decision Making based on different types of data fusion. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 135–144. Diego GALAR Anna GUSTAFSON Bernardo TORMOS Luis BERGES Maintenance Decision Making based on different types of data fusion Podejmowanie decyzji eksploatacyjnych w oparciu o fuzję różnego typu danych Over the last decade, system integration is applied more as it allows organizations to streamline business processes. A recent development in the asset engineering management is to leverage the investment already made in process control systems. This allows the operations, maintenance, and process control teams to monitor and determine new alarm level based on the physical condition data of the critical machines. Condition-based maintenance (CBM) is a maintenance philosophy based on this massive data collection, wherein equipment repair or replacement decisions depend on the current and projected future health of the equipment. Since, past research has been dominated by condition monitoring techniques for specific applications; the maintenance community lacks a generic CBM implementation method based on data mining of such vast amount of collected data. The methodology would be relevant across different domains. It is necessary to integrate Condition Monitoring (CM) data with management data from CMMS (Computer Maintenance Management Systems) which contains information, such as: component failures, failure information related data, servicing or repairs, and inventory control and so on. These systems are the core of traditional scheduled maintenance practices and rely on bulk observations from historical data to make modifications to regulated maintenance actions. The most obvious obstacle in the integration of CMMS, process and CM data is the disparate nature of the data types involved, and there have benn several attempts to remedy this problem. Although, there have been many recent efforts to collect and maintain large repositories of these types of data, there have been relatively few studies to identify the ways these to datasets could be related. This paper attempts to fulfill that need by proposing a combined data mining-based methodology for CBM considering CM data and Historical Maintenance Management data. It shows a system integration of physical and management data that also supports business intelligence and data mining where data sets can be combined in non-traditional ways. Keywords: data mining, RUL, data fusion, condition monitoring, CMMS. W ostatniej dekadzie coraz częściej stosuje się integrację systemów, która pozwala przedsiębiorstwom zwiększać wydajność procesów biznesowych. Nowością w zarządzaniu infrastrukturą techniczną jest zwiększanie efektywności już poczynionych inwestycji w systemy kontroli procesów. Pozwala to zespołom do spraw operacyjnych, utrzymania ruchu oraz kontroli procesów monitorować i ustalać nowe poziomy alarmowe na podstawie danych o stanie fizycznym maszyn krytycznych. Utrzymanie urządzeń zależne od ich bieżącego stanu technicznego (condition-based maintenance, CBM) to filozofia utrzymania ruchu opierająca się na tym masowym poborze danych, wedle której decyzje dotyczące naprawy lub wymiany sprzętu zależą od jego obecnego oraz przewidywanego przyszłego stanu technicznego. Ponieważ dotychczasowe badania były zdominowane przez problem technik monitorowania stanu dla konkretnych aplikacji, nie opracowano ogólnej metody wdrażania CBM opartej na eksploracji (data mining ) owych olbrzymich ilości zebranych danych, która miałaby zastosowanie w różnych domenach. Konieczna jest integracja danych z monitorowania stanu (condition monitoring, CM) z danymi dotyczącymi zarządzania pochodzącymi ze skomputeryzowanych systemów zarządzania utrzymaniem ruchu (CMMS), które zawierają informacje na temat uszkodzeń elementów składowych, dane związane z uszkodzeniami, a także informacje dotyczące obsługi lub napraw czy sterowania zapasami. Systemy te stanowią podstawę tradycyjnych praktyk obsługi planowej, a zasadzają się na całościowych obserwacjach dokonywanych na podstawie danych eksploatacyjnych, które pozwalają modyfikować regulowane działania obsługowe. Najbardziej oczywistą przeszkodą w integracji danych CMMS, danych procesowych oraz danych z monitorowania stanu jest rozbieżność ich natury. Dotychczas podjęto jedynie kilka prób rozwiązania tego problemu. Chociaż ostatnio wiele wysiłku włożono w gromadzenie i utrzymanie dużych zasobów tego typu danych, istnieje stosunkowo niewiele badań na temat możliwych sposobów powiązania owych zestawów danych. W prezentowanej pracy poczyniono próbę wypełnienia tej luki proponując metodologię łączoną opartą na eksploracji danych dla celów CBM, która bierze pod uwagę dane z monitorowania stanu i eksploatacyjne dane z zarządzania ruchem. W pracy przedstawiono integrację systemową danych fizycznych i danych z zarządzania, która wspiera także analitykę biznesową (business intelligence) oraz eksplorację danych, gdzie zestawy danych można łączyć w sposób nietradycyjny. Słowa kluczowe: eksploracja danych, pozostały okres użytkowania (RUL), fuzja danych, monitorowanie stanu, CMMS. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 135 Science and Technology 1. Introduction Maintenance can be considered as an information processing system that produces vast amount of data. However data is not synonymous with information; but that data must be processed with data analytical tools to extract the information, [5]. IT (Information Technology) and AI (Artificial Intelligence) tools development support the unprecedented transformation from the industrial age to the information age in maintenance using these existing and emerging technologies that analyze near real-time assets systems data to provide prediction and response maintenance capability. Several technological advances and initiatives at various levels have made a move toward CBM (Condition Based Maintenance) a reality for today’s industry. The transition to CBM requires a collaborative effort on a massive scale and is contingent on identifying and incorporating enhanced and emerging technologies into existing and future production systems. This will require new tools, test equipment, and embedded on-board diagnosis systems. Even more critical, the transition to CBM involves the construction of data-centric, platform-operating capabilities built around carefully developed robust algorithms. This will allow maintenance personnel in the field, support analysts, and engineers the ability to simultaneously, and in real-time, translate conditional data and proactively respond to maintenance needs based on the actual condition Nowadays, two main systems are implemented in most maintenance departments: Computer Maintenance Management Systems (CMMS) are the core of traditional maintenance record-keeping practices and often facilitate the usage of textual descriptions of faults and actions performed on an asset. Second one is condition monitoring systems; recently developed Condition Monitoring Systems (CM) are capable of directly monitoring asset components parameters; however, attempts to link observed CMMS events to CM sensor measurements have been fairly limited in their approach and scalability. A CBM strategy, where the optimal time to schedule a service visit is forecasted based on the condition of the equipment, is often proposed as an answer to the challenge of increase the efficiency and reduces the cost for the service of their equipment over their lifecycle, [10]. However, predictive maintenance approaches are frequently hampered. First, by the lack of knowledge of the features those give a good indication of the condition of the equipment. Second, by the processing power needed for prediction algorithms to forecast the future evolution of the selected features, especially, when large measurements are collected. To overcome these problems, this paper proposes to use data mining to improve the quality of the prognosis. Therefore, Figure 1. Functional model for maintenance documentation system 136 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology the development of future maintenance information systems in order to improve automatic condition monitoring systems enabled by embedded electronics and software in industrial machines, is one of the most important current research problems in this topic. In this paper, the issues and challenge of this necessary integration of data of different nature are presented. It can be argued that understanding the requirements and constraints in conjunction - from maintenance AI and IT perspectives - is necessary to provide different decisions for different end users. 2. Maintenance historical data 2.1. Existing data in maintenance function These plant inventory units are the target for maintenance actions. These actions are basically of two kinds according to [3]: • Those being done to correct an item after it has failed (corrective maintenance). It is required that for recording the reliability of an item, as a minimum corrective maintenance to correct a failure shall be recorded. • Those being done to prevent an item from failing (preventive maintenance). A part of this may only be checks. Recording actual preventive maintenance (PM) is recommended to be done, essentially in the same way as for corrective actions. This may give additional information as follows: -- the full lifetime story of an item (all failures and maintenance); -- the total resources used on maintenance (man-hours, spare parts); -- the total down time and hence, total equipment availability, both technical and operational; -- the balance between preventive and predictive maintenance (inspections, tests) to verify the condition of the equipment to decide if any preventive maintenance is required or not. Figure 2 shows the main types of maintenance actions being commonly performed. According to [2] a new maintenance type called “opportunity maintenance” can be included if one considers maintenance of an item that is deferred or advanced in time when an unplanned opportunity becomes available. Maintenance actions are the result of implemented maintenance program. Choice of applied methodology, ratio preventive-corrective, etc. is always up to maintenance manager who has plant inventory and maintenance information base to Maintenance documentation system, for recording and conveying information, is an essential operational requirement for all the elements of the maintenance management process. Maintenance documentation can be defined according to [11] as: Any record, catalog, manual, drawing or computer file containing information that might be required to facilitate maintenance work. Simultaneously, a maintenance information system can be defined as: The formal mechanism for collecting, storing, analyzing, interrogating and reporting maintenance information. The way in which a maintenance documentation system generally functions is shown in Figure 1, a model which has evolved, according to [12], over a number of years through extensive studies of both paper-based and computerized systems, and which therefore illustrates the principal features of both types – features which, inevitably, they have in common. The system can be considered to be made up of the following interrelated modules: 1. Plant inventory, 2. Maintenance information base, Maintenance Before failure After failure categories 3. Maintenance schedule, 4. Condition monitoring, 5. Maintenance control. The plant inventory (1) is a coded list of the plant units. Opportunity This is the main way into the system. The asset data shall be Corrective Preventive maintenance maintenance maintenance collected in an organized and structured way. The major data categories for equipment are the following: 1. Classification data, e.g. industry, plant, location, system; 2. Equipment attributes, e.g. manufacturCondition Periodic Testing inspection monitoring er’s data, design characteristics; 3. Operation data, e.g. operating mode, operating power, environment. Figure 2. Maintenance actions categorisation These data categories shall be general for all equipment classes. Additionally some data construct the model to be used. The maintenance information specific for each equipment class (e.g. number of stages for a base (2) is a database of maintenance information, e.g. unit life compressor) is needed. Finally, the classification of equipment plans, job catalog, etc. for each of the units. These data are charinto technical, operational, safety related and environmental acterised by: parameters is the basis for the collection of assets data due to • identification data; e.g. maintenance record number, rethe different nature of devices (safety instrumented systems, lated failure and/or equipment record; productive assets, maintenance tools, condition monitoring • maintenance data; parameters characterising a maintesystems etc.). This information is also necessary to determine if nance, e.g. date of maintenance, maintenance category, the data are suitable or valid for various applications. There are maintenance activity, impact of maintenance, items some data that are common to all equipment classes and some maintained; data that are specific for each equipment class. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 137 Science and Technology Table 1. Maintenance records recommended meeting most International Standards and general recommendations [9] •maintenance resources; maintenance man-hours per discipline and total, utility equipment /resources applied; •maintenance times; active maintenance time, down time. A common report for all equipment classes should be used for reporting maintenance data. The data required are shown in Table 1. For some equipment classes minor adaptations may be needed. The minimum data needed to meet the objectives of International Standards, Maintenance Association and CMMS manufacturer’s recommendations are identified in Table 1. Recording maintenance actions is crucial for a successful further knowledge extraction that is why all actions performed should be recorded. PM records are mainly useful for the maintenance engineer, but will also be useful for the maintenance engineer wanting to record -or estimate- the availability of equipment, and doing lifetime analysis not only taking failures into account, but also maintenance actions intended to restore the item to "as-good-as-new" condition. PMs are often performed on a higher indenture level (e.g. "package" level); hence there may not be data available that can be related to the items on the lower indenture level. This restriction must be considered when defining, reporting and analysing PM data. During the execution of PM actions, impending failures may be discovered and corrected as part of the PM activities. In this case the failure(s) shall be recorded as any other failure with the subsequent corrective action done even though it initially was considered to be a PM type activity. The failure detection method shall in this case be referred to as the type of PM being done. It is, however, realised that some failures, generally of minor character, may be corrected as part of the PM and not recorded specifically. The practice on this may vary between companies and should be addressed by the data collector(s) in order to reveal the possible type and amount of failures being included within the PM program. A final option is to record the planned PM program as well. In this case it is possible to additionally record the differences between the planned and the actual performed PM (backlog), [4]. An increasing backlog will be an indication that the control 138 of the conditions of the plant is being jeopardised and may in adverse circumstances lead to equipment damage, pollution or personnel injury. Regarding corrective maintenance, failure records are especially relevant for further knowledge extraction so failure data have to be recorded in a proper way to be suitable for further computation. A uniform definition of failure and a method of classifying failures are essential when data from different sources (plants and operators) need to be combined in a common maintenance database. These failure data are characterised by: • identification data; e.g. failure record number and related equipment that has failed; • failure data for characterizing a failure, e.g. failure date, items failed, failure impact, failure mode, failure cause, failure detection method, so on. The type of failure and maintenance data shall normally be common for all equipment classes, with exceptions where specific data types need to be collected. Corrective maintenance events shall be recorded in order to describe the corrective action following a failure. Finally the combination of plant inventory and maintenance base information produces the expected maintenance schedule. This schedule is a mixture of available techniques to fulfill stakeholder’s constraints and achieve company goals. This mixture is usually composed by some scheduled maintenance and condition monitoring to perform condition based maintenance. The maintenance schedule (3) is a schedule of the preventive maintenance jobs (over a year and longer) listed against each of the units in the life plans. The condition monitoring schedule (4) is a schedule of the condition monitoring tasks, e.g. vibration monitoring listed against each of the units in the life plans. Preventive maintenance records are required to retain the complete lifetime history of an equipment unit. The system has to plan and schedule preventive jobs (arising from the maintenance schedule), corrective jobs (of all priorities) and where necessary modification jobs. The jobs are carried out by trade-force via hard copy or electronic work or- Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology ders. Information coming back on the work orders (and other documents) is used to update the planning systems and provides information for maintenance control. The maintenance control system (5) uses information coming from a number of sources, work orders, stores, shift record, etc. to provide various reports for cost control, plant reliability control, etc. One of the main issues is the integration of these data with the rest of company records such us health and safety, finances, etc.. Up until about 10 years ago most CMMS were stand alone, i.e. they had no electronic linkage with other company software. The most recent computerized maintenance systems are integrated electronically (they are in the same database) with stores, purchasing, invoicing, company costing, payroll and also can have electronic links to project management and condition monitoring software. 2.2. The search of a comprehensive data format Each mentioned data becomes a database record, e.g. a failure event, shall be identified in the database by a number of attributes. Each attribute describes one piece of information, e.g. the failure mode. It is recommended that each piece of information be coded where possible. The advantages of this approach versus free text are: •facilitation of queries and analysis of data; •ease of data input; •consistency check undertaken at input, by having predefined code-lists; •minimise database size and response time of queries. The range of pre-defined codes should be optimised. A short range of codes will be too general to be useful. A long range of codes will give a more precise description, but will slow the input process and may not be used fully by the data collector. Selected codes shall, if possible, be mutually exclusive. The disadvantage of a pre-defined list of codes versus free text is that some detailed information may be lost. It is recommended that free text is included to provide supplementary information. A free-text field with additional information is also useful for quality control of data. This free text box is extremely risky in further data mining process due to difficulties of text recognition and interpretation, see Table 1. Different employees have different skills to describe failures, events and actions and expert systems are not so good to distinguish all these variations. For all mentioned categories, it is recommended to include some additional free text giving more explanatory information as available and deemed relevant, e.g. include a more verbal description of the occurrence leading to a failure event. This would assist in quality checking the information and browsing through single records to extract more detailed information. However users should be aware of the existing risk in automatic processing of these records. 2.3 Database structure The data collected shall be organised and linked in a database to provide easy access for updates, queries and analysis. Several commercial databases are available that can be used as main building block for designing a reliability database. Two aspects on organising the structure of data shall be addressed as follows: • Logical structure: This defines the logical link between the main data categories in the database. This model represents an application-oriented view of the database. The example in Figure 3 shows a hierarchical structure with failure and maintenance records linked to the classification/equipment description (inventory). Records describing preventive maintenance (PM) are linked to the inventory description in a many-to-one relation. The same applies for failures, which additionally have related corrective maintenance records linked to each failure. Each record (e.g. failure) may consist of several attributes (e.g. failure date, failure mode, etc.). • Database architecture: This defines the design of the database as to how the individual data-elements are linked and addressed. Four model categories are commonly available ranked in order of complexity and versatility: -- Hierarchical model: Data fields within records are related by a ‘family tree-like’ relationship. Each level represents a particular attribute of data; -- Network model: This is similar to the hierarchical model; however, each attribute can have more than one parent; -- Relational model: The model is constructed from tables of data elements, which are called relations. No access path is being defined beforehand; all manipula- Figure 3. Logical data structure Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 139 Science and Technology tion of data in tabular form is possible. The majority of database designs use this concept; -- Object model: Software is considered as a collection of objects that each has a structure and an interface. The structure is fixed within each object while the interface is the visible part that provides the link address between the objects. Object modelling enables the database design to be very flexible, extendable, reusable and easy to maintain. This model seems to be the most popular in new database concepts. 3. Condition Monitoring data and automatic asset data collection Condition monitoring involves comparing on-line or offline data with expected values; if necessary, it should be able to generate alerts based on preset operational limits. Health assessment determines if the health of the monitored component or system has degraded, and conducts fault diagnostics. The primary tasks of prognostics involve calculating the future health and estimating the remaining useful life (RUL). In reality, however reliable and effective CBM faces some challenges. First, initiating CBM is costly. Often the cost of instrumentation can be quite large, especially if the goal is to monitor equipment that is already installed. It is therefore important to decide whether the equipment is important enough. Implementing condition-based maintenance requires the setting of an information system to meet the basic requirements of: • Collection and processing of large quantity of information not previously available, regarding the condition of each part of a machine. •Initiate corrective maintenance actions within the leadtime (the period of time between the off-limits condition and an emergency shutdown). In this respect there may be two different situations which the examiner may encounter : -- The condition of machine is not yet close to breakdown. In this case the normal procedure through the maintenance planning section will be followed. -- The condition of machine is already well within the lead-time (near to breakdown). In this situation the information must be directly passed on to the maintenance supervision for carrying out emergency corrective maintenance actions. In order to operate the condition based maintenance program correctly, the maintenance personnel should introduce into the system: •Condition of machine, •Part of machine probably defective, •Probable defect, •Time during which failure must be repaired. By scrutinizing and correlating of diagnosis against actual findings during repair work, it will be possible: •To control the examiner training, •To improve the correlation between parameters chosen for condition measurement and actual defects found, •To obtain severity curves specific to each machine. Making the potential of condition monitoring a reality requires that large amounts of data be collected, monitored, filtered and turned into actionable information. The cheaper and more ubiquitous the computerized monitoring hardware be- 140 comes, the greater the volume of data and the more challenging it becomes to manage and interpret. The vast amount of diagnostic data produced by today’s smart field devices can be a very important source for accurate documentation of maintenance activities. But the sheer volume and complexity of such information can be daunting and difficult for maintenance personnel to manage. What’s needed is an effective means of compiling and organizing the data for day-today utilization by your staff, while preserving and recording significant events for future reference. Data is becoming more and more available. However, in most cases, this data may not be used due to its bad quality, or even properly stored for several reasons, [13]: • Project managers do not have sufficient time to analyse the computerised data so they don’t care about proper storage; • The complexity of the data analysis process is beyond the capabilities of the relatively simple maintenance systems commonly used; • There has been no well defined automated mechanism to extract, pre-process and analyse the data and summarise the results so stored data are not reliable. Maintenance personnel, not only cope with large amounts of field-generated data, they turn that information to their advantage in a number of ways. Real Time Condition Monitoring (RTCM) systems produce lots of warnings, alarms and reports that can be used by maintenance people for many purposes. In this way, the most important issues are identified and handled quickly. Ultimate goal is to fully integrate RTCM data with CMMS to generate work orders as needed. That will provide true automation from the time a field device begins to show signs of reduced performance until a work order is printed out in the maintenance department and a technician is dispatched to the scene. In Figure 4, this automation of work order dispatching is shown. This level of integration of CMMS and CM is feasible due to IT evolution. With the development of open communication protocols, the information accumulated by smart field devices can be captured by asset management software. It’s no longer necessary for technicians to carry handheld communicators or laptops into the plant to evaluate the condition of instruments, some of which are quite inaccessible or in hazardous areas, to be followed by manually documenting test results and current device status. Current applications compile databases of every smart instrument used for process control, including its design parameters, original configuration, maintenance history and current operating condition. With these online tools, technicians can obtain up-to-date information on any device and they never have to make manual entries back into a system. Every event is recognized and recorded, whether initiated by a technician or caused by an external force such as an equipment breakdown or power failure. This process produces one immediate result for shop floor level because work orders can be open and closed helped by devices that collect automatically information and send warning if something wrong happens. Users can refer to recorded alerts to identify any devices that have been problematic over time and what corrective steps may have been taken previously. Automated documentation provides a seamless record of events in a given production area, including communication failures, device malfunctions and process variables Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Figure 4.Two step integration of RTCM and CMMS databases that are out of range. Armed with this information, maintenance personnel are better equipped to understand and resolve nagging repetitive issues to improve the process. If there is an issue, or if maintenance personnel are experiencing a rash of issues, they can go back into the records and get a sense of what’s been going on over time. They can search by a specific device or by location. Since all records are date and time stamped, users can easily determine when and by whom a particular device was changed or tested, including “as found/as left” notations. With this information in a database that cannot be edited, it should never be necessary for technicians to spend time searching for historical information on a device. Since events can also be recorded manually, users can document unusual occurrences affecting the entire plant, such as a lightning strike or power outage, or individual events like device inspections. This decision level is extremely useful for technicians to take immediate actions. However vast amount of available information can produce new knowledge if it’s exploited with proper AI tools due to real physical integration in same database types and locations. Modern CMMS information is stored in very large relational, or tabular, databases. This format is appropriate for an integration investigation since there are a large number of software tools available to query and investigate the tables. For the historical analysis, only certain fields are required, thus allowing for the previously mentioned sensitive data to be removed or filtered. The data subset still contains a full history of component faults and related actions, providing a comprehensive maintenance history profile while alleviating security concerns. Importing CM data into this relational database is somewhat more challenging but possible, since each type of sensor generates different data classes, sampling rates, and number of compiled indicators. Furthermore, each manufacturer stores the collected information in unique proprietary formats, requiring platform-specific importation software to be written. However most CM software allows the CM data to be exported from the original interface so that it can be expanded and generalized. Although both the CMMS and the CM data now co-exist within a single database where it can be queried and explored, automating the discovery of linked events requires additional processing. Relating a given maintenance fault or action, which is textual, to sensor data, which is some arbitrary data class type, can only be accomplished through the compilation of overlapping metadata, [14]. The fields which are generated characterize the location and significance of events, creating a quantified set of parameters by which the disparate data can be compared. Metadata for CM records is generated differently depending on the data class involved. One-dimensional and dimensionless quantities can be assigned rarity parameters through statistical distribution analysis, and higher dimensional data requires using neural networks to identify anomalies. Determining rarity is often accomplished through simple single variable statistical analysis, while severity is typically derived from developers recommended threshold values. More complex domain types require more advanced, though typically wellunderstood analyses such as neural networks which can isolate anomalous points from multidimensional data. It is predicted that through the integration process, more advanced metrics and indicators can be discovered which implement previously unexplored relationships in the data, such as multi-parameter trending. This new discovered knowledge can help maintenance personnel to find out the Remaining Useful Life of the system in order to schedule operation and maintenance processes in function of such relevant information. This information affects replacement of assets, shutdown of the plant, overhauls etc.; so it constitutes the second decision level displayed in Figure 4 which is strongly related to business goals and useless for immediate interventions. 4. Data mining of maintenance historical data for RUL calculation Condition based maintenance (CBM) is the real target of all maintenance plans to optimize inspections and intervention avoiding wastes of time and money. It has evident benefits, including reducing maintenance cost, increasing machine reliability and operation safety, and improving time and resources management [1]. That is why major improvements can be Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 141 Science and Technology achieved in maintenance cost, unscheduled machine failures, repair downtime, spare parts inventory, and both direct and indirect overtime premiums, by using CBM. Although these benefits are well illustrated, two major problems hamper the implementation of predictive maintenance in industrial applications. First, the lack of knowledge about the right features to be monitored and second, the required processing power for predicting the future evolution of features, which is often not available. Data Mining (also known as Knowledge Discovery in Databases, or KDD) has been defined as "the nontrivial extraction of implicit, previously unknown, and potentially useful information, from data" [6]. Data mining and knowledge discovery can be applied to historical data from the field in order to optimally identify these relevant features for the condition of the equipment and the associated thresholds and contexts. Based on this information, a prediction model is fitted to the live data of the equipment, collected from customer’s premises, for predicting the future evolution Figure 5. Data mining steps with maintenance data of these features and forecasting the RUL (remaining useful life) and consequently the tures to reach the thresholds. This remaining life time time interval to the next maintenance action. To overcome the can be used to establish an optimal predictive maintelimited processing power of the machine’s processor and data nance. replications costs derived from the huge amount of collected Logical data mining process to optimally forecast the preand stored data, the concept of cloud computing comes up, dictive maintenance actions is as follows: starting point is the performing prediction through computation of remote located available historical database of machines running whose maindatabases. tenance performed actions can be altered in function of the reThe lack of knowledge for the proper extraction of right feasult of decision support system regarding scheduling process. tures can be compensated with data mining techniques proved In general, such a database exists for almost all machine manuto be useful for relevant features extraction. It has been proved facturers, but only a limited amount of information is currently in [8], [15] and [16] that application of data mining techniques used. to condition monitoring data is very useful for extracting relThe next step consists of data preparation, including data evant features which can be used as parameters for machine transformation from the original to unified data formats and diagnosis and/or prognostics. However, in many other indusdata cleaning such as removing outliers or calculating missing trial applications, no clear physical understanding of the procvalues. Note that the data preparation step may be time consumess is available and therefore to retrieve these relevant features, ing since the unified data format should be compatible with the a clear methodology is required. data mining software tool and still retain a physical interpretaThis paper proposes the use of data mining for CBM option of the data. timization in order to perform an accurate predictive mainteThe real data mining step consists of the data reduction nance scheduling. So far, prognostic has been tackled as an where significant information is found and non relevant inforindependent step from data mining, assuming the prediction of mation is discarded to reduce significantly the number of ata completely known parameter [7]. On the contrary, in a real intributes in the searching process. After this pruning process, the tegration of data mining and CBM approach, prognostics is an relevant features are extracted out of the selected. integral part of the flowchart and makes use of the relevant feaThe final step consists of prognostics, itself divided into tures extracted in data mining step. This approach enables the two sub steps. First, reliability estimation methods are applied possibility to compare different features evolution and combine to historical database in order to identify the optimal thresholds them to improve the accuracy of remaining useful life forecast. of the selected features. Secondly, a prediction algorithm is apThis approach consists of: plied to live data in order to forecast the time to schedule the •A data mining step on historical data where data prepaoptimal CBM. The different steps of the data mining approach ration, data reduction and relevant features extraction are summarized in Figure 5. are performed. •A prognostics step, where optimal thresholds are retrieved using prediction algorithms are applied on live data to estimate the remaining time for the relevant fea- 142 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology 5. Conclusion Main benefit from maintenance data integration process is insight into the future establishment of CM and CMMS data format standards. Early integration attempts will identify data structures that are most conducive and useful for long-term storage and searching. With the coordination of CMMS and CM developers, a general set of guidelines for file formats can be established which will enhance the potential for research by the scientific community, greatly increasing the usefulness of CMMS and CM platforms. Following the implementation of an integrated system, the usefulness of CM devices will expand from mere guidance tools to automated diagnostics and prognostics systems. These integrated systems will constantly compare sensor readings to the wealth of historical records and forecast likely maintenance events based upon historical precedence. This will allow for more efficient logistics and component performance evaluation. Furthermore, these systems will identify unexplained or common modes of failure directing the efforts of scientific component testing, the results of which will drive design modifications making the assets increasingly more reliable. This integration has two steps. First one is integration of technology and standards like MIMOSA are actively contributing to the development of a common hardware and software platform for data storage. Second step is related to the new knowledge extraction as a result of integration performed in first step. For this purpose data mining comes up as a very effective maintenance knowledge tool. Data mining has become useful over the past decade in maintenance to gain more information, to have a better understanding of the behaviour of running assets, and to find optimal maintenance policies derived from this new knowledge. Today, data mining is no longer thought of as a set of stand-alone techniques, far from the maintenance applications. Enterprises require more and more integration of data mining technology with relational CMMS and CM databases and their businessoriented applications. To support this move to applications, data mining products are shifting from stand-alone, technology to being integrated in the relational databases. The goal is to find new and interesting patterns in the data and somehow use the gained knowledge in maintenance decisions. The integration of mining results into the operational business is usually done in an ad-hoc manner. With the integration of mining, the focus shifts toward the deployment of mining in business applications. The audience includes the end users in the line of business who need an easy-to-use interface to the mining results and a tight integration with the existing environment. Simple predictive models, anticipation of assets behaviour, and automatic optimization of maintenance processes by means of data mining become more important than some general knowledge discovery. If the purpose of the model is to increase knowledge of the data, the knowledge gained needs to be organized and presented in a way that the end user, maintenance personnel can use it. Depending on the maintenance requirements, the deployment phase can be as simple as generating a report or as complex as implementing a repeatable data mining process. However, even if the data analyst does not carry out the deployment effort, the maintenance personnel, i.e. end user, must understand up front what actions need to be carried out to actually use the created models. A right deployment of the model and understanding by maintenance people are necessary to be sure the real benefits are harvested by the company through right maintenance decisions. The final product should manifest itself as an automated maintenance exploration interface. Users should be able to quickly identify possible diagnoses of faults and quickly retrieve historical maintenance actions that were effective in resolving the problem. Such a system would be easily scalable allowing for maintainers to have information on a variety of practices being performed across the field. In addition to the assurance that these systems deliver top performance, benefits of this integration include increased plant availability and lower maintenance costs because most faults are caught before they can evolve into problems requiring major repairs and/or costly process interruptions and downtime. Acknowledgement: The author B. Tormos wish to thank “Programa de Apoyo a la Investigación y Desarrollo (PAID-00-11) de la Universitat Politècnica de València” for supporting his research. References 1. 2. 3. 4. 5. 6. 7. 8. 9. Blair J, Shirkhodaie A. Diagnosis and prognosis of bearings using data mining and numerical visualization techniques. Proceedings of the 33rd Southeastern Symposium on System theory, 2001: 395–399 BRITISH STANDARD, BS 4778 (1991): Glossary of terms used in quality assurance (including reliability and maintainability). British Standards Institution, London IEC (1990). 60050 (191): Dependability and quality of service. International Electrotechnical Commission, Geneva, Switzerland. EN 15341: 2007, Maintenance Key Performance Indicators. European Committee for Standardization (CEN), Brussels, Belgium. Eriksson L, Johansson E, Kettaneh-Wold N, Wold S. Introduction to Multi- and Megavariate Data Analysis using Projection Methods (PCA & PLS). UMETRICS, Sweden, 1999. Frawley W, Piatetsky-Shapiro G, Matheus C. Knowledge discovery in databases: An overview. AI Magazine 1992; 13: 57–70 Goh K M, Tjahjono T, Subramaniam S. A review of research in manufacturing prognostics. 2006 IEEE International Conference on Industrial Informatics. pp.417–422. Isermann R. Fault-diagnosis systems. Model-Based Condition Monitoring: Actuators, Drives, Machinery, Plants, Sensors and Fault-tolerant Systems. New York, USA: Springer, 2006. ISO 14224 (2006) Petroleum and natural gas industries – Collection and exchange of reliability and maintenance data for equipment. International Organization for Standardization, Geneva, Switzerland. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 143 Science and Technology 10. Jardine AKS, Lin D, Banjevic D A. Review on machinery diagnostics and prognostics implementing condition-based maintenance. Mechanical Systems and Signal Processing 2006; 20:1483–510. 11. Kelly A. Maintenance Systems and Documentation. New York, USA: Butterworth-Heinemann, 2006. 12. Mather D. CMMS: a time saving implementation process. New York. USA: CRC Press, 2002. 13. Soibelman L, Kim H. Data preparation process for construction knowledge generation through knowledge discovery in databases. Journal of Computing in Civil Engineering 2002; 16 (1): 39–48. 14. Tianhao W, Khan F M, Fisher T A, Shuler L A, Pottenger W M. Posting Act Tagging Using Transformation-Based Learning. ICDM'02 1992 Proceedings of the Workshop on Foundations of Data Mining and Discovery, IEEE International Conference on Data Mining. 15. Wylie R, Mottola L, Kresta J, Monk R. Lessons Learned for the I02 Project. COM 2002 The Conference of Metalurgists (The International Symposium on Knowledge Management in the Metals Industry). Montréal, Québec, Canada. 16. Yang C, Létourneau S. Learning to Predict Train Wheel Failures. KDD 2005 Proceedings of the 11th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Chicago, Illinois, USA. Prof. Diego Galar Anna Gustafson, MSc. Division of Operation and Maintenance Engineering Luleå University of Technology Luleå, 97187 Sweden E-mail: [email protected] [email protected] Prof. Bernardo Tormos CMT- Motores Termicos Uiversitat Politècnica de València Valencia, Spain E-mail: [email protected] Prof. Luis Berges Department Design Engineering and Manufacturing University of Zaragoza Zaragoza, Spain E-mail: [email protected] 144 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Article citation info: JEDLIŃSKI Ł. Multi-channel registered data denoising using wavelet transform. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 145–149. Łukasz JEDLIŃSKI Multi-channel registered data denoising using wavelet transform Odszumianie danych rejestrowanych wielokanałowo z użyciem transformaty falkowej* In order to obtain information regarding given phenomenon or object, it is usually necessary to register selected measurement signals obtained using sensors. Unfortunately, obtained signals, apart form desired information, contain disturbances caused by, amongst many other, properties of the measurement channel and processes associated with object operation. In many cases it is necessary to measure the same value in different places and/or directions. Thus, there is a demand for a tool improving signal to noise ration of the multi-channel registered signals.Wavelet transform is a relatively new method of data processing used in different fields (e.g. technique and physics). In case of signals it can be used for denoising, compression, trend detection or discontinuity detection. In this work it was used to denoise vibration signals registered by two three-axis sensors. Object of investigation was the bevel toothed gear. Signals denoising was to improve efficiency of the diagnosis of transmission gears teeth damage. Keywords: denoising, wavelet transform, artificial neural network, spiral bevel gear. W celu uzyskania informacji o interesującym nas zjawisku lub obiekcie najczęściej rejestrowane są wybrane sygnały pomiarowe otrzymane za pośrednictwem czujników. Niestety uzyskane sygnały oprócz pożądanej informacji zawierają również zakłócenia, które są spowodowane m.in. właściwościami toru pomiarowego i procesami towarzyszącymi działaniu obiektu. W wielu przypadkach zachodzi potrzeba pomiaru takiej samej wielkości w różnych miejscach obiektu i/lub kierunkach. Potrzebne są zatem narzędzia do poprawy stosunku sygnału do szumu sygnałów rejestrowanych wielokanałowo.Transformata falkowa jest stosunkowo nową metodą przetwarzania danych, która znalazła zastosowanie w różnych dziedzinach takich jak technika i fizyka. W odniesieniu do sygnałów może być używana do odszumiania, kompresji, wykrywaniu trendu czy nieciągłości sygnału. W pracy tej transformata falkowa została użyta od odszumiania sygnałów drgań zarejestrowanych z dwóch trójosiowych czujników. Obiektem badań była przekładnia zębata stożkowa. Odszumianie sygnałów miało na celu poprawę skuteczności diagnozy uszkodzenia kół zębatych przekładni. Słowa kluczowe: odszumianie danych, transformata falkowa, sztuczne sieci neuronowe, przekładnia stożkowa. 1. Introduction In the vibroacoustic diagnosis, like in other fields of science, it is desired to improve achieved results [2]. In the recent years one can observe continuous development of the algorithms of diagnostic computing and signal processing methods [3, 11, 16]. Measuring and computer equipment allow signal measurement with much higher accuracy and from many channels simultaneously. In case of examining complex objects there is a necessity of registering many signals. At the beginning of the investigation more measurement points are selected to avoid losing important information and obtain best measurement points. Obtained measurement signals always contain disturbances. In a simple signal model [1, 4, 10] it is assumed that signal contains valuable component (contain useful information) and random component (noise). There are many methods of extracting useful information from the signal, e.g. by signal filtration, PCA main components analysis, signal averaging. Among these method it is worth to mention wavelet transform (WT) that has been found to be useful in signal denoising [14]. Generalisation of the denoising using WT for one signal is a procedure proposed in [1] for many signals – it was used in this work. 2. Test stand An investigated object is a bevel toothed gear. Its body was equipped with two three-axis vibration acceleration sensors marked with 1 and 2 in Fig. 1. X axis of the sensors is parallel to the direction of input shaft axis and Z axis is an axis of output shaft. The vibrations was registered for the gear in good conditions and with damaged surface of teeth due to seizing. More precise description of the test stand can be found in [9]. 3. Discrete wavelet transform (DWT) In case of the Continuous Wavelet Transform (CWT) the wavelet coefficients are computed for each scale what generates large amount of data and requires long computations. Dis- (*) Tekst artykułu w polskiej wersji językowej dostępny w elektronicznym wydaniu kwartalnika na stronie www.ein.org.pl Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 145 Science and Technology signal component) and using high-pass filter we obtain detail Dj (high frequency signal component)[7]. Fig. 2 shows sample signal registered with 10kHz frequency. Decomposition resulted in signal in 0 – 5kHz range at the first level of approximation A1 and detail D1 - range 5 – 10 kHz. If the original signal consisted of 100 samples then the filtration results in obtaining a detail with length equal approx. 100 samples and an approximation with length equal approx. 100 samples. A sum of the resulting signals is approximately twice larger than original signal. To avoid this increment of number of samples decimation is used (removing every second sample from the obtained signal). Then we can perform further decomposition. Usually, an approximation is analyzed in order to obtain next detail and approximation. Signal decomposition is performed for finished number of levels due to signal length or physical sense of obtained details and approximations [18]. Usually, decomposition level does not exceed 8. Fig. 1. View of the examined gear with marked position of the vibration sensors crete Wavelet Transform is based on CWT and allows scale selection. DWT can be defined as [17]: DWT ( j , k ) = S0j −1/ 2 t − kτ 0 S j 0 j S 0 ∑ x k ψ k (1) where: S0 > 1, τ0 > 1, ψ – basis wavelet, x|k| – investigate signal, j, k – positive integers. For fast and efficient computations of DWT one can use an algorithm introduced by Mallat [12,15], known as multiresolution analysis (MRA). As a result of its operation we obtain multi-resolution representation of the signal in a form of approximations and details. Complete theoretical basis of the MRA can be found in Mallat’s work [12] and the dependency describing it is expressed by formula [17]: j −1 x(t ) = ∑ S0 (τ )φ (t − τ ) + ∑ ∑ d j (τ ) j / 2ψ (2 j t − τ ) (2) τ τ j =0 where: τ – shift coefficient, S0 – scale coefficient, dj – wavelet coefficient for j scale, ф(t), ψ(t) – scaling and wavelet function. Details and approximations are computed thanks to filtration with a two-channel set of filters (quadrature reflection filters [5,6]). In order to obtain signal decomposition into a few levels the filtration operation should be iterated (Fig. 2). Using low-pass filter we obtain approximation Aj (low frequency 146 Fig. 2. Signal decomposition using multi-resolution analysis algorithm [17] Removing irrelevant information (noise) after wavelet decomposition can be realized using a few methods. Next signal approximations contain signal component with lower and lower frequencies, thus the selection of proper approximation as a signal representation is then noise elimination method. Other possibility is simple modification of first detail (or details) by changing wavelet coefficients to zero. These methods of signal denoising are not very precise. More sophisticated denoising algorithms base on zeroing detail wavelet coefficients basing on a criterion calculated separately for each detail. Apart from selection of the proper threshold criterion (below which wavelet coefficients are set to zero) it is also necessary to select a method of threshold realisation. Hard thresholding method realizes denoising process by setting zero value for the elements which absolute value are below threshold value (other elements are not changed). A variation of this method is soft thresholding, in which not zeroed elements are also changed what Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Fig. 3. 3D plot of vibration signal Fig. 4. Vibration signals for the damaged gear from first to sixth channel before and after denoising Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 147 Science and Technology eliminates discontinuities in the location where the elements have values equal to threshold value [13]. Signal reconstruction is performed similarly to the decomposition. One should oversample details and approximations before synthesis in the filters (their selection is critical for complete reconstruction of the original signal). Signal decomposition level was equal three and for the basic wavelet the Coiflet 1 was chosen (Fig. 5) 3.1. Denoising procedure Denoising procedure for the multi-dimensional data is a generalisation of single dimension data denoising. The considerations are in accordance with Aminghafari et al. [1]. Let us assume the following p-dimensional signal model: X(t) = f(t) + ε(t), t = 1,..., n (3) where: X(t), f(t), ε(t) are of 1x p dimension, f(t) – signal to be denoising, ε(t) – Gaussian noise with unknown covariance matrix E(ε(t) T ε(t))=Σε. Every component X(t) has form for 1 ≤ i ≤ p: Xi(t) = fi(t) + εi(t), t = 1,..., n (4) where: fi – belongs to certain functional space (most often L2 or Besov’s). Covariance matrix Σε that is to be additionally defined, shows stochastic dependence between X(t) components and spatial correlation models. Denoising procedure can be expressed using three steps [1] for the X matrix that has n x p dimensions and consists of p signals (columns of X matrix) in a way that n>>p: • For each column of X matrix perform a wavelet decomposition of J-th degree. In this degree J+1 matrixes D1,...,DJ are obtained – they contain degree detail coefficients from 1 to J of p signals and approximation coefficients AJ of p signals. The matrixes Dj and AJ have dimensions n2-j x p and n2-J x p; •Determine the estimator Σε of the noise covariance matrix and perform SVD (singular value decomposition) of the Σε matrix using orthogonal V matrix where Σε = VAVT. Then change the basis using transformation matrix V (precisely calculating Dj V, 1 ≤ i ≤ p) and perform single-dimensional filtering using threshold ti = 2λi log ( n ) for the i-th column of the matrix Dj V; •Perform reconstruction of the denoised matrix using simplified details matrix and approximation through basis changed using VT matrix and reciprocal wavelet transform. 4. Test results Vibration signals registered during tests were denoised using the procedure described above. Denoising was performed for the signals recorded for transmission in good condition and for damaged one (6 channels were registered). Figure 3 shows a graph of wavelet coefficients which illustrated the change of frequency vibration signal in time. Denoising results in time domain for the damaged gear are shown in Fig. 4. Basing on the graphs for the axes of sensors according to the direction of output shaft axis one can notice that corresponding signals exhibit the highest level of noise. 148 Fig. 5. Wavelet Coiflet 1 For the same criterion of calculating thresholds for the signal details the denoising was realized using soft and hard thresholding. Then the features of signals were computed and eight of twenty of them were selected using an algorithm presented in [8]. Selected features: • average value, • RMS value, • peak value, • peak factor, • backlash factor, • standard deviation, • energy ratio, • FM0. In order to compare the efficiency of signals denoising the same features were computed for the signals without processing. Condition classification was performed using artificial neural network - multi layer perceptron (MLP). Table 1. Condition classification results for the unprocessed signal Network name Quality (learning) Quality (testing) Quality (validation) All trials MLP 8-4-2 89,88 91,66 91,66 90,41 Table 2. Condition classification results for the denoised signal (soft thresholding) Network name Quality (learning) Quality (testing) Quality (validation) All trials MLP 8-4-2 95,53 95,83 95,83 95,62 Table 3. Condition classification results for the denoised signal (hard thresholding) Network name Quality (learning) Quality (testing) Quality (validation) All trials MLP 8-9-2 97,02 94,44 94,44 96,24 Network classification results are given in the tables above. Classification efficiency for the unprocessed signal reached approx. 90% and for the denoised signal – approx 5% better. Differences between hard and soft thresholding were insignificant. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology 5. Summary The work presents a method of denoising vibration acceleration signals registered simultaneously for the same object. Comparison of the signals before and after denoising showed that for both sensors (in the same direction of vibration registration – along the axis of the input shaft) there were the highest amount of noise. Signal denoising was performed using two methods - soft and hard thresholding. Then the neural network was to recognize gear’s condition. Both methods resulted in similar and satisfying output. Classification quality achieved approx 96%. In order to verify efficiency of the method, similar procedure was performed for the original (unprocessed signal). Classification correctness was then smaller by approximately 5%. Literature 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. Aminghafari M, Cheze N, Poggi J M. Multivariate denoising using wavelets and principal component analysis. Computational Statistics & Data Analysis 2006; 50: 2381–2398. Antoni J, Bonnardot F, Raad A, El Badaoui M. Cyclostationary modelling of rotating machine vibration signals. Mechanical Systems and Signal Processing 2004; 18: 1285–1314. Antoni J, Randall R B. Unsupervised noise cancellation for vibration signals: part I-evaluation of adaptive algorithms. Mechanical Systems and Signal Processing 2004; 18: 89–101. Barszcz T. Decomposition of vibration signals into deterministic and nondeterministic components and its capabilities of fault detection and identification. Int. J. Appl. Math. Comput. Sci. 2009; 19(2): 327–335. Batko W, Dąbrowski Z, Engel Z, Kiciński J, Weyna S. Nowoczesne metody procesów wibroakustycznych. Radom: Wydawnictwo Instytutu Technologii Eksploatacji, 2005. Białasiewicz J T. Falki i aproksymacje. Warszawa: WNT, 2004. Bilgin S, Colak O H, Koklukaya E, Ari N. Efficient solution for frequency band decomposition problem using wavelet packet in HRV. Digital Signal Processing 2008; 18: 892–899. Jedliński Ł, Jonak J. Optimum choice of signals’ features used in toothed gears’ diagnosis. Diagnostyka 2010; 55: 9–12. Jedliński Ł, Jonak J. Quality evaluation of the bevel gear assembly based on analysis of the vibration signal. Diagnostyka 2010; 53: 23–26. Krukowski W, Józefczyk I. Transformacja falkowa w diagnostyce urządzeń mechanicznych. Diagnostyka 2008; 46: 75–82. Łazarz B, Wojnar G, Czech P. Early fault detection of toothed gear in exploitation conditions. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2011; 1: 68–77. Mallat S G. A theory for multiresolution signal decomposition: the wavelet representation. IEEE Transactions on Pattern Analysis and Machine Intelligence 1989; 7: 674–693. Misiti M, Misiti Y, Oppenheim G, Poggi J M. Wavelet toolboxTM 4. User’s guide. The MathWorks, Inc. 2010. Peng Z K, Chu F L. Application of the wavelet transform in machine condition monitoring and fault diagnostics: a review with bibliography. Mechanical Systems and Signal Processing 2004; 18: 199–221. Seker S, Ayaz E. Feature extraction related to bearing damage in electric motors by wavelet analysis. Journal of the Franklin Institute 2003; 340: 125–134. Wang Y, Zuo M, Lei Y, Fan X. Improvement of local mean approximation in empirical mode decomposition for gear fault detection. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2010; 2: 59–66. Wu J D, Kuo J M. An automotive generator fault diagnosis system using discrete wavelet transform and artificial neural network. Expert Systems with Applications 2009; 36: 9776–9783. Zimroz R. Zastosowanie analizy falkowej w diagnostyce uszkodzeń lokalnych układów napędowych maszyn górniczych. Diagnostyka 2009; 49: 113-122. Łukasz Jedliński, MSc. (Eng.) Department of Mechanical Engineering Lublin University of Technology ul. Nadbystrzycka 36, 20-618 Lublin, Poland E-mail: [email protected] Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 149 Article citation info: TOMPOROWSKI A. Stream of efficiency of rice grains multi-disc grinding. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 150–153. Andrzej TOMPOROWSKI Stream of efficiency of rice grains multi-disc grinding Strumień wydajności wielotarczowego rozdrabniania ziaren ryżu* A search for design solutions for grain grinding devices offering energy-saving production processes justify the research into the improvement of the theory and the design of grinders. The efficiency, functionality and performance of the rice grains grinding are significantly influenced by processes, difficult to describe, occurring within the working space of the grinder. Although there are some studies on the principles of the functionality and analysis of multi-disc grinders, an attempt has not been made so far at describing the influence of features and behaviour of ground material on efficiency, performance and usefulness of the biomaterial comminution/grinding process. The basis for the improvement of the functionality of a grinding device is an analysis of the potential of existing solutions and a determination of the scope of effective design features of the working unit within the permissible area. The criteria may be fulfilled by objective-oriented control of the design features of the multi-disc unit. In order for these actions to be effective, relevant relationships need to be explored and a mathematical description needs to be developed for the flow of ground grains through the working space of a multi-disc grinder, as a result variable of the structure and the operation of a working unit. Keywords: grinding, biomaterials, efficiency. Poszukiwania rozwiązań konstrukcyjnych zespołów rozdrabniających ziarna zbóż, prowadzące do energooszczędnych procesów produkcyjnych uzasadniają podjęcie badań nad doskonaleniem teorii i konstrukcji rozdrabniaczy. Istotny wpływ na wydajność, funkcjonalność i sprawność procesu rozdrabniania ziarna ryżu mają trudne do opisania zjawiska zachodzące w przestrzeni roboczej rozdrabniacza. Pomimo, że dostępne są opracowania na temat podstaw funkcjonalności i badań rozdrabniaczy wielotarczowych, jak dotychczas nie podejmowano próby opisu wpływu cech i zachowań rozdrabnianego materiału na wydajność, sprawność i użyteczność procesu rozdrabniania biomateriałów.Podstawą do poprawy funkcjonalności działania maszyny rozdrabniającej jest przeprowadzone rozpoznanie możliwości istniejących rozwiązań oraz określenie zakresu, skutecznych w przetwórstwie, cech konstrukcyjnych zespołu roboczego z obszaru dopuszczalnego. Spełnienie kryteriów może być osiągnięte między innymi na drodze celowego sterowania cechami konstrukcyjnymi zespołu wielotarczowego. Aby jednak działania te przyniosły planowane korzyści, konieczne staje się poznanie zależności oraz opracowanie opisu matematycznego przepływu rozdrabnianego ziarna przez przestrzeń roboczą rozdrabniacza wielotarczowego, jako zmiennej wynikowej konstrukcji i działania zespołu roboczego. Słowa kluczowe: rozdrabnianie, biomateriały, wydajność. 1. Introduction An innovative approach to the design studies is based on developing operational objectives, means or techniques that have not yet been integrated into the practice and not to mention into the engineering knowledge. The verification procedures may relate to such problems of the grinding process as dynamics of changes in the initial length of grains, transport of the ground material (grains of various lengths) with a simultaneous quasicutting between subsequent discs, a stream of particles leaving the machine with a specific intensity and under specific conditions. Consequently, the knowledge on the grinder design, as a practical and empirical field of science, focuses on not only the current situation but also on the future potential that is currently only conceptual [1, 2, 3, 5]. Purpose of the study: The main purpose of the study is to determine factors, design features, functional features and their influence on the dynamics and efficiency of grain biomaterial (long grain rice) grinding. The analysis carries out with a multi-disc grinder constructed according to a patent held by the University of Technology and Life Sciences in Bydgoszcz [4]. Additional aim of the work was to describe the complexity of the phenomena, process and relationships of the multi-disc grinding. 2. Rice grains comminution model For the purpose of volumetric efficiency tests, long grains rice of stabilized humidity parameters and standardized grain size were used as the feed material. Therefore, an assumption was made that for the purpose of this study an output model of the feed material (grains) is the ground substance of even size and repeatable constant dimensions of a single grain. (*) Tekst artykułu w polskiej wersji językowej dostępny w elektronicznym wydaniu kwartalnika na stronie www.ein.org.pl 150 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology It is assumed that in the holes of grinding discs, rice grains take the position along their longer axis, perpendicularly to the intra-disc cutting (shearing) plane. The orientation of the ground medium ρ nm within the op- erating space of the multi-disc multi-hole grinder is described by the probability distribution of the grain length. Because the material in the holes of the same disc is subjected to the same cutting process in all holes, the indexed state is the number of a disc (n) and the number of a cut (m): ρ nm : ( 0, lmax ] → [0,1] , lmax ∫0 ρ nm dl = 1. (1) Dislocation of the material to the next disc involves passing through the cutting process stage, n ≤ m. The initial state of the material, in hole of the first disc (all grains are intact) be- 0 fore the first cut equals ρ0 and constitutes an input state of the material that will be described with a certain function focused on lmax value. An input state of the material during the grinding process is changed due to two factors, mechanisms (fig.1): grinding (quasi-cutting) and removal of grains of the desirable or smaller size from the intra-disc spaces of the grinding unit of the device. moves from a hole of the previous disc to a hole of the next disc without being sheared along the whole length of the device. Because some of the material is removed to the space between discs, the volume of holes in the next disc should always be smaller than in the previous disc (this is achieved through a smaller section of a hole or smaller thickness of a disc, however, there are some design limitations related to stresses that a disc must carry). Thickness (h) of disc n, for the purpose of calculations with be represented by yn,, whereas will represent the height to which the material fills a hole of disc n before k cut. 2. Each point of the section of the next hole should, at a certain moment of the relative motion of the disc, be located within the section of a hole of the previous disc. 3. When a cut is made, a hole in the next disc is closed for transfer to the following disc (there are not arising socalled opening corridors). Fig. 2. Sectional area and effective grinding surface for two grinding discs between edges of the grinding holes [2], FR –intra-hole surface area of cut (for example: F1, F2) Fig.1. Technique (method and design means) of disc arrangement in the grinder for quasi–cutting of the grain material [2]; 1-previous disc, 2-next disc, 3-input state of the material, 4-next state of the material, di – hole diameter, t01, t02, trl – scale, h1, h2 – disc thickness, s – gap, V1, V2 – linear velocity 3. Grinding process When holes of the two adjacent discs meet (fig.2) and their combined section begins to increase (0, max), the next hole is filled with a part of ground material from the previous hole. In order for the cutting process to be efficient and effective, a hole in the previous disc must be completely filled when the combined section of the holes begins to decrease (max, 0).. The three conditions must be fulfilled: 1. The quantity of the material in the previous hole must be always greater than the space available in the next hole. Otherwise, we can imagine that the material The fill of a hole in a layer changes as follows (fig.3): The analysis shows that k cut after n−1 disc occurs earlier than k cut after disc n. With such numbering of cuts, quasi-cuts (for each disc the numbers begins from the first cut), grains may, at each boundary of individual discs, be subjected to cutting identified with the same number. When a hole is filled, the shared part of the cross-section of the holes begins to diminish – cutting (quasi-cutting) process. According to the initial assumption, each grain within the section of grinding holes is subjected to the cutting process. The orientation of grains in relation to the plane on which the cutting process takes place is random with even distribution. Grains of each length will be disintegrated (ground) with equal probability into two smaller particles, with the total length equal to the length (size) before cutting. Cutting (to be precise quasi-cutting) always occurs in the material that before the complete filling was in the previous disc. The distribution of grain length, when the grain is cut, in the material that fills the empty space of disc changes according to the following relationship: (2) m x 1 lmax m ρ nm+1 ( x ) = An,m ρ nm = 1 − ρ x + ∫ ρn ( l ) dl , m n ( ) yn+1 − y n+1 yn+1 − y nm+1 x while in the material that remains in disc n: x 1 l ρ nm+1 ( x ) = B n,m ρ nm = 1 − m ρ nm ( x ) + m ∫ max ρ n ( l ) dl , (3) x y y n n Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 151 Science and Technology The functions obtained are non-negative being a sum of two non-negative components. When integration is performed from 0 to l, it may be easily established that those are the probability distributions: (4) lmax 0 ∫ ρ nm+1 ( x ) dx = 1 − x 1 1− m + m y n y n lmax 0 ∫ x 1 + y nm y nm lmax lmax x 0 ∫ ∫ ρ n ( l ) dldx = 1 − 1 x + y nm y nm lmax x ρ 0 0 n ∫ ∫ ( l ) dxdl = is: x ρ n ( l ) dx = 1 m and similarly for distribution ρ n+1 . Therefore, operators An,m and B n,m are correctly defined stochastic operators. To simplify the analysis, an assumption was made that after the cutting is completed, the distribution of grain length in disc n+1 is homogeneous (the cut fraction and the fraction situated in the hole before the cut mix), and therefore it is weighted av- erage of ρ nk+1 and ρ nk : ρ m+1 x ( ) ρ nm+1 ( x ) = Bn,m ρ n,m = n ρ nm+1 ( x ) = (5) lmax lmin ρ nm+1 ( x ) ) −1 x〉 lmax (6) x〈lmin 0 Level of material after cut m (before cut m+1) in gap n, y nm+1 ( = yn − yn+1 + y nk+1 ) lmin Bn,m ρ nm ( x ) xdx 1 − ∫0 (7) lmax m B x xd ρ x ( ) n ,m n ∫0 In order to obtain the distribution in the whole space of the hole before cut m+1 (after re-fill), the following weighted average must be used: y nm+1 m−1 yn − y nm+1 ρ n+1 + An,m ρ nm ( x ) yn+1 yn+1 (∫ ρ nm+1 ( x ) = yn − y nm+1 y m+1 An−1,m ρ nm−1 + n Bn,m ρ nm yn yn (8) Operator is no longer a linear operator as , because it depends on the level of the material that remains in disc n after cut m: , and is a function of the probability distribution in the material (that determines the quantity of material removed from the machine during the cutting process) [6]. For to be treated as linear operators, values of must be treated, at each stage of the procedure, as pre-determined and iterative – corresponding to the results of studies and experiments. 5. Flow of particles leaving the machine At cut m, the flow of particles leaving the machine through a gap between disc n and n+1 is described with the probability distribution: lmax ρ l dl n( ) m sn ( x ) = ∫x (∫ lmin lmax x 0 ∫ 0, ρ n ( l ) dldx ) −1 l 〈lmin (9) l 〉 lmin and its volume equals: Fig. 3. Diagram showing transport of the material subjected to the grinding process with quasi-cutting between the discs; where indicates a cutting phase ( V = yn − yn+1 + l y nk+1 min B ρ m ( x ) xdx ) ∫0lmax Bn,m ρ nm ( x ) xdx ∫0 n,m n (10) 6. Changes of length distribution in discs The grain length distribution of disc n after cut m is de- 4. Removal process After the grain is cut, two layers of the material move with respect to one another in the direction of movement of two adjacent discs and gradient of their mutual velocities. Particles of the material are removed from the previous hole (they are subjected to component of force, gravitational force perpendicular to the intra-disc gap), whereas they are not removed from the next hole (because the direction of perpendicular component of gravitational force is away from the gap). After the cutting, the length distribution is as follows: 152 m m m scribed with ρ n . Operators used for distribution ρ n in ρ n+1 m+1 and in ρ n are identified accordingly as An,m and Bn,m . Changes of distributions are shown in the efficiency diagram, fig.4. 0 State ρ0 is known (distribution of grain length for the input m material). State ρ n was obtained from operation of the sum of 0 products of operators A and B for state ρ0 . The products repre0 sent all paths to that state from state ρ0 . For example: Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology ρ00 ρ00 ↘ ρ00 B0,0 → ↘ ρ01 ↘ ρ10 B1,0 → ↘ ρ 20 ⋮ B2,0 → B0,0 → ρ02 B1,1 → ⋮ → ρ03 ρ12 B1,2 → ρ13 ↘ B2,1 → B0,0 ρ 22 ⋮ B2,2 → ρ 23 ⋮ B0,0 → B1,3 → B2,3 → ⋯ ⋯ ⋯ ⋱ Fig. 4. Efficiency-based diagram of grain length distributions – products of quasi-cutting, for specific discs ρ 23 = ( A1,2 A0,1B0,0 + A1,2 B11, A0,0 + B2,2 A11, A0.0 ) ρ00 (11) In general, operators A and B do not commutate. It is es- sential for further analysis to calculate such expressions. States ρ are positive elements of Banach space1 L (0, 1), ↘ ↘ ρ12 m . n ↘ ↘ ρ11 of such subsets, i.e. the number of products in the sum equals ρ00 ρm State n is obtained by application of A operator n times1 and B operator m times . The path is unambiguously determined by n-element subset of a set of steps within the complete path {1, . . . ,m}, steps within which A operator operates. The number 1) to be more precise, weighted average values of the result of operation of that operator and operators A and B are endomorphisms of that space. For practical reasons, approximate states ρ with positive elements Rd and operators A and B with matrixes d×I. Physically, it corresponds to the division of a l long particle into d undividable parts, l/d long – the length corresponding to the linear measure of grinding. 7. Energy flow With an assumption that stresses will propagate at infinite rate, when closing of the shared area of two holes begins, material condensates evenly in each grain until a limit condensation is exceeded and a grain breaks. Then, layers move with respect to one another with the friction force constant for a unit of adjacent surfaces of two layers. Apart from that, there is also some friction of the material against surfaces of the discs where there are no holes in the next disc. 8. Conclusions Phenomena, processes and relations of the multi-disc grinding, despite their complexity, are relatively easy to describe formally. Determination of process factors (actions and methods), design features (means, devices and systems), operational conditions and their influence on the dynamics and efficiency of the rice grinding process, with a multi-disc grinder used as an example, was possible, with an assumption that quasi-cutting stresses will propagate at an infinite rate, in relation to: -- probability distribution of grain length, -- probability distribution of particles in a stream leaving the machine through a gap between discs. The research financed by the funds of the National Centre for Research and Development in 2010/2013 as a development project References 1. 2. 3. 4. 5. 6. Detyna J. Analysis of nonequilibrium stases in the sieve separation process. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2011; 1(49): 78-85. Flizikowski J. The Construction of the Food Grinders. Wydawnictwo Uczelniane Akademii Techniczno-Rolniczej w Bydgoszczy, Bydgoszcz 2005. Flizikowski J., Bieliński K., Bieliński M.: Raising Energy-efficiency of Multidisc Grinder Used for Grinding Crop Seeds for Animal Fodder. Wydawnictwo ATR- OPO, Bydgoszcz 1994. Flizikowski J., Bieliński M. Multidisc Grinder Especially for Grains. Patent.RP-144 566. Kaleta A., Wojdalski J. Food Processing. Selected Engineering-production Processes and Energy Issues. Warsaw 2007. Razavi S.M.A., Farahmandrar R.: Effect of hulling and milling on the physical properties of rice grains. International Agrophysics, 2008, 22, 353-359. Andrzej Tomporowski, PhD (Eng.) Faculty of Mechanical Engineering University of Technology and Life Sciences in Bydgoszcz ul. Prof. S. Kaliskiego 7/ bud. 3.2.pok 202 85- 788 Bydgoszcz e-mail: [email protected] Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 153 Article citation info: LINGAITIS L P, MJAMLIN S, BARANOVSKY D, JASTREMSKAS V. Prediction methodology of durability of locomotives diesel engines. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 154–159. Leonas Povilas LINGAITIS Sergey MJAMLIN Denis BARANOVSKY Virgilijus JASTREMSKAS Prediction Methodology of Durability of Locomotives Diesel Engines Metodologia prognozowania trwałości silników diesla w lokomotywach The article testifies that technical maintenance and repair terms as well as durability can be predicted accordingly to the maintenance parameters of the diesel locomotives. It is determined that fixing fuel consumption and capacity of diesel locomotives and taking in comparison with limit values allows to set a performance date for technical maintenance. Introduced suggested aspects of interrepair resource of the diesels dependent on comparable fuel consumption and evaluating their operating probability without failure for the durability prediction of diesel locomotives. Currently, the most common are three strategies: until failure, scheduled – premonitory and adaptive (diagnostic). When the quantity of necessary technical maintenances is known for the specific kind of diesel locomotives it is possible to determine interrepair resource depending on comparative and required per hour consumption of fuel and predict their durability taking into account probability of operating without failures. Keywords: durability, technical maintenance and repairs, locomotive engines, resources, reliability, prediction. W artykule wykazano, że częstotliwość przeglądów technicznych į remontów spalinowozów z silnikami diesla można prognozować analizując parametry eksploatacyjne. Obecnie najbardziej rozpowszechnione są trzy strategie: do awarii, planowo-wyprzedzająca i adaptacyjna (diagnostyczna). W konkretnych warunkach zarządzania gospodarczego na pierwsze miejsce wysuwa się specyfika użytkowania spalinowozów, o zaraz po niej – normatywne wymagania techniczne, reglamentujące eksploatację spalinowozów. Ustalono, że poprzez odnotowywanie zużycia paliwa oraz mocy spalinowozów można określić czas eksploatacji, po upływie którego konieczne będzie przeprowadzenie przeglądu technicznego. Przedłożona została teoretyczna zależność okresu międzyremontowego diesli od porównawczego zużycia paliwa, wykorzystywana do oceny prawdopodobieństwa bezawaryjnej pracy oraz zaproponowana metodyka prognozowania trwałości spalinowozów z silnikami diesla. Znając właściwą dla danej marki spalinowozu liczbę przeglądów technicznych, można oszacować okres międzyremontowy w zależności od godzinnego i porównawczego zużycia paliwa oraz uwzględniając ich prawdopodobieństwo bezawaryjnej pracy, w ten sposób prognozując ich trwałość. Słowa kluczowe: trwałość, przegląd techniczny į remont, silniki spalinowozów, okres międzyremontowy, niezawodność, prognozowanie. 1. Introduction Operating locomotives on the railroad leads to natural obsolescence of the engines as well as other components and details that consequently increases the number of failures. In order to increase operating reliability and durability of locomotive‘s diesel engines it is necessary to monitor their technical condition employing determined system of their technical maintenance and repairs which essence is to rebuilt nominal or approximate values of diesel state parameters while maintaining certain purposeful complex of implements as elimination of failures in operating conditions requires plenty of time and material resources. In that manner resource is rebuilt, high working probability of diesels without failures is retained. Great influence on reliability and durability while operating diesels rationally plays technical maintenance and repairs 154 operations. Accordingly, an opportunity to thoroughly analyze technical maintenance and repairs system implementation conditions and methods arises while periodically examining condition and equivalence to normative technical basis that regulate those processes particularly when more new enginery has been acquired. The aim of this work – theoretical validation of technical maintenance and repairs as well as durability prediction of diesel locomotives according to exploitation parameters. 2. Research methodology and results There exists a normative technical maintenance and repairs system of the locomotives that regulates technical maintenance and repairs of the machines [16, 17, 19]. However, improving enginery and its maintenance and repairs technologies there is Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology an appreciable necessity to establish a corrected complex of new organizational technical means that, on one hand, settles united principles of technical maintenance and, on the other hand, various standards and regulations for the planning and management of technical services activities of railways locomotive depots. With no doubt, a weighty influence on the diesel locomotives engines technical state maintenance has strategies of technical maintenance and repair systems. Currently, the most common are three strategies: until failure, scheduled – premonitory and adaptive (diagnostic). Within the given odds and outs [16, 17] all of them have a right to exist, yet a selection or creation of suitable strategy, furthermore, when selecting combinations of their elements, must be weight up well and, of course, reasoned by the appropriate manner. Particularity of the locomotives usage at specific property management conditions comes up in the foreground and only afterwards normative technical requirements that regulates operation of the locomotives [1–6, 10–13]. Certain algorithms with the attaining information about technical state changes as well as special technical state determination methods of the parts, as technical objects and their components are created and presented in the tasks [14–16, 18]. System of actions pointed to the management of diesel locomotives reliability and durability must base reliable and full information upon certain time limit. Being aware of the operating conditions, normative legal fundamentals, initial state (at the beginning of operating or after repair), chosen technical maintenance and repair system, material conditions of technical supplies allows to determine state of the object fairly precise. Logic says that maximum of information helps in choosing optimal unit of technical means and actions on certain manufacture conditions. It is a pity that there is no settled constitutional attitude on the support of locomotives technical state. Diagnostics that should determine technical state of separate details and assemblies has a limited application possibility and cannot include a huge amount of factors that influence common technical state. Guaranteed terms of necessary to perform tasks and moment when it is required to maintain correction actions of the technical state cannot be held if we follow planned – premonitory repair system of technical maintenance whereas applying adaptive strategy is possible only if we obtain very reliable information about technical state of an object. Until failure strategy does not give an opportunity to interpose into the management process. There have been many trials to forecast resources and system of technical maintenances and repairs [14, 15, 18] taking into account operating rates until now, but this problem hasn‘t been solved yet. Effective fuel capacity decreases depending on the technical state of diesel locomotives [7–9]. Evaluation of effective capacity that corresponds to the certain operating regime, according to the operating characteristics of the diesel locomotives it is possible to write expression of the fuel consumption per hour: Ge = N e Pmax g max ∫ 0 ∫ 0 P ( g e )dPdge , (1) where Ne – effective diesel capacity, kW; P(ge) – probability function of operation without failure subjected to comparative fuel consumption (it is chosen according to the technical state and graphical interpretation of the diesel; Pmax – maximal probability value of the diesel operation without failure; gmax – maximal comparative fuel consumption that are taken constant during the operating process, g /kW year. If we trace an amount of used fuel of diesels during every hour while operating and compare those consumption with limit values, then time when we need to perform technical maintenance is expressed in such form: n −1 t"GM = m ∑ Geiti + ∑ Gejlimt j i =1 m j , ∑ Gei i = 1, 2, 3...n j = n...m (2) i =1 where m ∑ Gei i =1 – sum of diesel fuel consumption between the limit hours during the interval from j up to m, according to the operating conditions this interval has been chosen by the leadership (investigator) of the depot, with the difference not exceeding at any circumstances 8, kg/h; m ∑ Gei – sum of aver- i =1 age fuel consumption per hour without limited fuel consumption per hour, kg/h; ti – i-th run-in during limited fuel consumption per hour, h; tj – j-th run-in during limited fuel consumption per hour from the fixed beginning, h. Using expression (2), when we know values of minimal, maximal and limited fuel consumption per hour, it is possible to show times of technical maintenance performances graphically (Fig. 1). If we write expression (2) taking into account formula (1), we get dependency of technical maintenance performances to their comparative fuel consumption per hour evaluating possibility of operation without failures: (3) n −1 t"gP M = Pmax g max ∑ ti PeV ∫ i =1 lim ∫ 0 m P ( g eV)dPdgeV+ ∑ t j Pej j m P ∑ eV i =1 m Pmax g max ∫ lim ∫ 0 Plim g max ∫ 0 ∫ 0 P ( g ejg )dPdg ej P ( g eV)dPdgeV Graphical interpretation (3) is given in Fig. 2. There is also necessary to put minimal, maximal and limit values of the comparative fuel consumption per hour. If we constantly (all the time) obtain information about variation of diesel capacitance during exploitation it is possible to draw graphs of operating probability without failures dependent on the effective capacitance (Fig. 3). The graph shows that there exists point 2, 3, 5 that correspond to maximal values of probability working without failures during diesels exploitation. Points that are below show decrease of diesels reliability and demand renewal (improvement) of their technical state. As we can see from the graph, transitions from zones with minimal and maximal values of effective capacity can be of several types, i.e. interruption can be sudden in order to improve technical state of diesels, rise the probability of operating without failures or it can be performed Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 155 Science and Technology Fig. 1. Technical maintenance performance times according to fuel consumption per hour of the thermo diesels Fig. 2. Technical maintenance performance times according to comparative fuel consumption on the diesel locomotives 156 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Fig. 3. Dependency of probability working without failures on the effective capacitance of diesel locomotives Fig. 4. Dependency of technical maintenance performance dates of diesel locomotives on comparative consumption of fuel and operating without failures probability Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 157 Science and Technology during planned technical maintenances that also guarantees high operating probability without failures. Constant maintenance of diesels high technical state ensures increase in durability of diesel locomotives. Graphical interpretation of obtained formula (3), i.e. dependency of diesels operating probability without failures on effective capacity P (Ne), results in possibility to specify formula (3) and write it in such a form: (4) Pmax g max n −1 ∑ ti PeV i =1 t"gP M = ∫ ∫ 0 lim m P ( g eV)dPdgeV+ ∑ t j Pej j m P ∑ eV i =1 m Pmax g max ∫ ∫ 0 Plim g max ∫ ∫ 0 0 P ( g ejg )dPdg ej P ( g eV)dPdgeV This expression describes terms of technical maintenances with higher probability that can be shown graphically (Fig. 4). In this case it is necessary to determine limits of the diesels operating without failures probability additionally. Those limits are chosen according to which accurateness technical state diagnosis must be reached. When we know amount of corresponding technical maintenances for the certain kind of diesel locomotives, we can estimate interrepair resource depending on the comparative consumption of fuel evaluating probability of operation without failures: (5) n −1 Pmax g max gN P t" Me ∑ ti ∫ = i =1 lim ∫ 0 lim m P ( g eV) P ( N ei )dPdg eV + ∑ t j j m 1 Pmax g max ∑m ∫ i =1 lim ∫ 0 Plim g max ∫ 0 ∫ 0 P ( g ejg ) P ( N ej )dPdgej P ( g eV) P ( N ei )dPdgeV A A where TM−1 , TM−2 – amount TM-1 and TM-2 respectively. In accordance to (2) and (3), durability of diesel locomotives can be expressed as follows. Dependent on: – fuel consumption per hour: n −1 LG = ( ATR + 1)( ATM −1 + ATM −2 ) m ∑ Geiti + ∑ Gejg t j i =1 m j ∑ Gei i =1 (6) – comparative fuel consumption evaluating probability(7) of diesels operating without failures: Pmax g max n −1 ∑ ti NeV ∫ gP L = i =1 ∫ 0 g m Pg g max j 0 P ( g eV)dPdgeV+ ∑ t j Pej m N ∑ eV i =1 m Pmax g max ∫ ∫ 0 g ∫ ∫ 0 P ( g ejg )dPdgej × P ( g eV)dPdgeV ×( ATR + 1)( ATM −1 + ATM −2 ) where ATR – total amount of overhauls of the diesel locomotives. (8) Using (5), durability of diesel locomotives can be expressed in such shape: n −1 Pmax g max ∑ ti ∫ gNe P L = i =1 g ∫ 0 m Pg g max j 0 P ( g eV) P ( N ei )dPdg eV+ ∑ t j m 1 Pmax g max ∑m ∫ i =1 g ∫ 0 ∫ ∫ 0 P ( g ejg ) P ( N ej )dPdgej × P ( g eV) P ( N ei )dPdgeV ×( ATR + 1)( ATM −1 + ATM −2 ) Prediction of life in economic activities is very important moment, because the object is expensive. 3. Conclusions Performed theoretical investigations on the durability and dates of technical maintenance and repair performance estimation of diesel locomotives depending on exploitation parameters have shown that: 1. Fuel consumption per hour can be described as graphical interpretation depending on the probability of operating without failures and effective capacity. 2. It is possible to determine necessary date for technical maintenance performance depending on the state of exploitation parameters if we fix fuel consumption of the diesel locomotives during its operating hours and compare them to limit values. 3. When the quantity of necessary technical maintenances is known for the specific kind of diesel locomotives it is possible to determine interrepair resource depending on comparative and required per hour consumption of fuel and predict their durability taking into account probability of operating without failures. References 1. 2. 3. 4. 5. 6. 7. 8. Bai Y, Jia X, Cheng Z. Group optimization models for multi-component system compound maintenance tasks. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2011; 31(49): 42–47. Batko W, Borkowski B, Glocki K. Application of database systems in machine diagnostic monitoring. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2008; 1(37): 7–10. Bilinton R, Allan R. N. Reliability evaluation of power systems. 2nd Ed. New York: Plenum Press 1996; 500 p. Bonvillian W. B, Weiss C. Stimulating innovation in energy technology. Issues in Science & Technology 2009; 26(1): 51–56. Holger A. Oil’s not well. Fairfield Country Business Journal 1999; 38(31): 21–22. Huang H. Z, Tong X, Zuo M. J. Posbist fault tree analysis of coherent systems. Reliability Engineering System Safety 2004; 84(2): 141–148. Lingaitis L. P, Pukalskas S. Ecological aspects of using biological diesel oil in railway transport. Transport 2008; 23(2): 138–143. Lingaitis L. P, Pukalskas S. The economic effect of using biological diesel oil on railway transport. Transport 2008; 23(4):287–290. 158 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. Lingaitis L. P, Pukalskas S. Determining the Consumption of Biodiesel by Locomotive Engines. Proceedings of the 11th international conference “Transport Means 2007”. Kaunas: Technologija, 2007; 194–197 p. Manbachi M, Mahdloo F, Haghifam M, Ataei A, Yoo C. A new approach for maintenance scheduling of Power systems, using a genetic algorithm and Monte-Carlo simulation. Maintenance and Reliability 2010; 4(48): 82–90. Mohanta D. K, Sadhu P. K, Chakrabarti R. Deterministic and stochastic approach for safety and reliability optimization of captive power plant maintenance scheduling using GA/SA-based hybrid techniques: a comparison of results. Reliability Engineering and System Safety Journal 2007; 92: 187–199. Wu X, Liu Q. Reliability demonstration test model for binomial systems with reliability growth. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2010; 2(46): 55–58. Xing L, Dugan J. B, Morrisssette B. A. Efficient reliability analysis of systems with functional dependence loops. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2009; 3(43): 65–69. Бажинов А. Научные основы оценки ресурса силовых агрегатов транспортных машин с учетом условий эксплуатации: диссертация доктора технических наук [Bazhinov A. Scientific basis for assessing resource power trains transport vehicles, taking into account the operating conditions: doctoral dissertation of technological sciences]. Харьков 2001; 324 с. (in Russian). Болотин В. Ресурс машин и конструкций [Bolotin V. Resource of machines and structures]. Москва: Машиностроение, 1980; 448 с. (in Russian). Калявин В. Технические средства диагностирования [Kalyavin V. Diagnostics equipment]. Ленинград: Судостроение, 1984; 208 с. (in Russian). Канарчук В. Долговечность и износ двигателей при динамических режимах работы [Kanarchuk V. Durability and engines wear in dynamic modes]. Киев: Научная мысль, 1978; 256 с. (in Russian). Мямлин С. Моделирование динамики рельсовых экипажей: монография [Mjamlin S. Modelling the dynamics of railway vehicles]. Днепропетровск: Новая идеология, 2002; 240 с. (in Russian). Сухарев Э. Эксплуатационная надежность машин. Теория, методология, моделирование [Sukharev E. The operational reliability of machines. Theory, methodology, simulation]. Ровно: НУВХП, 2006; 192 с. (in Russian). Prof. Leonas Povilas Lingaitis Department of Railway Transport Vilnius Gediminas Technical University J. Basanaivičiaus str., 28-135, LT-03224 Vilnius, Lithuania E-mail: [email protected] Prof. Sergey Mjamlin Dneprpetrovsk National University of Railway Transport Ak. Lazaryana Str. 2, UA-49010 Dnepropetrovsk, Ukraine E-mail: [email protected] Assoc prof. Denis Baranovsky Department of Automation Control Systems and Electric Drive Kremenchuk Mykhailo Ostrohradskyi National University Pershotravneva Str. 20, UA-39600 Kremenchuk, Ukraine E-mail: [email protected] Virgilijus Jastremskas Department of Railway Transport Vilnius Gediminas Technical University J. Basanaivičiaus str., 28-135, LT-03224 Vilnius, Lithuania E-mail: [email protected] Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 159 Article citation info: Chłopek Z. Testing of hazards to the environment caused by particulate matter during use of vehicles. Eksploatacja i iezawodnosc – Maintenance and Reliability 2012; 14 (2): 160–170. Zdzisław Chłopek Testing of hazards to the environment caused by particulate matter during use of vehicles Badania zagrożenia środowiska cząstkami stałymi podczas eksploatacji pojazdów samochodowych* The study presents results of tests on emissions of fractions of PM10, PM2.5 and PM1 dusts. For modeling of emissions of fractions of PM2.5 and PM1 particles, results of empirical tests were used as carried out in air quality supervision stations located in the agglomeration of the city of Brno. The results of modeling of emissions of fractions of PM2.5 and PM1 particles did not make it possible to make unequivocal conclusions, which proves that the discussed problem has to be treated statistically. However, a significant relation between models of emissions of fractions of particulate matter and sources of emissions of dusts and conditions for distribution of the same were observed. Keywords: dusts, particulate matter, PM10, PM2.5, PM1, vehicles. W pracy przedstawiono wyniki badań imisji frakcji pyłów PM10, PM2.5 i PM1. Do modelowania imisji frakcji cząstek stałych PM2.5 i PM1 wykorzystano wyniki badań empirycznych, przeprowadzonych na stacjach nadzorowania jakości powietrza w aglomeracji czeskiego miasta Brna. Wyniki modelowania imisji frakcji cząstek stałych PM2.5 i PM1 nie umożliwiły sformułowania jednoznacznych wniosków, co dowodzi konieczności statystycznego potraktowania badanego problemu. Stwierdzono jednak istotną zależność modeli imisji frakcji cząstek stałych od źródeł emisji pyłów i warunków ich rozprzestrzenia. Słowa kluczowe: pyły, cząstki stałe, PM10, PM2.5, PM1, pojazdy samochodowe. 1. Introduction Hazards posed by dusts to the environment are commonly known. The harmful character of dusts for human health has been discussed in a lot of studies relating both to health aspects [1, 10, 16, 18, 28, 31, 34, 35] as well as evaluation of factors affecting emissions of dusts [2, 3, 5–9, 11–15, 17, 19, 21, 22, 24, 27, 29]. Sources of emissions of dusts include natural phenomena and civilization activities. Most significant natural sources of emissions of dusts include volcanic eruptions, deposits, marine aerosols, animal and plant sources as well as forest fires. On a global scale, the natural sources of emissions of dusts are dominant, however, in the areas characterized by particularly intense human activities, anthropogenic sources of dusts have strongest impacts upon contamination of the environment. The anthropogenic sources of dusts include all production processes and fuel combustion processes. Automobile industry plays a significant role in contamination of the environment with dusts, especially in large centers of urban agglomerations. The harmful character of dusts for human health depends on chemical and mineral composition and physical structure of dusts as well as sizes of dust particles [2, 3, 6, 7, 17, 21, 22, 35]. Depending on conventional sizes of dust particles, the following particles may be distinguished [2, 3, 6, 7, 17, 21, 22, 35]: -- TSP (total suspended particles) – a mixture of small particles of conventional sizes not exceeding 300 μm and suspended in the air (a dispersed phase of the solid body–gas two–phase system), -- PM10 suspended dust – of conventional sizes not exceeding 10 μm, -- PM2.5 fine dust – of conventional sizes not exceeding 2,5 μm, -- PM1 nanoparticles – of conventional sizes not exceeding 1 μm, constituting practically invisible dust [24, 29]. Particulate matter with conventional diameters exceeding 10 μm is mainly arrested in upper respiratory tract, where most of them are exhaled, PM10 particles (with exclusion of PM2.5 particles) even penetrate lungs and, although they do not accumulate in the lungs, they accumulate in the upper respiratory tract. PM2.5 particulate matter penetrates the deepest paths of lungs, where they accumulate. PM1 particulate matter even penetrates the circulatory system. Particularly toxic particulate matter includes dusts containing heavy metal compounds and polycyclic organic compounds, most of which are characterized by carcinogenic properties [31, 35] Apart from the negative impact of dusts upon human and animal health, dusts also affect plants, soil and water. Combined with sulfur dioxide, carbon oxide and other compounds, dusts contribute to the formation of the London fog. [2, 11]. Dusts also impair the greenhouse effect in the atmosphere [2, 11]. It should also be noted that dusts limit visibility, which affects road traffic safety. The hazardous character of air pollution is evaluated on the basis of imission of pollution – concentration of pollution dispersed in the air and measured at the height of 1.5 m above the ground level [26]. Exceeded admissible imissions of PM10 par- (*) Tekst artykułu w polskiej wersji językowej dostępny w elektronicznym wydaniu kwartalnika na stronie www.ein.org.pl 160 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology ticulate matter in economically developed countries are most common reasons for authorities to undertake repair activities relating to the environment quality. Since 2009 a reduction of imissions of PM2.5 particulate matter in the European Union has been observed. It is planned to control imissions of PM1 particulate matter in the future. The evaluation of particular sources of dust emissions as regards their negative impacts upon the environment is very difficult, as tests on the air quality in particular places include influence of all existing sources. Moreover, the quality of air is also affected by conditions of distribution of pollution. Therefore, it is purposeful to conduct comparative tests in places characterized by various shares of sources of pollution emissions and distribution of the same. On the basis of analyses of results of such tests it is possible to draw conclusions concerning impacts of particular sources of pollution emissions upon the air quality. The basic difficulty of such tests involves a relatively scarce network of air quality monitoring stations, which conduct constant measurements of imissions of complete fractions of particulate matter, i.e. PM10, PM2.5 and PM1 according to the present condition. In order to evaluate impacts of particular sources of emissions of dusts upon imission of fractions of particulate matter it is additionally necessary to perform measurements with the frequency enabling identification of dynamic properties of processes, which describe the phenomena causing emission of dusts, e.g. vehicle traffic. It has also been evaluated that, for such purposes, it is necessary to perform measurements with time intervals not exceeding 1 hour. Requirements are also posed to testing time, as it is purposeful to consider variability of the processes determining the anthropogenic emission of dusts connected with a weekly cycle as well as the variability resulting from seasons of the year. Additionally, long–term testing may also contribute effectively to decrease impacts of interferences upon testing results as connected with accidental factors such as weather conditions. As it is known, fluctuations of weather factors have a normal character in a given area and, therefore, their expected values reach zero with lengthening of the observation period. Therefore, it is purposeful that measurements should be conducted for at least one year. As regards numerous air quality monitoring stations found in Europe, there are such urban agglomerations that include several stations located in places with diversified character of emission sources and distribution of pollution. Additionally, the stations test complete fractions of particulate matter and pollutions such as nitric oxides and carbon oxide, the imission of which is argued as connected with imission of fractions of particulate matter [5–9, 12]. The similar collection of air quality monitoring stations may be found, among others in, the agglomeration of the city of Brno. For the purposes of analysis of imissions of fractions of particulate matter, this study has used results of tests made by three air quality monitoring stations located in Brno. Brno is a city located in Moravia, the Czech republic. It is located in the south–east flat part of the country at the confluence of the Svratka and Svitava Rivers. Brno has over 400000 inhabitants (2008 ) and it occupies the area of 230 km2. The air quality monitoring stations are owned by the Division of Environmental Protection of the Municipality of Brno. The tests used results of measurements made by the stations located in Svatoplukova, Zvonařka and Lány. The stations differ in the character of the area, in which they are located. Svatoplu- kova and Zvonařka stations are located at large main roads and Zvonařka station is located directly at the road. Lány station is located far from busy roads. The stations measure imissions of PM10, PM2.5 and PM1 particulate matter as well as nitric oxides and carbon oxide every hour. This study does not use results of tests on imissions of nitric oxides and carbon oxides used for development of behaviorist [4] models of imissions of PM10 particulate matter [6–9], as these could not be contained in this study. However, the selections of stations considered deliberately the possibility of recording of imissions of nitric oxides and carbon oxide, which could facilitate obtaining of complete materials for modeling of imissions of particular fractions of particulate matter. 2. Modeling of imissions of PM2.5 and PM1 particles Hazards to the environment may be evaluated on the basis of direct measurements of imissions of pollution, however this evaluation only relates to the place and time of measurements and generalization of the test results is not always qualified sufficiently. However, results of long–term tests in places with typical conditions of emission of pollution and distribution of pollution qualify for generalization of conclusions. In such cases, results of tests and modeling of imissions of pollution constitute a basis for evaluation of pollution of the environment. In other cases, hazards to the environment are evaluated on the basis of knowledge of emission of pollution and modeling of distribution of pollution. The knowledge of emission of pollution is possible owing to results of measurements and, in this case, there are the same restrictions as in the case of measurements of imissions. It is completely different in the case of mobile sources of emission such as, for example, vehicles. In this case, it is possible to model pollution only. As modeling of emission of pollution constitutes a basic tool for evaluation of hazards to the environment in most affected places, i.e. in city centers. Traditionally, all types of modeling connected with emission of pollution are referred to as modeling of emission of pollution, although in many cases the modeling applied formally to imissions. This simplification is justified in the possibility of concise formulation of opinions, although, formally it is inaccurate. Modeling of emissions of PM10 particulate matter does not constitute the subject of this study, however, it is inherently connected with modeling of imission of PM2.5 and PM1 particles. Modeling of emission of PM10 particles has been described extensively in literature [6–9, 12–15, 21, 22, 27]. The following two testing methods are used: -- modeling of emission of PM10 particles on the basis of knowledge of traffic and properties of vehicles and roads – models created on the basis of structural similarity [4], -- modeling of imission of PM10 particles on the basis of imission of nitric oxides and carbon oxide – models created on the basis of functional similarity (behaviorist models) [4]. -- models created on the basis of structural similarity consider the following sources of emission of PM10 particulate matter [6, 7, 13–15, 27]: -- vehicles, -- surface of the road, Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 161 Science and Technology -- solid contamination found on roads – in the form of excitation of dusts. -- sources of dusts emitted by vehicles include [2, 3, 5–9, 13–15, 27]: -- a combustion engine – particulate matter contained in the exhaust gas[11, 23, 29], -- friction pairs – found mostly in the braking system [2, 3] and coupling, -- tires, -- other parts of vehicles that are subject to wear and tear. Behaviorist models do not openly consider sources of emission of particulate matter, including those connected with automobiles and other particles. The behaviorist models use a significant statistical relation of imission of particulate matter and imission of other pollution and the theory and practice of automobile technology at least partly justifies such a relation, e.g. simultaneous increase of emission of particulate matter from combustion engines and other vehicle sources and emission of carbon oxide and nitric oxides with an increase of the vehicle velocity and, consequently, engine load. The behaviorist models usually argue for a linear relation between imission of particulate matter and other contamination. Generally, results of the analysis of models constructed on the basis of structural similarity cannot be compared to results of the analysis of models constructed on the basis of functional similarity, as structural models do not openly consider dust emission sources other than those connected with vehicle traffic. In reality, a wide scale of discretion of adoption of structural model coefficients, which are usually difficult to identify, causes it to become a significant reason for incomparability of results of the analysis of structural and functional models. The fraction of PM2.5 particles may be treated as a subset of PM10 fractions. Therefore, a linear relation between imission of PM2,5 particles– IPM2,5 and imission of PM10 particles– IPM10 is postulated: I PM 2,5 = k PM 2.5−10 ⋅ I PM 10 (1) where: kPM2.5–10 – coefficient of the model of emission of PM2.5 particulate matter; k PM 2.5−10 ∈ 0;1 Similarly to the modeling of imission of PM2.5 particles, PM1 particles may be treated as a subset of PM10 particles and PM2.5 particles. Thus, imission of PM1 particles – IPM1 may be modeled in a linear relation to imission of PM10 particles: I PM 1 = k PM 1−10 ⋅ I PM 10 (2) where: kPM1–10 – coefficient of the model of emission of PM1 particulate matter; k PM1−10 ∈ 0;1 and in a linear relation to imission of PM2.5 particulate matter: I PM 1 = k PM 1−2.5 ⋅ I PM 2.5 (3) where: kPM1–2.5 – coefficient of the model of emission of PM1 particulate matter; k PM1−2.5 ∈ 0;1 . 162 Identification of models of imission of PM2.5 particles (1) and imission of PM1 particles (2 and 3) involves determination of coefficients of kPM2.5–10, kPM1–10 and kPM1–2.5 models on the basis of results of empirical tests on imission of fraction of PM10, PM2.5 and PM1 particles. Identification results generally depend on conditions of emission of pollution and distribution of pollution as well as the period of measurements. 3. Testing of imission of PM10, PM2.5 and PM1 particles in selected air quality monitoring stations As used in this study, testing in air quality monitoring stations in Brno was conducted in the period from 1 January to 31 December 2010 with a sampling interval of 1 h. Fig. 1–3 present courses of imission of fraction of particulate matter for averaged values within the period of 1 week for time t as marked with day numbers – d and month numbers – m. The course of imission of fraction of particulate matter indicates a strong relation between the imission and seasons of the year: imission increases considerably in winter months. One may also observe a relation between imission of fractions and weekdays, which indicated a strong influence of civilization factors upon the imission. The mutual relationship between imissions of particular fractions is especially visible, which justifies adoption of linear models (1–3). Fig. 4–6 present statistical characteristics of the testes sets of imission of fractions of particulate matter1 : minimum value, maximum value, average value, standard deviation and span. There are considerable differences in extreme values of imission of particular fractions of particulate matter. The least values: maximum, minimum and average values of imission of particular fractions were recorded for Lány station (apart from the average value of imission of PM1 particles and minimum value of imission of PM10 particles that are very similar to the values recorded for Zvonařka station). It is interesting that the greatest maximum and average values of imission of all fractions were recorded in Svatoplukowa station located in the area with less intense road traffic than in the case of Zvonařka station. A strong correlation between imission of fraction of particulate matter results from the same. This is confirmed in the analysis of the correlations. The analysis was carried out with the use of Pearson’s theory of linear correlation [30] and non–parametrical methods [33]: Spearmann rang correlation [32], Kendall tau correlation [20] and Kruskal gamma correlation [25]. Fig. 7–9 present coefficients of Pearson r, Spearmann R, Kendall tau and Kruskal gamma correlations between the tested sets. The probability that the hypothesis assuming absence of correlation between the tested sets will not be rejected does not exceed 1∙10–6 in all cases. Results of the analysis of correlation of sets of imission of size fractions of particulate matter fully qualify for formulation of an opinion on a strong correlation between the tested sets. The values of Pearson correlation coefficient for particular sets and the probability that the hypothesis assuming absence of Pearson correlation between the tested sets 1In statistics and, in particular, in commercial applications, barely formal nomenclature is used, which does not always comply with the formalized mathematics. Therefore, terms such as “maximum value” should be treated as “the greatest value” and “minimum value” as “the least value”, as there are not extreme values within the meaning of terms applied in a mathematical analysis. However, due to the fact that such terms are common and make it possible to provide concise statements, this study uses them in descriptions. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Fig. 1. The process of imission I of PM10, PM2.5 and PM1 particles in Brno–Svatoplukova air quality monitoring station Fig. 2. The process of imission I of PM10, PM2.5 and PM1 particles in Brno–Zvonařka quality air monitoring station Fig. 3. The process of imission I of PM10, PM2.5 and PM1 particles in Brno–Lány air quality monitoring station Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 163 Science and Technology Fig. 4. Statistical characteristics of concentration of PM10: min – minimum value, max – maximum value, AV – average value; D – standard deviation, Δ – span Fig. 5. Statistical characteristics of concentration of PM2.5 particles: min – minimum value, max – maximum value, AV – average value; D – standard deviation, Δ – span Fig. 6. Statistical characteristics of concentration of PM1 particulate matter: min – minimum value, max – maximum value, AV – average value; D – standard deviation, Δ – span 164 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Fig. 7. Coefficients of Pearson r, Spearmann R, Kendall tau and Kruskal gamma correlations between sets of imission of PM10 and PM2.5 particles Fig. 8. Coefficients of Pearson r, Spearmann R, Kendall tau and Kruskal gamma correlations between sets of imission of PM10 and PM1 particles Fig. 9. Coefficients of Pearson r, Spearmann R, Kendall tau and Kruskal gamma correlations between sets of imission of PM1 and PM2.5 particles Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 165 Science and Technology Fig. 10. The process and AV average value of k coefficients of imission models of PM2.5 and PM1 particles in Brno–Svatoplukova air quality monitoring station Fig. 11. The process and AV average value of k coefficients of imission models of PM2.5 and PM1 particles in Brno–Zvonařka air quality monitoring station Fig. 12. The process and AV average value of k coefficients of imission models of PM2.5 and PM1 particles in Brno–Lány air quality monitoring station 166 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Fig. 13. Statistical characteristics of coefficients of imission model of PM2.5 particles: min – minimum value, max – maximum value, AV – average value; D – standard deviation, W – variability coefficient; Δ – span; ρ – relation between the span and average value Fig. 14. Statistical characteristics of coefficients of imission model (2) of PM1 particles: min – minimum value, max – maximum value, AV – average value; D – standard deviation, W – variability coefficient Δ – span; ρ – relation between the span and average value Fig. 15. Statistical characteristics of coefficients of imission model (3) of PM1 particles: min – minimum value, max – maximum value, AV – average value; D – standard deviation, W – variability coefficient Δ – span; ρ – relation between the span and average value Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 167 Science and Technology Fig. 16. AV average value of k coefficients of the model of emission of PM2.5 particles Fig. 17. AV average value of k coefficients of the model (2) of emission of PM1 particles Fig. 18. AV average value of k coefficients of the model (3) of emission of PM1 particles 168 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology justify adoption of linear models of imission of PM2.5 and PM1 particles. On the basis of empirical tests, parameters of imission models of PM2.5 and PM1 particles were identified. Fig. 10–12 present courses of the coefficient of imission models of PM2.5 and PM1 particles and the average value of those coefficients during tests. There is a visible regularity involving that in cold months coefficients of imission models of fractions of PM2.5 and PM1 particles are greater than in warm months, which means a greater share of fine particles in cold months. Fig. 13–15 present statistical characteristics of parameters of imission models of fractions of particulate matter. The variability coefficient and relation between the span and average value for coefficients of the models is considerably smaller than in the case of imission sets. The variability coefficient for coefficients of the models is (5 ÷ 20)%. Average values of imission models of fractions of particulate matter were compared in Fig. 16–18. The determined average values of coefficients of imission models of PM2.5 and PM1 particles are within normal limits found in literature [21, 22]. The results of identification of imission models of PM2.5 and PM1 particles cannot be interpreted unambiguously and one may even say that they are puzzling. The values of coefficients of models for Zvonařka and Lány stations are similar, especially for models of imission of PM2.5 particles and model (3) of imission of PM1 particles. In the case of the model (2) of imission of PM1 particles, the difference of the model coefficient for Lány and Svatoplukova stations is even greater than for Zvonařka and Lány stations. One should expect similar values of model coefficients for Zvonařka and Svatoplukova stations or Svatoplukova and Lány stations, which results from conditions of location of the stations and, in particular, from traffic in the roads found in the vicinity of the stations. 4. Conclusions Dusts constitute one of most severe hazards for the environment, especially in centers of large urban agglomerations. Evaluation of imission of particular fraction uses results of empirical tests carried out in air quality monitoring stations as well as results of modeling of imission of pollution. Testing of imission of particular fractions of dusts use emission models of PM10 particles that are constructed on the basis of structural similarity and models of distribution of pollution as well as imission models of PM2.5 and PM1 particles constructed on the basis of functional similarity. Identification of functional imission models of PM2.5 and PM1 particles (as carried out on the basis of results of meas- urements of imissions of fractions of PM10, PM2.5 and PM1 particles in 2010 in three air quality monitoring stations in Brno as characterized by diversified sources of emission of pollution and distribution of pollution) made it possible to draw the following conclusions: 1. A strong correlation may be noticed between sets of imissions of particular fractions of particulate matter in all stations. 2. There is a strong relation between imission of fractions of particulate matter and seasons of the year: imission is much greater in cold seasons of the year. 3. There are also relations between imission of fractions of particulate matter and days of the week, which indicates a strong impact of civilization factors upon the imission. 4. There is a visible mutual relation between imission of particular fractions, which justifies adoption of linear models of imission of PM2.5 and PM1 particles. 5. There are great differences in extreme values of imissions of particular fractions, which is confirmed by great values of the variability coefficient and relation between the span and average value. 6. The determined average values of coefficients of imission models of PM2.5 and PM1 particles are within normal limits found in literature [21, 22]. 7. There is a visible regularity involving that in cold months coefficients of imission models of fractions of PM2.5 and PM1 particles are greater than in warm months, which denotes a greater share of fine particles in cold months. 8. The results of identification of imission models of PM2.5 and PM1 particles cannot be interpreted unambiguously. No results were obtained indicating an impact of road traffic upon composition of size fractions of particulate matter. The ambiguousness of results of identification of imission models of PM2.5 and PM1 particles indicates that it is necessary to treat this issue in a more comprehensive way. One may justify the expectation that on the basis of a larger set of result of empirical tests, which also include results from the stations located in other areas, it is possible to draw more unambiguous and general conclusions. Despite the partly critical evaluation of results of testing of imission models of PM2.5 and PM1 particles it may be stated that modeling of size fractions of particulate matter in accordance with the criterion of functional similarity is the only effective method of testing of hazards to the environment caused by dusts. Literature 1. 2. 3. 4. 5. 6. Canagaratna M. Chase studies of particulate emissions from in–use New York City vehicles. Aerosol Science and Technology 2004; 38(6): 555–573. Chłopek Z, Jakubowski A. A study of the particulate matter emission from the braking systems of motor vehicles. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2009; 4: 45–52. Chłopek Z, Jakubowski A. The examination of the reduction of particulate matter emission from motor vehicle braking systems. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2010; 4: 29–36. Chłopek Z, Piaseczny L. Remarks about the modelling in science researches. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2001; 4: 47–57. Chłopek Z, Skibiński F. Wprowadzenie do tematyki emisji cząstek stałych PM2,5 z transportu samochodowego. Transport Samochodowy 2010; 3: 73–87. Chłopek Z, Żegota M. Badania emisji cząstek stałych PM10. Rozdział monografii „Edukacja ekologiczna. Podstawy działań Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 169 Science and Technology 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. naprawczych w środowisku”. Nałęczów: Polskie Towarzystwo Inżynierii Ekologicznej, 2004: 114–120. Chłopek Z, Żegota M. The emission of particulate matter PM10 from vehicles. Eksploatacja Niezawodnosc – Maintenance and Reliability 2004; 1: 3–13. Chłopek Z. Modele behawiorystyczne emisji cząstek stałych PM10 ze źródeł trans¬portu drogowego. Zeszyty Naukowe Instytutu Pojazdów Politechniki Warszawskiej 2011; 1: 111–118. Chłopek Z. Ocena stanu zagrożenia środowiska przez cząstki stałe PM2,5 ze źró¬deł transportu drogowego. Zeszyty Naukowe Instytutu Pojazdów Politechniki Warszawskiej 2011; 1: 101–110. Christoforou C S. Trends in fine particle concentration and chemical composition in Southern California. Journal of the Air & Waste Management Association 2000; 50(1): 43–53. Eastwood P. Particulate emissions from motor vehicles. Chichester: John Wiley, 2008. Environment Australia: Emission estimation technique manual for aggregated emission from motor vehicles. Version 1.0. Canberra, 2000. EPA: Addendum to Emission Factor Documentation for AP–42, Section 11.2.5 and 11.2.6 (now 13.2.1). Paved Roads. Final Report. September 1997. EPA: Compilation of Air Pollutant Emission Factors AP–42, 5. Edition, Vol. I, Chapter 13, Section 13.2.1. Miscellaneous sources. October 1997. EPA: Compilation of Air Pollution Emission Factors – AP 42, Emission Factor Documentation, Section 13.2.1, Paved Roads. Appendix C.1: Procedures for Sampling Surface Dust Loading and Appendix C.2: Procedures for Laboratory Analysis of Surface Loading Samples. 1993. Forsberg B. et al. Comparative health impact assessment of local and regional particulate air pollutants in Scandinavia. Journal of the Human Environment 2005; 34(1).: 11–19. Harrison R M, Jones A M, Lawrence R G. Major component composition of PM10 and PM2.5 from roadside and urban background sites. Atmos. Environ 2004; 38: 4531–4538. Hoek G. et al.: Association between mortality and indicators of traffic–related air pollution in the Netherlands: a cohort study. Lancet 2002; 360:1203–1209. Houthujis D. et al. PM10 and PM2.5 concentrations in central and eastern Europe: Results from the CESAR study. Atmospheric Environment 2001; 35: 2757–2771. Kendall M G. A New measure of rank correlation. Biometrika 1938; 30: 81–89. Keogh D U, Ferreira L, Morawska L. Development of a particle number and particle mass vehicle emissions inventory for an urban fleet. Environmental Modelling & Software 2009; Vol. 24, Issue 11: 2009 1323–1331. Ketzel M. et al. Estimation and validation of PM2.5/PM10 exhaust and non–exhaust emission factors for practical street pollution modeling. Atmospheric Environment 2007; Vol. 41, Issue 40: 9370–9385. Kinney P L. et al. Airborne concentration of PM2.5 and diesel exhaust particles on Harlem sidewalks. A community – based pilot study. Environmental Health Perspectives 2000; 108: 213÷218. Kittelson D B. Engines and nanoparticles: A review. J. Aerosol Sci. 1998; Vol. 29, No. 5/6. Elsevier Science Ltd.: 575–588. Kruskal H, Wallis W A. Use of ranks in one–criterion variance analysis. Journal of the American Statistical Association 1952; 47(260): 583–621. Leksykon ekoinżynierii. Pod red. Gabriela Borowskiego. Warszawa: Polskie Towarzystwo Inżynierii Ekologicznej, 2010. Lohmeyer A, Düring I. Validierung von PM10–Immissionsberechnungen im Nahbereich von Straßen und Quantifizierung der Staubbildung von Straßen, Lützner Straße in Leipzig. Sächsisches Landesamt für Umwelt und Geologie. Dresden über Staatliche Umweltbetriebsgesellschaft. Radebeul: Februar 2001. MacNee W, Donaldson K. Exacerbations of COPD – environmental mechanism. Chest 2000; 117: 390–397. Mayer A. et al. Nanoparticle–emissions of EURO 4 and EURO 5 HDV compared to EURO 3 with and without DPF. SAE Emissions Measurement and Testing 2007: 335–43. Pearson K. On the theory of contingency and its relation to association and normal correlation. Drapers’ Company Research Memoirs. Biometric 1904; Ser. I. Schwartz J, Laden F, Zanobetti A. The concentration–response relation between PM2.5 and daily deaths. Environmental Health Perspectives 2002; Vol. 110, No. 10: 1025–1029. Spearman Ch.: The proof and measurement of association between two things. Americal Journal of Psychology 1904; 15: 72–101. Wasserman L. All of nonparametric statistics. Springer, 2007. Wilkinson P. et al. Case–control study of hospital admission with asthma in children aged 5–14 years: relation with road traffic in north west London. Thorax 1999; 54: 1070–1074. Yuh–Shen Wu et al. The measurements of ambient particulates (TSP, PM2,5, PM2.5–10), chemical component concentration variation, and mutagenicity study during 1998–2001 in central Taiwan. Journal of Environmental Science and Health, Part C Environmental Carcinogenesis and Ecotoxicology Reviews 2002; Vol. 20, Issue 1: 45–59. Prof. Zdzisław Chłopek The Motor Transport Institute in Warsaw ul. Jagiellońska 80, 03–301 Warsaw E–mail: [email protected] 170 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Article citation info: HARDYGÓRA M, KOMANDER H, BŁAŻEJ R, JURDZIAK L. Method of predicting the fatigue strength in multiplies splices of belt conveyors. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 171–175. Monika HARDYGÓRA Henryk KOMANDER Ryszard BŁAŻEJ Leszek JURDZIAK Method of predicting the fatigue strength in multiplies splices of belt conveyors Metoda prognozowania trwałości zmęczeniowej złączy wieloprzekładkowych taśm przenośnikowych* Method of testing the stress distribution in overlap adhesive joints of multiply conveyor belts is presented The impact of strength properties of belts and adhesive rubber on the stress scale and fatigue life of splices were defined. The relations enabling to calculate the strength magnitude and predicting the durability of conveyor’s belts splices were given. Keywords: belt conveyors, conveyor belts, conveyor’s belt splices. Przedstawiono metodę badań rozkładu naprężeń w spoinie klejowej zakładkowych złączy taśm przenośnikowych wieloprzekładkowych. Określono wpływ właściwości wytrzymałościowych taśm i gumy klejowej na wielkość naprężeń i na trwałość zmęczeniową złączy. Podano zależności pozwalające na obliczenie wielkości naprężeń i prognozowanie trwałości złączy taśm przenośnikowych. Słowa kluczowe: przenośniki taśmowe, taśma przenośnikowa, złącza taśm przenośnikowych. 1. Introduction Multiplies conveyor belts manufactured in parts of certain length are connected into longer sections or loops according to the pattern showed on figure 1. The specificity of splice construction is that in cross-sections of plies̕ contacts is one ply less than in belts being connected. Loss of belt strength in the splice area is inversely proportional to the number of belt plies. breakage of plies on the joint of the first or the last notch. Often in practical use the connections of splice are unstuck, what is the beginning of its destruction. It results from the fact that under the fatigue loads the adhesive joint is damaged [1,2,3]. Problems related with establishing what properties of conveyor belts and vulcanization materials have the essential impact on the stresses in the adhesive joint and its fatigue life, were the subject of studies carried out in Laboratory of Belt Transportation of Mining Laboratory at Wroclaw University of Technology [4,5]. The results are presented in the paper. 2. Tests of stresses in adhesive joint of conveyor belt splice. Fig. 1 Dimensional diagram of a 4-plies conveyor belt splice Test results of static strength of splices indicate that loss of strength is bigger than the one resulting from the loss of one ply. This is caused by shearing stresses in the adhesive joint, which are the biggest on the ply connections and strengthen th e function of notch at that point of splice. During the test of static tension of properly manufactured splice it is destroyed due to The size of stresses in adhesive joint of splice was tested by measuring the angle of non-dilatational strain of γ joint. γ angle is defined as ratio of absolute ΔS non-dilatational strain and the distance between relocating edges of the joint under the F stretch force (fig.1). tgγ = ∆S g (1) In view of properties of fabric plies and the rubber of adhesive joint, the angle of γ non-dilatational strain is not the same at the whole length of separate notch of the slice. (fig.2), thus (*) Tekst artykułu w polskiej wersji językowej dostępny w elektronicznym wydaniu kwartalnika na stronie www.ein.org.pl Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 171 Science and Technology its measurement must be made at many points along the length of the splice. Fig. 2 Strain diagram of the adhesive joint along the length of the 1st step of a belt splice The measurements of γ angles were made on the full-scale splices of four-ply belts. To render the test results independent from the impact of possible production defects, the splices were made in the special manner. Individual notches were obtained through cutting the belt plies at certain places equal the notch length. 50 mm wide samples for tests were cut out along the belt axis and then were loaded by the force giving the stress accounting for 15% of belt strength. Basing on measured γ angles, the chart of the γ=f(lx) relation where lx is a distance between the measuring point and the place of contact of cut plies was plotted. Example results of γ measurements after the approximation by the trend line are presented on figure 3. Fig. 4 Dimensional deformations diagram of adhesive joint component of a belt splice along the splice, the tests of adhesive rubber were carried out to get the τ=f(ε) relation, what in turn helped to recalculate the results into the τ=f(lx) relation. Example results of calculations of stress distribution in the joint of EP1000/4 and PP1000/4 belt splices were showed on figure 5. Fig. 5 Stress distribution in the adhesive joint along the length of EP1000/4 and PP1000/4 belt splices Fig. 3 Distribution of non-dilatational strain angles γ in adhesive joint along the length of EP1000/4 and PP1000/4 belt splices The charts of γ=f(lx) relations were approximated by trend lines and the relative elongations of ε joint were calculated using the formula (2): where: ε = 1 − cos γ − ε t ⋅ν cos γ + ν ⋅ sin γ (2) εt – relative elongation of belt at the stress of 15 % of belt strength, ν - Poisson’s ratio of adhesive rubber. The foregoing relation was determined investigating the schema of adhesive joint deformation presented on figure 4, assuming that relative elongation of the joint is ε=(g1-g)/g and taking into consideration the impact of belt elongation as well as Poisson’s ratio of adhesive rubber. The calculations gave the charts of distribution of adhesive joint elongations along the splice length i.e. ε=f(lx). In order to obtain the picture of stresses distribution in the adhesive joint 172 The results of measurements of basic properties of spliced belts, properties of the adhesive rubber as well as results of tests on stresses magnitude at outside connections of splices are presented in Table 1. Stretch resistance of belts (Rt) and elongations ( εt) were determined according to EN ISO 283:2008 standard, modulus of elasticity of belt Mt according to EN ISO 9856:2005 standard, stretch strength of adhesive rubber according to ISO 252:2007 standard, modulus of adhesive rubber according to ISO 252:2007 standard at elongation of 100% reached after 3 hour of loading the samples with constant force. In the tests both γ angles of adhesive joint and modules of joint rubber were determined after three hours after their stress because of big impact of creeping effect of those materials on their elongation during the initial period of stretching forces application. The results of tests presented in table 1 were analyzed to present the relation of τ stresses versus Mt/Rt and Mg variables. The following form of this relation was assumed (3): where: τ = C( Mt x ) M g y Rt (3) C – constant, x, y – exponents in demand. This function was brought to the linear by taking a logarithm of both sides of the equation and thus the linear regression could be applied. The C, x and y values in demand were determined using STATGRAPHICS Centurion XV (v.15.2.06, Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Table 1. Results of strength properties tests of belts and stresses in adhesive joint of splice Properties of connected belts Item Type of material plies Stretch strength Rt [kN/m] Modulus of belt elasticity Mt [kN/m] Properties of adhesive rubber Relative elongation at 15% Rt εt [%] Modulus of rubber Mg [MPa] Maximum stress in adhesive joint τ [MPa] 1 PP 1356 4639 6.28 6.2 1.3 1.70 2 PP 1258 4375 5.45 10.9 1.7 1.65 3 EP 1050 12665 1.11 12.4 0.8 0.43 4 EP 997 14717 1.07 15.2 1.2 0.79 5 PP 2359 11030 4.27 7.0 1.4 1.95 6 EP 1771 26577 1.13 6.2 1.3 0.74 7 EP 1942 18683 1.97 7.2 0.8 0.62 8 PP 1752 8697 6.17 9.6 0.9 1.00 edition Professional) software of StatPoint Inc. The academic license of it is possessed by the Institute of Mining Engineering at Wroclaw University of Technology. As a result of multiple linear regression with logarithmic transformation, the following form of demanded relationship was chosen: Stretch strength TS [MPa] R τ = 2, 434( t )0,525 M g 0.99 Mt (4) Correlation coefficient of this function in relations with the test results was R2=90.6%, while its corrected value was 86.9%. Since the obtained P value in the table of variance analyze ANOVA was lower than 0.05 (P-value=0.0027), the statistical dependence between the variables at the 95.0% level of confidence was identified. The chart of function (4) is presented on figure 6. Fig. 6. The relationship of the dependent variable τ from independent variables Mt/Rt and Mg (4) The tests showed that the strength properties of belts being connected and the adhesive rubber have the essential impact on the magnitude of stresses in the adhesive joint. Maximal stresses occurring on the outside contacts of the splice plies may differ even several times depending on the elasticity modulus of the belt and modulus of the rubber, as it is showed on the example charts for splices of EP100/4 i PP1000/4 belts on figure 5, The approximate value of maximal stresses in the adhesive joint may be evaluated using the formula (5): τ = 2, 4 Rt M g Mt (5) 3. Tests of fatigue life of splices In the adhesive joint of splice subjected the cyclic fatigue loads, the heat is accumulated and therefore its temperature increases. The increase of temperature depends on the size and frequency of loads as well as on the type of the joint rubber and properties of the belt textile plies. Even at the same load conditions the temperature difference between individual splices may be up to tens of degrees. In such case comparing the test results for different types of splices is not possible. During the fatigue test it was assumed that the temperature of adhesive joint of splices being tested cannot exceed the range of 23°C ± 2°C. It was experimentally established that the above range of temperatures is possible to obtain under the following test conditions: -- Range of load form 5% to 20% of the belt strength -- Frequency of loads – 0.3 Hz -- Sinusoidal characteristics of loads -- Ambient temperature – 18°C Temperature of joint was controlled using the pyrometer. The tested samples had a shape of small scull which was 50mm wide in the tested part and 100 mm wide in holding part. The total length of samples was 1150mm. After the analyze of test results of stress magnitude in the adhesive joints, the central notches of splices were shorten to 150mm. Outside notches had the standard length. The experiments were performed on the machine for dynamic tests of HC-25type, produced by ZwickAmster. As a criterion for evaluation the fatigue life of splices number of fatigue cycles causing the splice delamination equal 3mm. Basic parameters of tested splices and numbers of fatigue cycles obtained during the tests are showed in table 2. The total of 12 samples of splices were tested. The number of Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 173 Science and Technology Table 2. Results of fatigue tests of splices Item Material type of plies 1 PP 1356 4639 6.3 440 2 EP 1050 12665 11.4 212 000 Tensile strength belt Rt [kN/m] 3 EP 997 14717 14.6 359 000 EP 1771 26577 9.2 145 000 5 EP 1942 18683 7.5 8600 6 PP 1752 8697 12.1 1900 LC = 12, 3362( M t 1,90881 1.92878 ) T Rt (6) The correlation coefficients received for the relationship (6) are: R2=94.7% and the corrected value is 91.2%. Figure 7 presents the relationship described by the equation (6). Fig. 7 Dependence of the number of cycles LC of the independent variables Mt/Rt and T (6) 174 Number of fatigue cycles after which the splice is delaminated LC 4 fatigue tests showed in the table is the average for two tested samples. Table 2 contains also the properties of splices, which according to the previous analyze, had the essential impact on their fatigue life. The T adhesive strength between connected plies was determined according to ENISO 252:2008 standard. Results of fatigue tests listed in table 2, were analyzed to present the relationships between number of cycles and Rt, Mt, Mg and T variables. It was assumed that number of cycles may depend on 3 variables i.e. Rt/Mt, Mg and T and in STATGRAPHICS software the procedure of multiple regression model selection was used for them. It turned out that the best, from the point of corrected value of R2, is the model based on two variables Rt/Mt, and T (R2=95.04%, and its corrected value is 91.74%). As a result of non-linear multiple regression analyze, the following relations for number of cycles LC was obtained: Adhesive strength of joint T [N/mm] Elasticity modulus of belt Mt [kN/m] For practical purposes relationship (6) may be simplified to the form (7): LC = 8, 05( Mt 2 2 ) T Rt (7) The correlation coefficients received for relationship (7) are: R2=94.6% and its corrected value is 94.6%. 4. Summary The test results presented in the paper showed how the stresses behave in the adhesive joint along each notch of multiply splices of conveyor belts. It was identified that extreme values of these stresses occurring at splice joints depend mainly on elasticity modulus of the belt being connected, its strength and modulus of the adhesive rubber. Basing on the calculated relationship (4 or 5) producers of conveyor belts and materials used to join them, can select their parameters to have the maximal stresses in the adhesive joint at the level of 0.5 MPa, under the belt load amounting 15 % of its strength. Fatigue tests of splices demonstrated that the essential impact on its durability has Rt/Mt the unit elasticity modulus of the jointed belts, and T adhesive strength of rubber used to join the plies. Tests enable to define the relationship (6, 7) which allows to calculate the fatigue life of splice, measured in number of fatigue cycles causing the start of delamination of splice joints under the cyclic loads within the range from 5% to 20% of belt strength. It was stated that depending on the strength parameters of jointed materials, the boundary number of fatigue cycles may be from several hundred to several thousands, what among other things, explain the reason of ungluing of numerous splices in working practice. The deduced formulas calculating the number of LC fatigue cycles enable to select very simply, the belts and adhesive materials properties to obtain the high fatigue life of splices and thus their better reliability. Conducted research has indicated significant parameters affecting the strength and durability of multi-plies conveyor belt splices. It seems advisable to test using developed method more splices having different properties and construction. This will clarify any of the models shown in the article and allow on more accurate prediction of fatigue life of belt splices. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Acknowledgements: The research work financed with the means of the National Center for Research and Development (Poland) in the years 2009-2012 as a development project. References 1. 2. 3. 4. 5. Błażej R, Hardygóra M. Modelling of shear stresses in multiply belt splices. Bulk Solids Handling. 2003, 23/4, 234-241. Kozhushko G.G., Kopnov V.A. Fatigue strength functions in shear loading of fabric conveyor belts. International Journal of Fatigue 1995, 17/8, 539-544. Mazurkiewicz D. Problems of numerical simulation of stress and strain in the area of the adhesive-bonded joint of a conveyor belt, Archives of Civil and Mechanical Engineering, 2009, IX/2, 75-91. Research Project No. 4 T 12A 06630. Elaborating the rules for selecting the strength parameters of conveyor belts and jointing them materials to obtain maximum strength and durability of joints. Mining Institute, Wroclaw University of Technology, Wrocław 2009. Research Project No. N R09 0019 06. Belt conveyor with increased economic and energetic efficiency constructed and operated according to the rules of sustainable development, Mining Institute, Mining Institute, Wroclaw University of Technology, Wrocław 2011. Prof. Monika Hardygóra Institute of Mining Engineering Wroclaw University of Technology Machinery Systems Division KGHM CUPRUM Ltd. - Research & Development Henryk Komander, MSc Ryszard Błażej, PhD (Eng.) Assoc. Prof. Leszek Jurdziak Institute of Mining Engineering Wroclaw University of Technology Machinery Systems Division Plac Teatralny, 50-051 Wrocław, Poland E-mail: [email protected] [email protected] [email protected] [email protected] Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 175 Article citation info: PARCZEWSKI K, WNĘK H. Make use of the friction coefficient during braking the vehicle. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 176–180. Krzysztof PARCZEWSKI Henryk WNĘK Make use of the friction coefficient during braking the vehicle Wykorzystanie przyczepności podczas hamowania pojazdu* In this publication is presented use the tyre-road friction during vehicle braking. Results presented in this publication are based on the road tests of the vehicle equipped in the anti-lock brake system (ABS). Two kinds of tests applied were carried out - the road tests of vehicle making the manoeuvre of braking on the straight section of the road and on the curve of the road. The braking forces and the friction coefficients for the individual wheels of the vehicle were defined on the basis of road tests, including the border values of the friction coefficient. Keywords: friction coefficient, vehicle testing, braking vehicle, curvilinear track, slip of wheels, ABS. W publikacji przedstawiono zagadnienie wykorzystania przyczepności opony do nawierzchni jezdni podczas hamowania. Wyniki prezentowane w publikacji oparto na badaniach drogowych pojazdu osobowego wyposażonego w układ zapobiegający blokowaniu kół. Przeprowadzono dwa rodzaje prób stosowanych do badania wykorzystania przyczepności - badania pojazdu wykonującego manewr hamowania na prostoliniowym odcinku drogi oraz na łuku drogi. Na podstawie badań określono siły hamowania oraz współczynniki przyczepności dla poszczególnych kół pojazdu, w tym wartości graniczne współczynnika przyczepności. Słowa kluczowe: przyczepność, badania pojazdu, hamowanie pojazdu, tor krzywoliniowy, poślizg kół, ABS. 1. The introduction The tyres and the road surface condition were accountable for passing on forces from the vehicle to road during the motion of the vehicle. The values of forces transferred on the surface of the road depend on the parameters of vehicle and its motion. They are limited the friction forces on contact area of the tyre road. The phenomenon of friction occurs on the wheel contact area with the road and encloses all conditions and mechanisms be present during this co-operation. At present, vehicles are equipped in arrangements preventing locking wheels while braking (ABS) what limits the range of the changes to be in operation on the area of limited wheel slip of the braking system. The problem of the co-operation of the wheel with road were introduced in the work while braking the vehicles on the rectilinear road and on the curve of the road, the appointed border values of the friction coefficients and the variation of these coefficients got from road tests. 2. Tyre – road friction During the motion of the vehicle the mechanism of formation of forces on the contact area of the tyre with the road appear from contiguous and normal intensity of stress in the area of this contact area. Wheel load on the road are different in every place of contact area with the tyre and change both in the longitudinal and lateral direction. While rolling the problem of the assymetry of the trace of the co-operation comes still. Every unit of the tyre, being in the contact area, is responsible for the transfer of longitudinal and lateral forces. There are two primary mechanisms [1, 2, 3, 4, 5, 6, 10] responsible for formation of the friction forces between tyre and road: hysteresis and adhesion. The adhesion comes into rise on the surface of the adhesion the force in the result of intermolecular bonds between the gum of the tread and the aggregate in the road surface. This influence is reduced with the presence of dirts or water in the area of contact. The mechanism of the bulk hysteresis comes into being in the result of the loss of energy while deforming the gum on agregate in the road. The friction comes into being in the order of this mechanism he is not „so affected” on dirts and the presence of water. The motion of the vehicle can be divide on compliant with the longitudinal axis of the vehicle and in perpendicular direction to this axis. The tyre-road friction can be describeed using the coefficient of adhesion (the ratio of the tyre-road friction force to the wheel load force) [2]. The coefficient of adhesion is understood, as the relation of maximum contiguous resultant force transfers by wheel to the load force working on this wheel. The temporary coefficients of adhesion were marked during the analysis using relationship: µ= W FZ The tire-road friction forces enclose together force transferred on the surface of the road in longitudinal direction XK and (*) Tekst artykułu w polskiej wersji językowej dostępny w elektronicznym wydaniu kwartalnika na stronie www.ein.org.pl 176 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology μm - friction coefficient (the maximum value of the relative friction force which can be got in given con ditions), v - drive velocity, R - radius of the track, g – acceleration of gravity (9.81 m/s2). During the manoeuvres of speeding up or drive with the steady velocity (on the flat road) the whole wheels frictions is used relatively seldom. The full use of friction is more considerably often in cases of the braking manoeuvres on the straight road or on the curve track of the road, particularly during realize manoeuvres on the wet road, covered by snow or icy. The use of anti-lock brake systems (ABS) limits the wheels slip. This will be result in the limitation of braking forces acting on the individual wheels of the vehicle. The exploration of use of the friction of the vehicle during braking were introduced below. Fig. 1. Mechanisms of tyre-road friction lateral direction YK. The resultant force W is limited the friction force of the wheel to the road surface Fμ. W= X K2 + YK2 and W ≤ Fµ Considering individual wheels separately, we can assign the border values of the friction forces which can be transferred to road surface. The above mentioned relationship will simplify oneself during the vehicle motion on the straight, level section of the road because of the possibility of the omission of transverse forces. The whole wheel-road friction force can be used on braking in such case. In the case of the vehicle motion on the curvilinear track of the road, the influence of lateralis force is smaller if the radius of turn is larger. In the case, when on wheel acts simultaneously longitudinal and transverse force with a simplify [9] one can record relationship defining the friction coefficient as: 3. Exploration of the tyre-road friction of the vehicle. 3.1. The assumption to exploration. Two kinds of tests applied to the explore of use of the tyreroad friction were carry over - the testing of vehicle making the manoeuvre of braking: on the straight road section and on the curvelinear track. µ = µ x2 + µ y2 where: μx - the coefficient of longitudinal friction, μy - the coefficient of transverse friction. While braking on the curve of the road the possible to use friction force in the longitudinal direction were limited by the centripetal force depend on the drive velocity and radius of turn. From this regard only part of the friction force can be used on braking the vehicle. The analysis the motion of the driving vehicle can mark what part of the friction coefficient can be used on braking on the circular track: where: v2 µh = µm 2 − g ⋅ R μh - part of the friction coefficient used on braking the vehicle, Fig. 2. The tracks of tests: a) braking on the straight section of the road, b) braking on the curve of the road From the safety considerations the exploration was made on the dry and clean aggregate surface. The sensors were used to exploration thrusts installed in the brake system, head to the measurement of the longitudinal and lateral velocity, the sensor of force on the pedal of the brake, sensors to measure accelerations of the vehicle in directions X and Y, sensors to the measurement of the angular speeds of the motion of the car body and sensors to measurement of the turn angle and moment on the steering wheel [7, 8]. The weight of the vehicle resulted from his own weight, weight of measuring equipment and a driver. Thus when the vehicle decelerate during braking load is transfered from the rear to the front axle in proportion to acceleration. This results in the change of the border friction forces and in the effect of the use of the anti-lock brake system (ABS), the limitation of the braking force generated through brakes in- Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 177 Science and Technology dividually for any wheel. The loads of the vehicle wheels were calculated on the basis of the measurements of the location of the vehicle centre of gravity, and longitudinal and lateral forces resulting from the motion conditions. The change of the location of the centre of gravity resulting from the inclination of the vehicle was not considered to calculate the forces of the load on respective wheels. Braking forces for individual wheels were assigned on the basis of measured pressure in the brake system and the geometrical parameters of brakes. The influence of the wheels inertia was considered on braking forces caused on the change of their rotative velocity. 3.2. Test of braking on the rectilinear section of the road. First test was carried out on the rectilinear section of the road. The driver keep up for the rectilinear direction of the track. He pressed on the pedal of the brake after the obtainment of the suitable velocity. The force of the pressure on the brake pedal assured to be active the anti-lock brake system (ABS). Fig. 5. The process of braking forces acting on the respective wheels of the vehicle The force of the pressure on the pedal assured the working of the system ABS. On figure 6 was introduced the track vehicle motion got on the drive test. Below were showed courses of loads changes of wheels (fig. 7), on the next graphs were showed forces acting on respective wheels of vehicle (fig. 8). Fig. 3. The process of the driving velocity of vehicle during the test of braking on the rectilinear section of the road Fig. 6. The track of the vehicle motion during the test of braking on the curvelinear road Fig. 4. The process of the loadings of the vehicle wheels during the test of brakeing on the rectilinear section of the road On graphs were described visible braking load of front wheels and the clear difference of the quantity of braking force at the front and rear axis. Certain translation in operation of the brakes of right and left wheels were result from the inhomogeneity of background and small asymmerty of the loads of the vehicle. 3.3. Test of braking on the curve of the road. Second test was carried out on the curved section of the road. The driver provided for steering wheel in such way the vehicle drived on the circular track. After conquest about 15 m on the circular track, the driver pressed on the brake pedal. 178 Fig. 7. The process of the wheel loads during the test of braking on the curvelinear road On the figures were introduced visible changes of the loads on right and left wheels while braking on the curvilinear track. It can see also the clear difference of the quantity of braking forces the front and rear axis, corrected regard of the schedule of wheel loads and centripetal force acting on the vehicle. One can notice that the loads of the rear wheel left is close to the zero what is produce desired results the limitation of the brak- Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Fig. 8. The process of braking forces acting on the respective wheels of the vehicle ing pressure by anti lock system in the circuit of the brakes of rear wheels in the initial phaze of braking and the same fall of the braking forces to small values. In the case of braking on the curvelinear track of the road the level of used friction coefficient grow up from the beginning of braking to the maximum value together with with decrease of the drive velocity. The value of friction coefficient is larger for the front right wheel (with cornering load) than for left wheel. They stabilize the coefficients value after decrease of the velocity of the drive. The friction coefficients of rear wheels are clearly smaller in the initial stage of braking and they grow up to maximum values. Differences between the coefficients values for front wheels, result from the their inaccuracy of estimation caused omission of the inclination influence of the side car and from the considerable difference of the loads of the right and left side of the vehicle. The exploration of braking the vehicle on the curve of the road allowed to delimitation of border total coefficient of friction (fig. 11) appointed on the basis of the friction ellipse. 4. Analysis of tests results of the friction forces utilization Utilization of the friction forces of the vehicle wheels during the road tests of braking on the rectilinear section and on the curve of the road was calculated on the basis introduced above analyses and the results of road tests. On figures 9 and 10 was presented values appointed, the used coefficients of friction and the border values of these coefficients resulting from the conditions of the motion. Fig. 11. The border values of friction coefficients got during the test of braking on the curve of the road (the fricrion ellipse was marked the thick line) 5. Recapitulation and conclusions. Fig. 9. Coefficients of braking friction on the rectilinear section of the road. One can notice that in first case the maximum value of the used friction coefficient for front wheels oscillates around value 0.75, and is larger for rear wheels and oscillates around value 0.8. Fig. 10. The friction coefficients while braking on the curvelinear track of the road The exploration of the friction forces acting between wheels and surface of the road, showed that the friction force (while emergency braking on the rectilinear road) is used in the complete since the initial moment of braking, until to the stop of the vehicle. Uploading of the front axis and unloading of the rear axis produced desired results the clear differentiation of pressure in brake circuits what allows to complete use of the wheels friction forces. Small difference among the individual wheels of one axis, are results depends on the local conditions of friction and is generating by small inequality and dusty surface of road. In the case of braking on the curve of the road the limitation of longitudinal friction results from the occurrence of centripetal force. The system ABS (preventing locking the wheels while braking) does not allow to achieve large longitudinal force, assure suitable conditions on proceed lateral forces and keep of the stability of the vehicle motion. The correction arise from the motion on the curvelinear track of the road gets smaller together with from the drive velocity is smaller. The clear differentiation of the individual wheels loads, particularly the sides - right and left, it arise from the working of centripetal force. Asymmerty generated by the load the vehicle by the only driver additionally influence on the quantity of individual loads and unreeled forces braking. Similarly as while braking on the straight line section of the road, the considerable differences of pressure be presented in the circuits of brake front and rear wheels. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 179 Science and Technology The friction forces of front wheels is used in the complete, however in the case of the rear wheels full use of friction follows just near the lower velocities of the drive of which the rear left wheel loses the contact with the road temporarily and rear right put under load partly. The clear growth of braking force on rear wheels follows what causes the considerable enlargement of force braking after the crossing of the border speed where switch off the ABS system. This state was showed on drawing 11, on which also is presented the border values of the friction coefficient appointed from the ellipse of friction (got from road testing). 6. References 1. 2. 3. 4. 5. Andrzejewski R. Dynamika pneumatycznego koła jezdnego. Warszawa: WNT 2010. Arczyński St. Mechanika ruchu samochodu. Warszawa: WNT 1993. Fundowicz P. Droga hamowania na łuku drogi. Zeszyty Instytutu Pojazdów, Politechnika Warszawska, 2010; 1(77): 103-110. Gillespie T D. Fundamentals of vehicle dynamics. Warrendale: SAE Inc. 1992. Grzegożek W. Modelowanie dynamiki samochodu przy stabilizującym sterowaniu siłami hamowania. Kraków: Zeszyty Naukowe Politechniki Krakowskiej, Seria Mechanika, monografia 275, 2000. 6. Pacejka H B. Tire and vehicle dynamics. Warrendale: SAE 2006. 7. Parczewski K, Wnęk H. Wpływ niesprawności zawieszeń na stateczność ruchu pojazdu – porównanie badań symulacyjnych i pomiarów. Archiwum Motoryzacji 2006; 2: 159-169. 8. Parczewski K, Wnęk H. Wykorzystanie modelu samochodu do analizy ruchu pojazdu po torze krzywoliniowym. Eksploatacja i Niezawodność – Maintenance and Reliability 2010; 4: 37-46. 9. Prochowski L, Unarski J, Wach W, Wicher J. Podstawy rekonstrukcji wypadków drogowych. Warszawa: WKŁ 2008. 10. Smith R H. Analyzing friction in the design of rubber products and their paired surfaces. CRC Press 2008. Krzysztof Parczewski, PhD (Eng.) Henryk Wnęk, PhD (Eng.) Department of Internal Combustion Engines and Vehicles University of Bielsko-Biała ul. Willowa 2, 43-300 Bielsko-Biała, Poland e-mail: [email protected], [email protected] 180 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Article citation info: PIENIAK D, NIEWCZAS A, KORDOS P. Influence of thermal fatigue and ageing on the microhardness of polymer-ceramic composites for bio-medical applications. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2012; 14 (2): 181–188. Daniel PIENIAK Agata M. NIEWCZAS Paweł KORDOS Influence of thermal fatigue and ageing on the microhardness of polymer-ceramic composites for biomedical applications Wpływ zmęczenia cieplnego oraz starzenia na mikrotwardość kompozytów polimerowo – ceramicznych do zastosowań biomedycznych* Studies presented in this paper, concern polymer-ceramic composites applied in the conservative dentistry. The aim of the study was to evaluate a long-term impact of the humid environment and cyclic thermal loads on the microhardness of new silorane-based composites and two methacrylate-based composites. The composite samples were subjected to normal saline environment with cyclically variable temperatures (5°C and 65°C), using a special thermal shock simulator. Microhardness was measured with Vicker’s method before the fatigue test and after a series of 4000 thermal cycles. It is known that microhardness of silorane-based composite in opposite to methacrylate-based composites not decrease under the influence of cyclic thermal loads. It was found slight increase of microhardness under conditions of conducted tests. The ageing studies were also conducted consisting in microhardness evaluation of the composite samples in 6 months period. During that time the samples were kept in normal saline. The studies of hardness were carried out after each month of the exposure time. No long-term impact of normal saline environment with constant temperature on the microhardness of the studied materials has been noticed. Keywords: thermal fatigue, ageing, microhardness, dental composite fillings. Badania prezentowane w niniejszej publikacji dotyczyły kompozytów polimerowo-ceramicznych stosowanych w stomatologii zachowawczej. Celem pracy była porównawcza ocena wpływu długotrwałego oddziaływania wilgotnego środowiska oraz cyklicznych obciążeń cieplnych na mikrotwardość nowego kompozytu bazującego na siloranach oraz dwóch tradycyjnych kompozytów bazujących na związkach metakrylanu. Próbki z kompozytów poddano oddziaływaniu środowiska soli fizjologicznej o cyklicznie zmiennych temperaturach (5°C i 65°C) wykorzystując specjalny symulator szoków termicznych. Wykonywano pomiary mikrotwardości metodą Vickersa przed rozpoczęciem testu zmęczenia cieplnego oraz po serii 4000 cykli termicznych. Wykazano, że w przeciwieństwie do tradycyjnych kompozytów stomatologicznych mikrotwardość kompozytu bazującego na siloranach nie zmniejsza się pod wpływem cyklicznego oddziaływania szoków termicznych odpowiadających warunkom fizjologicznym jamy ustnej. W warunkach przeprowadzonych badań stwierdzono nieznaczny wzrost tej mikrotwardości. Przeprowadzono również badania starzeniowe polegające na ocenie mikrotwardości próbek kompozytów przez okres 6 miesięcy. W tym okresie czasu próbki przechowywano w soli fizjologicznej. Pomiary mikrotwardości wykonywano po każdym miesiącu ekspozycji. Wykazano, że długotrwałe oddziaływanie środowiska soli fizjologicznej w warunkach stałej temperatury nie zmienia mikrotwardości żadnego z badanych materiałów. Słowa kluczowe: zmęczenie cieplne, starzenie, mikrotwardość, kompozytowe wypełnienia stomatologiczne. 1. Introduction Polymer-ceramic composites are commonly applied as dental fillings materials. They are used due to their aesthetic properties (wide range of colours) and good mechanical properties. Polymer composite materials based on the methacrylate compounds consist of the matrix, which is a light-cured resin (up to 40% of the structure volume) and inorganic micro- or macro-nanofillers in most cases based on the silicon compounds (above 60% of the structure volume) [2,11,17]. Additionally, composite is composed of photo-initiators and pre-adhesive agents. A short characteristic of the polymer composite compounds is given in table 1. Recently, in dental practice more and more the new siloranebased composite are used. A silorane-based composite has been introduced with distinctive polymerization characteristic to reduced polymerization shrinkage. The silorane matrix is formed by opening-ring during polymerization process. The silorane molecule represents a hybrid built-up of siloxane and oxiran (*) Tekst artykułu w polskiej wersji językowej dostępny w elektronicznym wydaniu kwartalnika na stronie www.ein.org.pl Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 181 Science and Technology Table. 1. Characteristics of the selected polymer-ceramic composites applied in dentistry. Indication Name of the compound Characteristics Organic composite phase (polymer) Bis-GMA Bisphenol A-Glycidyl Methacrylate Bis-GMA monomers have a large molecular weight and they undergo free-radical polymerization creating polymer rich in cross bonds and initiators. TEGMA Tetraethyleneglycol Dimethacrylate These are dissolving monomers, they make composite material less viscous, their kontent influence also the level of polymerization shrinkage. UDMA Urethane Dimethacrylate - PCDMA polycarbonate dimethacrylate Inorganic composite phase - This chase consists of among the others: lithium aluminium silicates, crystalline quartz, barium-aluminium-barium-silicon glass, strontium-barium-aluminium-fluorosilicate glass, silica, silanized lime Based on ethe smolecules size inorganic phase can be divided into macro – (1 -30 μm) of qartz, glass, ceramic or micro-fillers (0,007 do 0,04 μm) prepared of silicon dioxide or in a different way – prepolymerized as a result of technological performance (1-200 μm), agglomerated (1-25 μm) or spheroidal (20-30 μm). Photo-initiator CQ camphorquinone structural moieties [15]. The silorane-based composite opens new vistas in reduction of marginal microleakage phenomenon [5,15,19]. Based on the literature [14] it is known that siloranebased composite has very good mechanical properties: high impact resistance, flexural strength and satisfactory hardness. Whereas the stability question is unknown in conditions logterm influence humidity environment and cyclically thermal fatigue correspond to oral cavity. Therefore, authors undertake a study of comparison the influence of ageing and thermal fatigue on microhardness new silorane-based composite and properties two “traditional” methacrylate-based composites. Microhardness is a composite property, which is correlated with resistance to wear, also in case of thermal fatigue [3]. Investigations of microhardness allow evaluating mechanical properties of the composite. As it has been demonstrated in [13] there is a strong correlation between composite microhardness and elasticity modulus values, photo-polymerization depth, and the strongest with a polymerization shrinkage degree. In the paper concerning relation between physical-mechanical properties of the polymer composites and their application [9,16] a relationship between composite microhardness and degree of its wear in in vitro simulation conditions has been shown. Additionally, a correlation with the degree of composite filler conversion has been demonstrated [6]. Microhardness studies can be also used to evaluate a local gradient of photo-polymerization, which is a specific homogeneity of composite in the area of impact of the lamp light spectrum [8,18], influence of polymerization time and the type of the lamp’s light. It can be also applied as a measure of residual mechanical properties in the ageing and fatigue studies. Undertaking in vitro thermal fatigue simulation studies of the mechanical tooth-composite filling system, the loads conditions reflecting physiological conditions in the human oral cavity should be ensured. The following parameters should be controlled: temperature of the operating liquid (artificial saliva 182 Fotoinitiators are used to polymerization by the generation of free radicals (in case of light photo-polymerization with wave length of approx. 400-500 nm, blue or violet light or normal saline), retention time of the operating liquid in the container with samples, or the studied sample in the container with operating liquid, as well as number of load cycles (thermal shocks). In the previous studies different assumptions have been made with regards to the experimental parameters. Lower operating liquid temperature applied in the experiments, was between 2 and 24ºC [7], whereas heated liquid temperature was in a range of 45ºC [4] and 60 ºC [20]. Retention time of the liquid in the container with samples was from 15 even up to 180 seconds, while number of cycles varied from 25 to 1 million thermal cycles [1,7]. Currently, most often the following experimental parameters are assumed: • Cooled operating liquid temperature − 5 ºC, • Heated operating liquid temperature from 55ºC to 65 ºC, • Retention time of the operating liquid in the container with samples – 30 seconds, • Number of thermal cycles – from a few up to a few thousands. 2. Materials and methods In both conducted tests of thermal fatigue and ageing, the same composites applied in stomatology were considered. The commercial methacrylate-based materials, such as: Ice (SDI), Venus (Heraeus) and new silorane-based Filtek Silorane (3M ESPE) – table 2. The examples of SEM analysis results are presented in figure 1. There are visible molecules of composite polymer phase (larger), and also molecules of inorganic filler (smaller). The latter ones are the molecules with more regular shape, often close to spherical, with a similar size of grains. From the selected materials disk shape samples with 14 mm diameter and 1 mm thickness were made. Photo-polymerization Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Table 2. List of the studied composites Material Type Manufacturer Ice nanohybrid SDI Venus microhybrid Heraeus-Kulzer Filtek Silorane silorane 3M ESPE Resin Filler content (wt%) Size of filler molecules multifunctional methacrylic ester 77,5 inorganic filler 0,04 – 1,5 µm bisphenol-A glycidyl methacrylate (Bis-GMA) trietylen glycol dimethacrylate (TEGMA) 80 barium glass inorganic filler 0,04 -1 µm 76 inorganic filler 0,04 – 1,7 µm silorane a c b Fig. 1. SEM images of the sections of the studied materials: a) Filtek Silorane, b) Ice, c) Venus Fig. 2. Production of samples made of light-cured composites: 1 – lamp, 2 – exposed composite sample, 3 and 5 – plexiglass plates, 4 – steel form process was conducted with the use of halogen lamp as shown in figure 2. The exposure time of the samples was 40 seconds, according to materials manufacturers’ recommendations. In these studies a thermal shocks simulator was applied, designed in order to investigate dental materials. The device was made by the staff of the Mechanical Department of Technical University in Lublin in cooperation with Medical University in Lublin. The thermal shocks simulator (fig. 3) consists of the microprocessor control system and hydraulic system. The device enables creation of thermal shocks in the samples placed in the measuring container located in the simulator. Operation of the device consists of the cyclic pumping in and out of the operating liquid from the measuring container. The container is alternately filled with heated (65°C) or cooled operating liquid (5°C) from two independent temperature conditioning systems [11]. Fig. 3. Experimental stand for durability studies with specific thermal shocks module components: 1 – micro-processor control system, 2 – control valves, 3 – chewing simulator, 4 – peristaltic pomp, 5 – cooling thermostat, 6 – heating ultra-thermostat Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 183 Science and Technology Time of the subsequent procedures performance within each thermal shocks cycle was programmed and repeatable. Retention time of the cooled and heated liquid was 30 seconds, time pumping in and out of operating liquid was 10 seconds (fig. 4). Ageing studies were also performed (for 6 months period), based on micro-hardness evaluation of composites as a function of exposure time in saline. One of the objectives of the studies was to obtain a reference sample for the comparison with the results obtained in thermal fatigue test. 3. Studies results Results of the performed fatigue tests with the use of thermal shocks simulator are shown in fig. 5 and table 3. The results of ageing studies are presented as a box plot in figure 6. 4. Discussion Fig. 4. Thermal shock algorithm with a single pumping of the operating liquid Microhardness studies were conducted according to Vicker’s method using Futertech FM 700 (Future-tech Corp. Japan), with load of 50g. A specified penetration time of the indenter was 15 seconds. The measurements were taken at ten points of the samples’ surfaces. Measuring coordinates were determined in order to include possibly a whole sample surface. They were identical for all samples. The studies were conducted both at the exposed (lc) and non-exposed (nlc) samples surface. Microscopic analyses of the composites structure were carried out in the Laboratory of Electron Microscopy of the Catholic University in Lublin. Microscopic observations were conducted by means scanning electron microscope (SEM) by Zeiss. The effect of the thermal influence on the decrease of microhardness methacrylate-based composites Venus and Ice. The highest drop of microhardness after 4000 thermal cycles (4kTC) for Venus material was noticed, which can be seen in the graph (fig. 5c) and was confirmed by the t-Student test results (tab. 4). T parameter values for this material were the highest in the most comparisons. The influence of the thermal fatigue on the faster weakening of non-exposed surface (nlc) than the exposed one (lc) of Venus material has been demonstrated. Investigations of Ice material confirmed the influence of thermal shocks on the micro-hardness decrease. It was not as high as in case of Venus material, however it was visible. This relationship was confirmed by t-Student test. The highest values of t parameter were obtained in comparison of the results before and after thermal fatigue test performance. In case of FSi (Filtek Silorane) material a decline in average microhardness values after 4000 TC cycles was not shown. However, a slight increase was noticed. Statistically significant increase of micro-hardness at the exposed surface was also confirmed by t-Student test results. However, in case of this material impact of 4000 thermal cycles was the least visible and it resulted in different consequences that the observed for the two other materials. That fact is connected with different phase composition Filtek Silorane, a new silorane-based composite. Similar observation was presented in the literature [14]. Analysing the results of ageing studies it can be concluded that despite a slight micro-hardness fluctuations during the test Table 3. Results of microhardness. Descriptive statistics. TC – thermal cycles, lc – expose surface, nlc – non-exposed surface Group Valid N Mean Median Minimum Maximum Std.Dev. Coef.Var. Filtek Silorane 0TC lc 40 51,97275 52,20000 48,80000 57,00000 1,791623 3,447235 Filtek Silorane 0TC nlc 40 49,38750 49,71000 42,76000 55,48000 2,742934 5,553903 Filtek Silorane 4kTC lc 40 54,80075 55,24500 47,42000 61,31000 3,102902 5,662153 Filtek Silorane 4kTC nlc 40 50,15750 50,02500 42,09000 56,24000 3,091503 6,163590 Ice 0TC lc 40 51,95350 51,86500 45,53000 57,71000 2,736424 5,267064 Ice 0TC nlc 40 49,96100 49,84000 45,18000 54,28000 2,700346 5,404909 Ice 4kTC lc 40 47,23800 47,22500 41,92000 51,75000 2,737509 5,795141 Ice 4kTC nlc 40 44,58550 44,99500 35,28000 52,56000 4,141693 9,289327 Venus 0TC lc 40 48,60775 48,39500 43,94000 55,73000 2,613053 5,375794 Venus 0TC nlc 40 48,88650 49,11000 42,90000 54,27000 2,974151 6,083787 Venus 4kTC lc 40 37,78125 37,88000 33,82000 41,77000 1,973859 5,224440 Venus 4kTC nlc 40 35,14625 34,96000 30,37000 41,45000 2,322400 6,607817 184 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology a b c Fig. 5. Stemplot of micro-hardness of the studied materials after fatigue tests: a) Filtek Siloran, b) Ice, c) Venus Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 185 Science and Technology a b c Fig. 6. Relationship of micro-hardness of the studied materials and ageing time in saline: a) Filtek Siloran, b) Ice, c) Venus 186 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 Science and Technology Table 4. T-Student test results of microhardness. TC – thermal cycles, lc – exposed surfaces, nlc – non-exposed surfaces No Gr. 1 vs Gr. 2 T df p 1 FSi 0TC lc vs. Fsi 0TC nlc 4,99 78 0,000004 2 FSi 0TC lc vs. Fsi 4kTC nlc 3,21 78 0,00191 3 FSi 4kTC lc vs. Fsi 0TC lc 4,99 78 0,000004 4 FSi 4kTC lc vs. Fsi 0TC nlc 8,27 78 0 5 FSi 4kTC lc vs. Fsi 4kTC nlc 6,7 78 0 6 FSi 4kTC nlc vs. Fsi 0TC nlc 1,18 78 0,242252 7 Ice 0TC lc vs. Ice 0TC nlc 3,28 78 0,001563 8 Ice 0TC lc vs. Ice 4kTC lc 7,71 78 0 9 Ice 0TC lc vs. Ice 4kTC nlc 9,39 78 0 10 Ice 0TC nlc vs. Ice 4kTC lc 4,48 78 0,000025 11 Ice 0TC nlc vs. Ice 4kTC nlc 6,88 78 0 12 Ice 4kTC lc vs. Ice 4kTC nlc 3,38 78 0,001138 13 Venus 0TC lc vs. Venus 4kTC lc 20,91 78 0 14 Venus 0TC lc vs. Venus 4kTC nlc 24,35 78 0 15 Venus 0TC nlc vs. Venus 0TC lc 0,45 78 0,657331 16 Venus 0TC nlc vs. Venus 4kTC lc 19,68 78 0 17 Venus 0TC nlc vs. Venus 4kTC nlc 23,03 78 0 18 Venus 4kTC lc vs. Venus 4kTC nlc 5,47 78 0,000001 period (fig. 6), an unequivocal impact of the exposure in saline on the changes in micro-hardness have not been demonstrated. 5. Conclusions: 1. 2. Changes of microhardness for FSi (Filtek Silorane) material due to thermal shocks were insignificant. The impact of thermal fatigue on the microhardness decrease of the methacrylate-based composites Venus and Ice has been demonstrated. 3. 4. No impact of ageing in humid environment on the microhardness of the studied materials has been noticed. It has been concluded that the applied method of the studies can be useful for the preliminary qualification of the operational usability of the dental composite materials. References 1. Achilias D S, Karabela M M, Sideridou I D. Thermal degradation of light-cured dimethacrylate resins Part I. Isoconversional kinetic analysis. Thermochemica Acta 2008; 472: 74–83. 2. Andrzejczuk M, Lewandowska M, Kurzydłowski K J. Właściwości mechaniczne światłoutwardzalnych kompozytów zbrojonych mikro- i nanocząstkami. Kompozyty (Composites) 2005; (5)1: 75-79. 3. Assuncao W G, Gomes E A, Barao V A R, Barbosa D B, Delbenc J A, Tabata L F. Effect of storage in artificial saliva and thermal cycling on Knoop hardness of resin denture teeth. Journal of Prosthodontic Research 2010; 54: 123–127. 4. Chan K C, Swift E J. Marginal seal of new generation dental bonding agents. Journal of Prosthet. Dentistry 1994; 72: 420–423. 5. Eick J D, Kotha S P, Chappelow C C, Kilway K V, Giese G J, Glaros A G. Properties of silorane-based dental resins and composites containing a stress-reducing monomer. Dental Materials 2007; 23: 1011-1017 6. Ferracane J L. Correlation between hardness and degree of conversion during the setting reaction of unfilled dental restorative resins. Dental Materials 1985; 1: 11—14. 7. Gale M S, Darvell B W. Thermal cycling procedures for laboratory testing of dental restorations. Journal of Dentistry 1999; 27: 89–99. 8. Geis-Gerstorfer J. In vitro corrosion measurements of dental alloys. Journal of Dentistry 1994; 22: 247-51. 9. Heintze S D, Zellweger G, Zappini G. The relationship between physical parameters and wear of dental composites. Wear 2007; 263; 1138–1146. 10. Hill T, Lewicki P. Statistics: methods and applications: a comprehensive reference for science, industry and data mining. wyd. StatSoft 2006. 11. Joyston-Bechal A, Kidd E, Joyston-Bechal S. Essentials of dental caries: the disease and its management. 2nd ed. wyd. Oxford University Press, Oxford 1998. Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 187 Science and Technology 12. Kordos P, Hunicz J, Niewczas A. The station designed for accelerated fatigue tests of dental materials. Ekspoloatacja i Niezawodnosc – Maintenance and Reliability 2009; 1(41): 63-69. 13. Li J, Li H, Foka A S L, Watts D.C. Multiple correlations of material parameters of light-cured dental composites. Dental Materials 2009; 25: 829–836. 14. Lien W, Vandewalle K S. Physical properties of a new silorane-based restorative system. Dental Materials 2010; 26: 337-344 15. Moszner N, Salz U. New development of polymeric dental composites. Prog Polym Sci 2001; 26: 535-536 16. Sanders-Tavares da Cunha Mello F, Feilzer A J, de Gee A J, Davidson C L. Sealing ability of eight resin bonding systems in a Class II restoration after mechanical fatiguing. Dental Materials 1997; 13: 372-376. 17. Szafran M, Rokicki G, Bobryk E, Szczęsna B. Effect of filler’s surface treatment on mechanical properties of ceramic-polymer composites used in dentistry. Kompozyty (Composites) 2006; 6(3): 78-82. 18. Versluis A, Tantbirojn D, Douglas W H. Do dental composites always shrink toward the light? Journal of Dental Research 1998; 77: 1435–45 19. Weinmann W, Thalacker C, Guggenberger R. Siloranes in dental composites. Dental Materials 2005; 21: 68-74 20. Yoshida K, Matsumura H, Atsuta M. Monomer composition and bond strength of light-cured 4-META opaque resin. Journal of Dental Restoration 1990; 69: 849–851. Daniel Pieniak , PhD (Eng.) Faculty of Transport and Computer Science University of Economics and Innovation in Lublin Mełgiewska 7-9 str., 20-209 Lublin e-mail: [email protected] Agata M. Niewczas, DMD PhD Department of Conservative Dentistry Medical University of Lublin, Karmelicka 7 str., 20-081 Lublin, e-mail: [email protected] Paweł Kordos, PhD (Eng.) Faculty of Transport and Computer Science University of Economics and Innovation in Lublin Mełgiewska 7-9 str., 20-209 Lublin e-mail: [email protected] 188 Eksploatacja i N iezawodnosc - Maintenance and Reliability Vol.14, No. 2, 2012 2012 International Conference on Quality, Reliability, Risk, Maintenance, and Safety Engineering QR2MSE 2012 June 15-18, 2012 · Chengdu, Sichuan, China http://www.qr2mse.org CALL FOR PAPERS ORGANIZERS University of Electronic Science and Technology of China (UESTC) Maintenance Professional Committee, China Ordnance Society SPONSORS IEEE Reliability Society, Society The Korean Reliability Society, Society Reliability Engineering Association of Japan, Japan European Safety and Reliability Association (ESRA), CRC for Integrated Engineering Asset Management, National Natural Science Foundation of China THEME The 2012 International Conference on Quality, Reliability, Risk, Maintenance, and Safety Engineering (QR2MSE 2012) aims to provide an international forum for exchange of innovative ideas, research results, and applications of reliability and quality tools in design, manufacturing, and operation and maintenance of engineering systems. Presenters may address any aspect of quality, reliability, risk, maintenance, and safety. Participations from academia, academia industry, industry and government are all welcome. welcome Papers dealing with case studies, studies reliability data generation, experimental results, best design practice, or effective maintenance solutions are of particular interest. All papers accepted will be included in the conference proceedings and IEEE Xplore. Every effort will be made to ensure that the conference proceedings be indexed by EI and ISTP. Selected papers will be published in special issues/sections of international journals. TOPICS OF INTERESTS (not limited to the following topics) • • • • • • • • • • • • • • • • • • • Quality engineering Quality assurance and cost issues Quality planning and measurements Total quality management techniques Experimental design for quality control Methodologies for quality control Reliability theory and application Reliability methodology in design Reliability testing and statistics y and safety y evaluation Product reliability Human factors and reliability Software reliability and testing Structural reliability Mechanism reliability Mechanical reliability Network reliability Electric power systems reliability Reliability tools for product development Risk assessment methods • • • • • • • • • • • • • • • • Survival data analysis Data collection and analysis Prognostics and health management Condition based maintenance Equipment management and maintenance e-Maintenance Maintenance engineering theory and application Reliability centered maintenance Software So a e maintenance ai e a ce Emergency maintenance Maintenance support modeling and simulation Maintainability engineering theory and application Contractor logistics theory and practice Robust design Optimal design Multidisciplinary design optimization SUBMISSION OF PAPERS Submit abstracts and papers via the conference website. Please check the conference website at www.qr2mse.org for more information and updates, or contact us at [email protected]. IMPORTANT DATES Full Paper Submission Deadline Full Paper Acceptance Notification Camera Ready Papers Due Feb. 29th, 2012 Mar. 31st, 2012 Apr. 30th, 2012 All submitted papers will be reviewed for merits and contents. Accepted papers will be published in the Conference Proceedings and IEEE Xplore. Outstanding papers will be considered for publication in a special issue of international journals. Dr inż. Andrzej Loska Instytut Inżynierii Produkcji Politechnika Śląska ul. Roosevelta nr 26, 41-800 Zabrze, Polska E-mail: [email protected] Uwagi o modelowaniu procesów eksploatacyjnych z wykorzystaniem technik scenariuszowych Słowa kluczowe: eksploatacja, utrzymanie ruchu, scenariusze eksploatacyjne, procesy eksploatacyjne, modelowanie Streszczenie. Celem artykułu jest zaprezentowanie wyników przeprowadzonych badań w zakresie sposobów wykorzystania metod modelowania procesów eksploatacyjnych w zastosowaniach praktycznych (przemysłowych). W pierwszej kolejności dokonano identyfikacji i klasyfikacji stosowanych w tym zakresie rozwiązań, zwrócono przy tym uwagę na problemy związane z modelowaniem procesów w aspekcie szerszego ujęcia organizacyjno-technicznego. W oparciu o przeprowadzone badania, które miały charakter zarówno podstawowy (teoretyczny), jak i przemysłowy, sformułowano problem badawczy. Jego istota sprowadza się do trudności wykorzystania modeli procesów eksploatacji w odniesieniu do funkcjonujących organizacji utrzymania ruchu przedsiębiorstw przemysłowych ze względu na ich „płaski” charakter. W oparciu o sformułowany problem badawczy zaproponowano sposób jego rozwiązania poprzez zastosowanie do modelowania procesów eksploatacji technik scenariuszowych. Następnie po dokonaniu przeglądu stanu wiedzy (literatury) w aspekcie budowy scenariuszy eksploatacyjnych, opracowano i przedstawiono szczegółowo procedurę prowadzenia badań, których efektem będzie opracowanie metodologii tworzenia scenariuszy eksploatacyjnych, z uwzględnieniem sposobu praktycznego ich wykorzystania. Artykuł jest wynikiem realizacji części badań w ramach projektu badawczego, finansowanego ze środków Narodowego Centrum Nauki (nr 5636/B/T02/2011/40 pt.: Wykorzystanie metod scenariuszowych w modelowaniu procesów eksploatacyjnych. 1. Wprowadzenie Zarówno w eksploatacji systemów technicznych, jak i w praktycznych zastosowaniach tej teorii, wiele uwagi poświęca się problemowi modelowania. Obecne zadania służb technicznych większości przedsiębiorstw przemysłowych wykraczają poza klasyczne ramy planowania, realizacji i rozliczania prac obsługowo-naprawczych, szczególnie w zakresie sposobów racjonalizacji i optymalizacji procesów decyzyjnych zarówno w krótkim, jak i dłuższym horyzoncie czasowym. W aspekcie szybkiego rozwoju strategii i metod związanych z zarządzaniem eksploatacją i utrzymaniem ruchu, a także stosowanych narzędzi informatycznych umożliwiających pozyskiwanie, gromadzenie i przetwarzanie dużej ilości różnorodnych danych opisujących obiekty techniczne oraz warunki ich funkcjonowania, konieczne i możliwe wydaje się prowadzenie badań eksploatacyjnych związanych przede wszystkim z opracowaniem sposobu podejmowania optymalnych decyzji strategicznych dotyczących między innymi technicznie, organizacyjnie i ekonomicznie uzasadnionych terminów realizacji prac obsługowo-naprawczych, wycofywania obiektów z eksploatacji, długoterminowych sposobów postępowania w ramach określonych filozofii eksploatacyjnych czy zakresów realizacji zadań. Wymaga to budowy odpowiednich modeli zarówno obiektów technicznych, zdarzeń, jak i realizowanych procesów eksploatacyjnych. Modele te powinny dobrze odzwierciedlać wybrane aspekty "rzeczywistości" eksploatacyjnej oraz muszą opierać się na fundamentach wynikających z teorii modelowania. Niniejszy artykuł stanowi kolejny etap badań prowadzonych przez autora w Instytucie Inżynierii Produkcji Politechniki Śląskiej nad możliwością wspomagania procesu podejmowania decyzji eksploatacyjnych w przedsiębiorstwach przemysłowych. Badania zmierzają do opracowania sposobu modelowania procesów eksploatacyjnych zarówno aspektów technicznych, jak również uwarunkowań nietechnicznych (np. organizacyjnych, ekonomicznych) realizowanych prac obsługowo-naprawczych, co pozwoli na wykorzystanie takich modeli w ujęciu praktycznym. 2. Praktyczne uwarunkowania klasycznych sposobów modelowania procesów eksploatacyjnych Przyjmuje się, że proces eksploatacyjny jest to zbiór uporządkowanych działań, realizowanych z udziałem obiektu i w odniesieniu do obiektu technicznego w trakcie jego eksploatacji. Przebieg poszczególnych działań jest warunkowany indywidualnymi cechami eksploatowanych obiektów oraz ich otoczeniem organizacyjno-technicznym, co skutkuje unikalnością poszczególnych procesów eksploatacyjnych. W celu wyodrębnienia i zdefiniowania rozpatrywanych w artykule problemów badawczych, konieczne jest dokonanie jednoznacznej identyfikacji i klasyfikacji procesów eksploatacji. Podstawą do takiej klasyfikacji może być sposób opisu (modelowania), warunkujący realizację szczegółowych zadań związanych z takim procesem oraz podejmowaniem decyzji dotyczących tego procesu. Według takiego kryterium klasyfikacji, można wyróżnić dwa podstawowe sposoby opisu (modele) procesów eksploatacji [7]: a. model procesu eksploatacji jako sekwencji zdarzeń, według którego działania obsługowonaprawcze są podejmowane i realizowane na podstawie informacji o określonych zdarzeniach występujących w różnych chwilach czasowych (rys. 1) Z11 Z0 Z21 T1 Z31 T2 T3 Z22 T4 t 0 t1 t2 t3 t4 Rys. 1. Model procesu eksploatacji jako sekwencja zdarzeń [7] Zij - zdarzenia eksploatacyjne, Ti - przedział czasu pomiędzy zdarzeniami, ti - chwile, w których występują zdarzenia b. model procesu eksploatacji jako ciągu stanów, według którego działania eksploatacyjne są podejmowane i realizowane na podstawie chwilowych stanów obiektu identyfikowanych w równych odstępach czasowych (rys. 2). St(0) St(1) ∆t St(2) ∆t St(3) ∆t ∆t t 0 t1 t2 t3 Rys. 2. Model procesu eksploatacji jako sekwencja zdarzeń [7] St(i) - identyfikowane stany techniczne, ∆t - przedział czasu pomiędzy chwilami identyfikacji stanu technicznego, ti - chwile identyfikacji stanu technicznego Rozpatrując model procesu eksploatacji jako sekwencji zdarzeń, możemy wyróżnić dwa typy możliwych do wystąpienia zdarzeń eksploatacyjnych: zdarzenia zamierzone i zdarzenia niezamierzone. Każdy typ zdarzenia warunkuje inny sposób postępowania, co przejawia się w odmiennych zasadach eksploatowania obiektów zawartych w ramach polityki eksploatacyjnej. Zastosowanie modelu procesu eksploatacji jako ciągu stanów do identyfikacji sposobu podejmowania decyzji eksploatacyjnych wymaga dokonywania okresowej (stałej w czasie) kontroli stanu technicznego obiektu, co stanowi podstawę decyzyjną realizacji określonych w czasie i zakresie zadań eksploatacyjnych oraz wymusza przyjęcie odmiennych w stosunku do poprzednich, procedur postępowania. Praktyka eksploatacyjna pokazuje, że najczęściej opis przebiegu procesów eksploatacyjnych odniesiony do konkretnego egzemplarza eksploatowanego obiektu wymaga jednoczesnego wykorzystania omówionych powyżej modeli. Przykładowo, obiekty techniczne, dla których prowadzone są działania związane z monitorowaniem stanu technicznego (czyli są związane z modelem procesu eksploatacji jako ciąg stanów), mogą ulegać uszkodzeniom (czyli zdarzeniom związanym z modelem procesu eksploatacji jako sekwencja zdarzeń). Przedstawiony powyżej sposób modelowania procesów eksploatacji, w praktyce przemysłowej przekłada się na opracowywanie i wdrażanie ogólnych i szczegółowych rozwiązań w postaci określonych strategii eksploatacyjnych. W szczególności, na podstawie zależności procesów eksploatacyjnych od sposobu podejmowania decyzji w eksploatacyjnych systemach organizacyjno-technicznych można stwierdzić, że najczęściej decyzje wynikają z wykorzystania jednej z poniższych strategii eksploatacyjnych: 1. strategii eksploatacji według uszkodzeń (ang. BM - Breakdown Maintenance), opartej na modelu procesu eksploatacji jako sekwencji zdarzeń niezamierzonych, 2. strategii eksploatacji według ilości wykonanej pracy (ang. PM - Preventive Maintenance), opartej na modelu procesu eksploatacji jako sekwencji zdarzeń zamierzonych, 3. strategii eksploatacji według stanu technicznego (ang. PDM - predictive maintenance), opartej na modelu procesu eksploatacji jako ciągu stanów. W praktyce przemysłowej, wiążąc poszczególne strategie z procesem decyzyjnym dotyczącym konkretnych obiektów technicznych (a niekoniecznie całego ciągu technologicznego), wszystkie trzy strategie tworzą politykę eksploatacyjną przedsiębiorstwa, przy czym jedna z tych strategii ma najczęściej charakter dominujący (stanowi podstawę decyzyjną w odniesieniu do większości lub najistotniejszych obiektów technicznych), a pozostałe zaś stanowią jej uzupełnienie. 3. Identyfikacja problemów badawczych i propozycja ich rozwiązania Przedstawione w poprzednim punkcie i opisane w literaturze modele procesów eksploatacyjnych umożliwiają rozwiązywanie istotnych praktycznych problemów dotyczących użytkowania i obsługiwania maszyn i urządzeń. Jednakże wymienione powyżej możliwości są obarczone określonymi ograniczeniami, wynikającymi z charakteru omawianych metod. Podstawowym ograniczeniem tego rodzaju jest możliwa do uwzględnienia w modelu liczba cech (parametrów), opisujących stan obiektu/sygnału diagnostycznego lub zdarzenie eksploatacyjne. Ograniczenie tego rodzaju wpływa w istotnym stopniu na dokładność identyfikacji konkretnego stanu lub zdarzenia. Trudne także jest uwzględnienie w modelach omawianego typu cech jakościowych oraz aspektów nietechnicznych (organizacyjno-ekonomicznych), które w znaczący sposób wpływają na realizację poszczególnych zadań eksploatacyjnych. Do najważniejszych ograniczeń opisanych wcześniej sposobów opisu procesów eksploatacyjnych można zaliczyć: • znaczne uproszczenia modeli procesów w odniesieniu do organizacji, w której funkcjonują obiekty techniczne stanowiące przedmiot modelowania, • "płaski” charakter modeli pokazujący relacje przyczynowo-skutkowe w sposób jednowymiarowy bez uwzględniania interakcji z szeroko rozumianym otoczeniem w postaci dodatkowych aspektów nietechnicznych (np. organizacyjnych czy ekonomicznych), • trudności z uwzględnieniem cech jakościowych oraz aspektów nietechnicznych (organizacyjno-ekonomicznych) w modelach omawianego typu, które w znaczący sposób wpływają na realizację poszczególnych zadań eksploatacyjnych (także w obszarze zarządzania eksploatacją i utrzymaniem ruchu maszyn i urządzeń). Przedstawione ograniczenia nie pozwalają zaspokoić potrzeb organizacji utrzymania ruchu przedsiębiorstw przemysłowych w zakresie efektywnego wykorzystania typowych modeli dla potrzeb optymalizacji procesów decyzyjnych w odniesieniu do funkcjonujących obiektów technicznych. Wychodząc z istniejącego stanu wiedzy w omawianej dziedzinie oraz przedstawionych powyżej przesłanek, można sformułować główny problem badawczy niniejszego artykułu: stosowane obecnie modele i metody modelowania wykazują istotne, wymienione powyżej ograniczenia, które nie pozwalają na efektywne ich wykorzystanie w optymalizacji procesów decyzyjnych w odniesieniu do eksploatowanych obiektów technicznych, a co za tym idzie nie stanowią skutecznego rozwiązania oceny i poprawy efektywności zarządzania eksploatacją i utrzymaniem ruchu. Rozwiązaniem problemu optymalizacji procesu decyzyjnego może być wykorzystanie metod scenariuszowych, czyli zastosowanie sposobów modelowania, wykorzystujących wielowątkowy obraz bieżącej rzeczywistości przy równocześnie możliwym spojrzeniu „wyprzedzającym” czas bieżący. Techniki scenariuszowe należą do metod prognozowania, które dotychczas były stosowane w obszarze nauk ekonomicznych do prognozowania gospodarczego i zarządzania strategicznego. W dziedzinie nauk technicznych, jak na razie nie zdobyły sobie większego uznania. Wśród kilku metod tworzenia scenariuszy najlepszą w tym przypadku wydaje się metoda zaproponowana przez H. Kahna, który określił ją mianem „pisania scenariuszy” (scenariowriting). Metoda ta polega na opisie zdarzeń i wskazania ich logicznego i spójnego następstwa w celu ustalenia, w jaki sposób rozwijać się będzie dany obiekt lub sytuacja. Przyjmuje się w tym przypadku określony punkt odniesienia, którym w przypadku zarządzania eksploatacją może być na przykład przeszły lub bieżący stan techniczny. Główny nacisk kładzie się na te zdarzenia i sytuacje, które mogą stanowić podstawę przyszłego stanu lub ciągu zdarzeń. W wyniku tego otrzymuje się zbiór możliwych do wystąpienia zdarzeń lub inaczej obrazów przyszłości. Obok wielu sposobów tworzenia scenariuszy z wykorzystaniem różnych metod, w omawianej metodologii wyróżnia się pewną liczbę typów czy wariantów scenariuszy. W odniesieniu do rozpatrywanej dziedziny możliwe, celowe i sensowne jest wykorzystanie dwóch takich typów, które odpowiadają potrzebom i możliwościom modelowania procesów eksploatacyjnych [23]. W pierwszej kolejności należy do nich scenariusz badawczy (rys. 3a), w którym na podstawie określonych hipotez określa się zdarzenia możliwe do wystąpienia (stanowiące zbiór efektów stanu/sytuacji bieżącej). Ten typ scenariusza odpowiada problemom planowania eksploatacji, które mają bezpośrednie przełożenie na definiowanie założeń strategii eksploatacyjnych. W specyficznych sytuacjach możliwe jest zastosowanie scenariusza antycypacyjnego (rys. 3b), w którym zidentyfikowane relacje umożliwiają określenie skutków zaistniałej sytuacji. Ten typ scenariusza będzie stanowił podstawę analizy zdarzeń przeszłych, w szczególności analiz awaryjności, których najważniejszym elementem jest określenie przyczyn zaistniałych sytuacji lub niepowodzeń w realizacji prac obsługowonaprawczych. a) b) Rys.3. Typy scenariuszy: a) scenariusz badawczy, b) scenariusz antycypacyjny [23] Jak widać z powyższego, technika budowy scenariuszy ma wiele wspólnego z technikami prognostycznymi. Wychodząc z aktualnej sytuacji (zbioru cech opisujących procesy eksploatacyjne), można wyróżnić te czynniki wpływające na sytuacje przyszłe, które są znane. Odniesienie tych czynników do najbliższej przyszłości powoduje nieznaczną ich zmianę, natomiast w zakresie okresów dłuższych zmiany te mogą być znaczne. Ma to swoje odzwierciedlenie zarówno w zmianach poszczególnych czynników, jak i wpływie zmian na poszczególne obiekty czy systemy. Poprzez wykorzystanie do modelowania procesów eksploatacji metod scenariuszowych, można pokazać sposób i zakres osiągnięcia możliwych alternatywnych scenariuszy opisujących skutki podjętych decyzji. W ramach dalszych badań proponuje się wykorzystanie scenariuszy zdarzeń i procesów eksploatacyjnych jako ujęcia metodologicznego, które daje nowe jakościowo możliwości w obszarze modelowania procesów eksploatacyjnych dla potrzeb skutecznego podejmowania decyzji/zarządzania eksploatacją w systemach technicznych. Kluczowe elementy metodologii scenariuszowej oraz specyfika obszaru eksploatacji systemów technicznych warunkują zbiór koniecznych działań (celów szczegółowych), których realizacja pozwoli na rozwiązanie sformułowanego problemu badawczego. Działania te obejmują: • zdefiniowanie potrzeby i podstawy generowania scenariusza/zbioru możliwych scenariuszy w oparciu o określone modele eksploatacyjne, które w tym przypadku mogą wynikać z kryteriów niezawodnościowych (np. zbioru wybranych ilościowych wskaźników eksploatacyjnych) • określenie wewnętrznej, formalnej struktury opisu scenariusza (identyfikacja zbioru parametrów, czyli składników ilościowych, oraz cech, rozumianych jako składniki jakościowe obrazu danej sytuacji/zdarzenia), • uzupełnienie scenariuszy sytuacji dotyczących obiektu jako takiego przez odwzorowanie jego otoczenia (czyli przez analogię, autor artykułu proponuje oprócz scenariusza opisać „scenografię”, w której scenariusz się dzieje), • rozwiązanie problemu praktycznego wykorzystania scenariuszy w planowaniu prac obsługowo-naprawczych, a jednocześnie optymalizacji procesów decyzyjnych dotyczących eksploatowanych systemów technicznych z uwzględnieniem zagadnienia wielowariantowości możliwych zdarzeń oraz symulacji zachowań obiektu w bliższej i dalszej przyszłości. 4. Znaczenie modelowania procesów eksploatacyjnych z wykorzystaniem technik scenariuszowych Problem wykorzystania technik scenariuszowych w eksploatacji występuje w literaturze incydentalnie. Jest to związane głównie z pojedynczymi opisami praktycznych rozwiązań scenariuszy do postępowania w wybranych sytuacjach awaryjnych (np. w [24]). Szczegółowa analiza literaturowa zarówno krajowa, jak i zagraniczna wykazała, że badania nad możliwością wykorzystania scenariuszy w odniesieniu do kompleksowych aspektów strategicznych eksploatowania obiektów technicznych są w Polsce prowadzone przede wszystkim w Instytucie Inżynierii Produkcji Politechniki Śląskiej. Ma to swoje odzwierciedlenie w publikacjach [8, 13, 16, 15, 18]. W zakresie problematyki poruszanej w artykule występują publikacje opisujące poszczególne odrębne elementy związane z tą tematyką, w szczególności: 1. modelowanie wybranych elementów eksploatacji, 2. metody i techniki scenariuszowe, 3. zarządzanie eksploatacją i utrzymaniem ruchu w ujęciu strategicznym wraz z komputerowym wspomaganiem w tym obszarze zadaniowym. Pierwszy obszar jest opisywany dość szeroko w aspekcie modelowania różnych elementów procesów eksploatacji. Wartościową w tym zakresie pozycją jest [13], w której przeprowadzona została klasyfikacja i przegląd najważniejszych modeli eksploatacyjnych. Pozostałe pozycje opisują mniej lub bardziej szczegółowo poszczególne grupy modeli eksploatacyjnych (modele zdarzeń np. [7, 27, 25], modele diagnostyczne np. [1, 2, 27, 28], modele obiektów technicznych [13, 19, 7], czy modele procesów eksploatacyjnych [7, 13, 27, 22, 26]. Drugi obszar jest opisywany w literaturze z punktu widzenia nauk ekonomicznych, szczególnie w ujęciu zarządzania strategicznego [3, 4, 5, 6, 23]. Trzeci obszar jest opisywany głównie w artykułach pism branżowych, szczególnie w [17]. Wyniki przeprowadzonego przeglądu istniejącego stanu wiedzy można podsumować w dwóch aspektach: 1. Z punktu widzenia krajowych i zagranicznych publikacji oraz innych opracowań, istnieją odniesienia do metod i zagadnień cząstkowych będących elementem proponowanego rozwiązania. 2. W ujęciu kompleksowym, wykorzystanie technik scenariuszowych w odniesieniu do modelowania procesów eksploatacyjnych można uznać za zagadnienie nowe nie publikowane zarówno w kraju, jak i za granicą. 5. Koncepcja badań nad opracowaniem sposobu wykorzystania metody scenariuszy w modelowaniu procesów eksploatacyjnych Rozpoczęte i prowadzone przez autora badania nad opracowaniem sposobu wykorzystania scenariuszy w modelowaniu procesów eksploatacyjnych obejmują swoim obszarem szeroki zakres prac zarówno o charakterze podstawowym (teoretycznym), jak i przemysłowym i rozwojowym. Badania te zmierzają z jednej strony do opracowania metodyki budowy scenariuszy eksploatacyjnych jako podstawy i sposobu modelowania i optymalizacji procesów eksploatacyjnych, z drugiej zaś do opracowania praktycznych narzędzi pozwalających na wykorzystanie opracowanej metodologii w warunkach przemysłowych. Schemat koncepcji prowadzenia takich badań przedstawiono na rys. 4. Jego poszczególne elementy zostaną opisane w kolejnych podpunktach. Etap 1: Identyfikacja i pozyskiwanie informacji i wiedzy dla potrzeb modelowania obiektów, zdarzeń i procesów eksploatacji Etap 2: Budowa i badanie modeli obiektów, zdarzeń i procesów eksploatacyjnych Etap 3: Badanie i ocena wpływu zdarzeń i procesów na stan eksploatacyjny typowego przedsiębiorstwa przemysłowego Etap 4: Budowa i badania modeli scenariuszy zdarzeń i procesów eksploatacji Etap 5: Budowa systemu obsługi zdarzeń i procesów eksploatacyjnych Rys.4. Schemat sposobu prowadzenia badań 5.1. Etap 1: Identyfikacja i pozyskiwanie informacji i wiedzy dla potrzeb modelowania obiektów, zdarzeń i procesów eksploatacji Pierwszym etapem badań jest pozyskanie i zgromadzenie odpowiedniego zasobu informacji i wiedzy o obiektach i wszelkich technicznych i nietechnicznych "okolicznościach" ich funkcjonowania, stanowiących potencjalny przedmiot i zakres scenariuszy eksploatacyjnych. Jakość scenariusza, a przez to dokładność prognozy wynika z możliwie dużej różnorodności i szczegółowości informacji, dlatego poszukiwanie i pozyskiwanie powinno mieć charakter uporządkowany. Podstawą takiego uporządkowania są na przesłanki wynikające z trzech płaszczyzn odniesionych do funkcjonowania wybranych obiektów eksploatacji: technicznej, organizacyjno-decyzyjnej i wymiany informacji. 1. Płaszczyzna techniczna obejmuje zbiór czynników o obiektach i realizowanych z ich udziałem procesach eksploatacyjnych. Działania zmierzające do uzyskania informacji omawianego typu dotyczą: • identyfikacji złożoności obiektów technicznych będących przedmiotem modelowania, • uporządkowania procedur obsługowo-naprawczych realizowanych w ramach zidentyfikowanych wcześniej obiektów technicznych, • inwentaryzacji zasobów eksploatacyjnych (siły roboczej, części zamiennych, narzędzi, usług zewnętrznych). 2. Płaszczyzna organizacyjno-decyzyjna obejmuje czynniki wynikające ze sposobu funkcjonowania typowego działu utrzymania ruchu zarówno w aspekcie wewnętrznym, jak również w odniesieniu do szeroko rozumianego otoczenia zewnętrznego. Podstawą może tu być jeden z modeli: ogólny model zarządzania eksploatacją [7] lub model BCM (Business Centered Maintenance) [9]. W tym przypadku identyfikowane są cechy charakterystyczne typowego wzorcowego działu utrzymania ruchu przedsiębiorstwa, co pozwala na zgromadzenie bogatego zasobu informacji i wiedzy dla potrzeb opracowania odpowiednich modeli, a następnie scenariuszy eksploatacyjnych. 3. Płaszczyzna wymiany informacji obejmuje kryteria związane zarówno ze sposobem przepływu informacji w ramach typowej organizacji utrzymania ruchu oraz do/z organizacyjnych jednostek zewnętrznych. Szczególnej uwagi wymaga identyfikacja struktury obiegu informacji, która powinna rozpocząć się od wyodrębnienia obiektu, jakim jest organizacja utrzymania ruchu (rys. 5). Jednostki otoczenia zewnętrznego Segment zasobów i usług zewnętrznych Jednostki struktury organizacji ORGANIZACJA UTRZYMANIA RUCHU Jednostka główna/Inne jednostki organizacyjne Rys.5. Schemat wzorcowej Organizacji Utrzymania Ruchu w relacji z otoczeniem Obiekt taki stanowi uogólniony, uproszczony model przepływu informacji pomiędzy rozpatrywaną jednostką organizacyjną a jednostkami zewnętrznymi. Przy czym za jednostki zewnętrzne uważa się zarówno te, które są zawarte w strukturach organizacyjnych systemu przedsiębiorstwa (systemu technicznego), jak również te, które do struktur tych nie należą. Inwentaryzacja przepływu informacji powinna dotyczyć identyfikacji: • wejścia i wyjścia informacyjnego, • struktury przepływu informacji wewnątrz organizacji utrzymania ruchu. Badania w ramach tego etapu muszą być odniesione do realnie istniejących przedsiębiorstw, w celu uwzględnienia cech wynikających ze specyfiki różnych branż. Zakłada się gromadzenie i porządkowanie informacji i wiedzy, jednocześnie prowadząc analizy począwszy od celu i funkcji działań eksploatacyjnych, przez identyfikację i modelowanie obiektów technicznych, plany strategiczne i prewencyjne, aspekty organizacyjne do elementów kontrolnych. Wynikiem etapu 1 powinien być zbiór informacji i wiedzy związany z funkcjonowaniem typowej wzorcowej organizacji utrzymania ruchu z uwzględnieniem wykorzystanych do tego celu narzędzi informatycznych (np. system klasy CMMs/EAM czy narzędzie modelowania procesów ARIS Toolset). 5.2. Etap 2: Budowa i badanie modeli obiektów, zdarzeń i procesów eksploatacyjnych Celem badań prowadzonych w ramach tego etapu jest zbudowanie modeli obiektów, zdarzeń i procesów eksploatacyjnych. Badania te muszą być prowadzone w oparciu o wyniki etapu 1, czyli zebrany zasób informacji i wiedzy o obiektach i realizowanych procesów eksploatacyjnych. Ze względu na dużą różnorodność modeli stosowanych w obszarze eksploatacji, będą one budowane i porządkowane, w oparciu o schemat obrazujący relacje pomiędzy wybranymi cechami eksploatacyjnymi (rys. 6). Rys.6. Hierarchiczna struktura informacji eksploatacyjnej [7, 12] Każdy z przedstawionych poziomów może obejmować zbiór modeli charakterystyczny dla omawianych aspektów eksploatowania systemów technicznych. Przy czym, ze względu na hierarchiczny układ zagadnień eksploatacyjnych powodujący pełną współzależność elementu nadrzędnego i elementów podrzędnych, budowa i porządkowanie modeli będzie odbywać się rozpoczynając od poziomu najniższego. Według powyższego układu, można przyjąć, że: • w zakresie modelowania obiektów technicznych (cech obiektów) konieczne jest wykorzystanie metod modelowania strukturalnego i funkcjonalnego, jak również metod modelowania grupowego i indywidualnego, • w odniesieniu do zdarzeń eksploatacyjnych konieczne jest modelowanie zdarzeń zamierzonych i niezamierzonych, w oparciu o istniejące w tym zakresie narzędzia (karty zdarzeń, drzewa zdarzeń, drzewa logiczne itp.) [2, 12] z uwzględnieniem specyfiki procesu decyzyjnego będącego wynikiem lub skutkiem zaistnienia danej sytuacji, • procesy eksploatacji będą przedmiotem modelowania z użyciem map procesów oraz narzędzi informatycznych (np. Aris Toolset), co pozwala na przeprowadzenie symulacji w odniesieniu do przyszłych zachowań obiektów technicznych. Wynikiem realizacji badań w ramach tego etapu będzie techniczno-organizacyjny kompleksowy zbiór modeli eksploatacyjnych, jako podstawa scenariuszy eksploatacyjnych oraz systemu obsługi zdarzeń i procesów eksploatacyjnych. 5.3. Etap 3: Badanie wpływu zdarzeń i procesów na stan eksploatacyjny typowego przedsiębiorstwa przemysłowego Badania w ramach tego etapu będą prowadzone przy założeniu, że stan eksploatacyjny przedsiębiorstwa przemysłowego (z technicznego punktu widzenia) jest określany wskaźnikami rangowymi, których wykorzystanie (podział na określone klasy) pozwoli na określenie skutków oddziaływania poszczególnych zdarzeń i procesów na przedsiębiorstwo. Przedmiotem badań w ramach tego etapu będzie w pierwszej kolejności identyfikacja klas stanowiących podstawę wyodrębnienia wartości wskaźnika rangowego w oparciu o zaproponowany zbiór miar pozwalających na ilościową ocenę eksploatowania obiektów technicznych oraz funkcjonowania służb utrzymania ruchu. Ze względu na dużą różnorodność tego typu wielkości, celem badań w tym zakresie będzie ocena ich przydatności oraz wybór tych, które w kontekście dalszych prac będą najlepiej odzwierciedlały efektywność, jakość i niezawodność funkcjonowania rozpatrywanych obiektów technicznych. Następnie, w oparciu o opracowane w ramach etapu 2 modele, przeprowadzona zostanie analiza i dokonana identyfikacja wpływu poszczególnych zdarzeń i procesów na stan przedsiębiorstwa charakterystycznych dla każdej zidentyfikowanej klasy. Wymaga to przeprowadzenia szeroko zakrojonych prac klasyfikujących techniczne i nietechniczne (organizacyjne, ekonomiczne) czynniki zdarzeń i procesów wpływające na stan przedsiębiorstwa. Tak nakreślone badania będą realizowane na trzy sposoby: • metodą wprowadzenia „próbnej” awarii obiektu i symulacji potencjalnych skutków, • metodą szczegółowej analizy przyczynowo-skutkowej zaistniałych w przeszłości zdarzeń i realizowanych w wyniku tego działań (procesów eksploatacji), porządkowanych z uwzględnieniem modeli wynikających z metodologii Root Cause Failure Analysis (RCFA), • metodą szczegółowej analizy przyczynowo-skutkowej ewentualnych bieżących zdarzeń i ich skutków (występujących w okresie prowadzenia badań), z których informacje będą uzupełniały materiał badawczy. Po zgrupowaniu wyników badań i opracowaniu wykazów przyczynowo-skutkowych określane będzie przyporządkowanie zdarzeń i procesów dla obiektów eksploatacji do odpowiedniego wskaźnika rangi. 5.4. Etap 4: Budowa i badania modeli scenariuszy zdarzeń i procesów eksploatacji W ramach tego etapu, wychodząc z wyników badań przeprowadzonych w ramach poprzednich etapów, zbudowane zostaną modele procesów eksploatacyjnych poprzez wprowadzenie do nich elementów metodologii scenariuszowej, przez co realne będzie uwzględnienie wieloaspektowego charakteru zarządzania procesami eksploatacyjnymi i możliwości podejmowania optymalnych decyzji. Ze względu na specyfikę eksploatacyjną, będą brane pod uwagę dwa typy scenariuszy [4,23]: scenariusze możliwych zdarzeń i scenariusze symulacyjne. Etap ten zostanie zrealizowany w dwóch krokach: 1. W oparciu o wybrany sposób tworzenia scenariuszy, zostanie opracowany zbiór scenariuszy eksploatacyjnych dla wybranego obszaru technicznego-organizacyjnego. Wybór obszaru będzie warunkowany wynikami badań realizowanych w poprzednich etapach, w szczególności opracowanymi modelami obiektów, zdarzeń i procesów eksploatacji (etap 2) oraz badaniami nad wpływem poszczególnych zdarzeń i procesów na stan przedsiębiorstwa (etap 3). 2. W oparciu o opracowany zbiór scenariuszy zdarzeń i procesów zostaną opracowane wytyczne do sposobu wykorzystania technik scenariuszowych jako podstawa modelowania zdarzeń i procesów eksploatacyjnych. Pozwoli to na podejmowanie strategicznych decyzji związanych z eksploatowaniem obiektów technicznych i funkcjonowaniem służb utrzymania ruchu. Wynikiem realizacji badań w ramach tego etapu będzie zbiór modeli procesów eksploatacyjnych zbudowanych z wykorzystaniem metody scenariuszy wraz z wytycznymi określającymi sposób ich wykorzystania w średnio- i długoterminowym planowaniu i realizacji prac obsługowo-naprawczych. 5.5. Etap 5: Budowa systemu obsługi zdarzeń i procesów eksploatacyjnych Celem badań w ramach tego etapu będzie opracowanie założeń do budowy systemu obsługi zdarzeń i procesów eksploatacji i budowa takiego systemu, w oparciu o wyniki wcześniejszych etapów. Jako punkt wyjścia do tych badań zakłada się, że budowany system posiada dwie warstwy: • warstwę akwizycji danych i wiedzy o zdarzeniach i procesach eksploatacji, w postaci komponentów zapewniających komunikację ze źródłami danych i wiedzy w systemie informatycznym przedsiębiorstwa, • warstwę wspomagania decyzji eksploatacji przez kontrolę przepływu zdarzeń i procesów, z uwzględnieniem opracowanych modeli scenariuszy eksploatacyjnych w postaci zbioru komponentów systemu wspomagania obsługi zdarzeń i procesów eksploatacji. Wynikiem badań w ramach tego etapu będzie prototypowy system obsługi zdarzeń i procesów eksploatacyjnych, którego działanie potwierdzi praktyczną realizowalność przyjętych w ramach badań założeń. 6. Podsumowanie Uwzględnienie wieloaspektowego i wielowariantowego charakteru zarządzania procesami eksploatacyjnymi, co przekłada się także na możliwość podejmowania optymalizacji, w praktyce przemysłowej wydaje się być bardziej realne w ujęciu „scenariuszowym” niż w ujęciu klasycznym. Postać modelu, jak również zbiór kryteriów muszą uwzględniać szeroki wachlarz działań technicznych i nietechnicznych, jak również sposób ich powiązania zarówno „wewnątrz” modelowanego procesu, jak i w jego bliższym i dalszym otoczeniu. Może to być realne pod warunkiem odpowiedniej budowy scenariuszy, używanych do opisu sytuacji eksploatacyjnych. Granica złożoności opisu scenariusza wydaje się tkwić nie tyle w samej metodzie, ile w możliwościach narzędzi – głównie informatycznych – wykorzystywanych dla potrzeb praktycznych zastosowań tej metody. Przedstawiony sposób modelowania procesów eksploatacyjnych stanowi kluczowy element badań autora prowadzonych w Instytucie Inżynierii Produkcji Politechniki Śląskiej. Uzasadnienie podjęcia tej tematyki można określić w dwóch obszarach: 1. w obszarze teorii eksploatacji, 2. w obszarze zastosowań metod i narzędzi wspomagających w praktyce przemysłowej. Pierwszy obszar uzasadniający podjęcie tej tematyki wynika z nieuzasadnionego, zdaniem autora, braku większego zainteresowania zastosowaniem technik scenariuszowych w dziedzinie nauk technicznych, pomimo że techniki te są znane od wielu lat i stosowane w obszarze nauk ekonomicznych. Zdaniem autora, badania tutaj omawiane mogą stanowić wartość dodaną w dziedzinie eksploatacji maszyn i urządzeń. Drugi obszar wynika z bogatych, praktycznych i wieloletnich doświadczeń autora w zakresie funkcjonowania służb technicznych (służb utrzymania ruchu) przedsiębiorstw przemysłowych wielu branż (spożywczej, motoryzacyjnej, papierniczej, chemicznej, wodociągowo-kanalizacyjnej i in.). Obecnie najważniejsze problemy eksploatacyjne przedsiębiorstw przemysłowych dotyczą planowania głównych inżynierów długoterminowego, z uwzględnieniem momentów wymian zużytych elementów i całych obiektów technicznych w warunkach konieczności ciągłego obniżania kosztów. Problemy te nie występują w odosobnieniu technicznym, ale należy uwzględniać jednocześnie organizacyjne, ekonomiczne i normatywno-prawne aspekty funkcjonowania przedsiębiorstwa. Dotychczas problemy te rozwiązywane były w dwóch odrębnych aspektach: • technicznym, czyli w oparciu o wskaźniki niezawodnościowe uzupełnione w niektórych przypadkach wynikami badań diagnostycznych, • organizacyjno-ekonomicznym, czyli w oparciu o wskaźniki wynikające z kosztów działalności technicznej przedsiębiorstwa. Zastosowanie technik scenariuszowych w modelowaniu procesów eksploatacyjnych może stanowić skuteczne narzędzie rozwiązywania problemów planowania i realizacji prac obsługowo-naprawczych wielu przedsiębiorstw przemysłowych, a tym samym umożliwi realizację szczegółowych działań w zakresie: • wdrażania nowoczesnych filozofii zarządzania utrzymaniem ruchu (TPM, RCM, WCM), • budowy inteligentnych systemów obsługi zdarzeń i procesów eksploatacyjnych, • opracowywania procedur eksploatacyjnych w odniesieniu do projektowanych zakładów przemysłowych, • prowadzenia różnych analiz (np. analizy awaryjności), a przede wszystkim • optymalizowania procesu podejmowania decyzji eksploatacyjnych. Literatura 1. Cempel C. Diagnostyka wibroakustyczna maszyn. Warszawa: WNT, 1989. 2. Cholewa W., Kaźmierczak J. Data Processing and Reasoning in Technical Diagnostics. Warszawa: WNT, 1995. 3. Ducot C., Lubben G.J., A Typology of Scenarios. Futures 1980: No. 1. 4. Gierszewska G., Romanowska M. Analiza strategiczna przedsiębiorstwa. Warszawa: PWE, 2003. 5. Heijden K. Planowanie scenariuszowe w zarządzaniu strategicznym. Kraków: Dom Wydawniczy ABC, 2000. 6. Kahn H, Wiener A.J. The Year 2000. A Framework for Speculation on the Next Thirty Three Years. Macmillan, New York 1967. 7. Kaźmierczak J. Eksploatacja systemów technicznych. Gliwice: Wydawnictwa Politechniki Śląskiej, 2000. 8. Kaźmierczak J., Loska A. Scenariusze eksploatacyjne. Służby Utrzymania Ruchu 2007; 3: 64-69. 9. Kelly A.: Strategic Maintenance Planning. Butterworth-Heinemann, Oxford 2006. 10. Legutko S.: Development trends in machines operation maintenance. Eksploatacja i Niezawodność - Maintenance and Reliability 2009; 2(49: 8-16. 11. Levitt J.: The Handbook of Maintenance Management. New York: Industrial Press Inc., 1997. 12. Loska A. Bazy danych w zarządzaniu eksploatacją maszyn i urządzeń. Rozprawa doktorska. Gliwice: Politechnika Śląska, 2002. 13. Loska A. Koncepcja wykorzystania scenariuszy w modelowaniu procesów eksploatacyjnych. Ekonomika i Organizacja Przedsiębiorstwa ORGMASZ 2005; 4: 106. 14. Loska A. Modelowy sposób funkcjonowania służb utrzymania ruchu - strategie eksploatacyjne, Służby Utrzymania Ruchu; 2010: 3(21), 15. Loska A. Pozyskiwanie informacji dla potrzeb budowy scenariuszy eksploatacyjnych. Świnoujście-Kopenhaga 2006: IV Międzynarodowa Konferencja Naukowo-Techniczna EXPLO-SHIP, 197-206. 16. Loska A. Przegląd metod modelowania jako podstawa budowy scenariuszy eksploatacyjnych. Opole: Oficyna Wydawnicza Polskiego Towarzystwa Zarządzania Produkcją, 2010, t. 2, 152-161. 17. Loska A. Sposób wspomagania zarządzania utrzymaniem ruchu z wykorzystaniem wybranych narzędzi informatycznych, Służby Utrzymania Ruchu 2008-2009. 18. Loska A. Zastosowanie metodologii Business Centered Maintenance w pozyskiwaniu i porządkowaniu informacji dla potrzeb konstruowania scenariuszy eksploatacyjnych. Ekonomika i Organizacja Przedsiębiorstwa ORGMASZ 2008; 6: 77-78. 19. Loska A., Senczyna S., Kaźmierczak J. Modelowanie struktury technicznej przedsiębiorstwa dla potrzeb komputerowego wspomagania zarządzania eksploatacją. Ekonomika i Organizacja Przedsiębiorstwa ORGMASZ 2007; 7: 81-82. 20. Moubray J. RCM II - Reliability Centered Maintenance. Oxford: Elsevier Ltd., 2007. 21. Niebel W.B. Engineering Maintenance Management. New York: Marcel Dekker Inc., 1994. 22. Percy D. F., Kobbacy K. A. H. Preventive maintenance modelling. A Bayesian perspective, Journal of Quality in Maintenance Engineering 1996; vol. 2, nr 1: 15-24. 23. Praca zbiorowa po red. Cieślak M. Prognozowanie gospodarcze. Metody i Zastosowanie. Warszawa: PWN, 2001. 24. Praca zbiorowa pod red. Koradecka D. Bezpieczeństwo pracy i ergonomia. Warszawa: Centralny Instytut Ochrony Pracy, 1999. 25. Skotnicka-Zasadzień B., Biały W.: An analysis of possibilities to use Pareto chart for evaluating minimg machines’failure frequency. Eksploatacja i Niezawodność Maintenance and Reliability 2011; 3(51): 51-55. 26. Staniszewski R. Sterowanie procesem eksploatacji. Warszawa: WNT, 1990. 27. Żółtowski B, Niziński S. Modelowanie procesów eksploatacji maszyn. Bydgoszcz, Sulejówek: MARCAR, 2002. 28. Żółtowski B.: Podstawy diagnostyki maszyn. Bydgoszcz: Akademia TechnicznoRolnicza, 1996. Hubert Dębski Grzegorz Koszałka Mirosław Ferdynus WYKORZYSTANIE MES W ANALIZIE STRUKTURY NOŚNEJ RAMY NACZEPY O ZMIENNYCH PARAMETRACH EKSPLOATACYJNYCH Słowa kluczowe: metoda elementów skończonych, rama nośna, obliczenia wytrzymałościowe, naczepa, transport nienormatywny Streszczenie W pracy zaprezentowano numeryczną analizę konstrukcji ramy naczepy o zmiennej długości i podwyŜszonej nośności przeznaczonej do transportu ładunków ponadnormatywnych. Prowadzone badania miały na celu opracowanie adekwatnych modeli numerycznych MES umoŜliwiających identyfikację wytęŜenia konstrukcji oraz stanu odkształcenia w warunkach obciąŜeń eksploatacyjnych. Zastosowanym do analizy narzędziem numerycznym był program Abaqus/Standard, umoŜliwiający prowadzenie obliczeń w zakresie geometrycznie nieliniowym z wykorzystaniem przyrostowo-iteracyjnej metody Newtona-Raphsona. W wyniku przeprowadzonych prac ustalono newralgiczne obszary ramy w których występowały niebezpieczne napręŜenia. UmoŜliwiło to modyfikację konstrukcji pozwalającą na zmniejszenie napręŜeń do bezpiecznego poziomu. 1. Wprowadzenie Projektowanie nowoczesnych konstrukcji jest zadaniem złoŜonym, wymagającym uwzględnienia wielu cech materiałowych i mechanicznych przy optymalizacji geometrii projektowanej struktury. Dotyczy to zwłaszcza elementów konstrukcyjnych o znaczeniu krytycznym, których nośność determinuje wytrzymałość całego ustroju nośnego. Proces projektowania takich elementów wymaga wykorzystania nowoczesnych narzędzi, umoŜliwiających poszukiwanie najlepszych rozwiązań konstrukcyjnych [10-12]. Współczesnym narzędziem dającym szerokie moŜliwości analizy parametrów wytrzymałościowych projektowanych konstrukcji jest oprogramowanie numeryczne z grupy CAE wykorzystujące metodę elementów skończonych [3, 4, 6, 7, 8]. Oprogramowanie takie znajduje obecnie szerokie zastosowanie w wielu dziedzinach przemysłu, a w szczególności w przemyśle kosmonautycznym, lotniczym i motoryzacyjnym. Grupę konstrukcji nośnych wobec których stawiane są wysokie wymagania wytrzymałościowe i sztywnościowe stanowią ramy współczesnych środków transportu. W szczególności do rozwiązań konstrukcyjnych poddanych działaniu wysokich obciąŜeń eksploatacyjnych moŜna zaliczyć ramy naczep przeznaczonych do transportu ładunków o ponadnormatywnych masach i wymiarach. Projektowanie tego typu konstrukcji wymaga uwzględnienia róŜnych konfiguracji eksploatacyjnych naczepy, umoŜliwiających w razie potrzeby zwiększenie długości i szerokości platformy przewozowej [5, 9, 10]. Zmiana konfiguracji konstrukcji podczas jej eksploatacji wpływa w znaczący sposób na zmianę charakteru jej obciąŜenia, co wymaga dodatkowego uwzględnienia w procesie projektowokonstrukcyjnym struktury. W pracy zaprezentowano koncepcję ramy centralnej z moŜliwością szerokiego zakresu zmiany jej długości w zaleŜności od potrzeb eksploatacyjnych. Przyjęto rozwiązanie bazujące na dwóch współpracujących belkach o strukturze cienkościennej konstrukcji skrzyniowej, z regulacją długości ramy realizowaną poprzez wzajemne wysuwanie i wsuwanie elementów nośnych względem siebie. W obliczeniach uwzględniono szczegóły konstrukcyjne, takie jak otwory technologiczne, mogące mieć istotny wpływ na ocenę stanu wytęŜenia konstrukcji ramy. Obliczenia przeprowadzono przy wykorzystaniu programu Abaqus/Standard, stanowiącego nowoczesne narzędzie numeryczne wykorzystujące metodę elementów skończonych w zakresie geometrycznie i fizycznie nieliniowym [1, 2 ,6, 9]. 2. Cel i zakres badań Celem prowadzonych badań była ocena wytrzymałości i sztywności opracowywanego rozwiązania konstrukcyjnego ramy nośnej naczepy, przeznaczonej do transportu ładunków długich o cięŜarze do 45 ton. W przyjętej koncepcji ramy, składającej się z dwóch niezaleŜnych części: przedniej i tylnej, moŜliwa była łatwa zmiana jej długości dzięki moŜliwości wysuwania części przedniej z części tylnej ramy. UmoŜliwiało to dopasowanie długości naczepy do przewoŜonego ładunku. Opracowane nowe rozwiązanie konstrukcyjne wymagało dokładnej analizy stanu odkształcenia i wytęŜenia, którą przeprowadzono w oparciu o metodę elementów skończonych. Przed przystąpieniem do obliczeń MES, analitycznie wyznaczono momenty zginające działające na ramę naczepy dla róŜnych kombinacji jej długości oraz rozmieszczenia przewoŜonego ładunku. W rozwaŜaniach tych uwzględniano tylko takie przypadki, które uznano za prawdopodobne i dopuszczalne (np. obciąŜenie maksymalnie wydłuŜonej ramy punktową siłą, o wartości odpowiadającej maksymalnej dopuszczalnej masie ładunku, przyłoŜoną w środku jej długości uznano za przypadek mało prawdopodobny i wymieniony w instrukcji do naczepy jako niedopuszczalny). Na podstawie powyŜszych rozwaŜań wybrano przypadki do analiz numerycznych MES (rys. 1). Rys. 1. Schematy obciąŜenia ramy zsuniętej (a) i rozsuniętej (b) ładunkiem Ze względu na analizę konstrukcji w zakresie obciąŜeń eksploatacyjnych, nie dopuszczając tym samym moŜliwości wystąpienia odkształceń trwałych w elementach konstrukcji, zdecydowano się na przyjęcie modelu materiału liniowo-spręŜystego. Uwzględniając zastosowane w modelu numerycznym interakcje kontaktowe oraz moŜliwość wystąpienia duŜych odkształceń zdecydowano się na przeprowadzenie obliczeń w zakresie geometrycznie nieliniowym z wykorzystaniem przyrostowo-iteracyjnej metody NewtonaRaphsona [2, 6, 9]. 3. Budowa modelu numerycznego MES Zdecydowano się na rozwaŜenie dwóch konfiguracji pracy ramy: ramy zsuniętej oraz ramy maksymalnie rozsuniętej. W obydwu wariantach zastosowano te same parametry konstrukcyjne oraz materiałowe. Konstrukcję ramy zaprojektowano w taki sposób, aby część tylna będąca podstawowym elementem konstrukcji, dla której warunki brzegowe stanowi układ zawieszenia pojazdu, zapewniała jednocześnie warunki podparcia dla części przedniej, będącej ruchomym elementem ramy. MoŜliwość zmiany długości ramy następuje w wyniku wsuwania i wysuwania części przedniej z konstrukcji części tylnej ramy. Widok ogólny modelu numerycznego ramy przedstawia rys. 2. a) b) Rys. 2. Modele dyskretne ramy z zaznaczonym sposobem modelowania obciąŜenia zewnętrznego: a) konfiguracja zsunięta, b) konfiguracja rozsunięta ObciąŜenie modelu numerycznego stanowił cięŜar własny konstrukcji jak równieŜ obciąŜenie masami skupionymi pochodzącymi od cięŜaru ładunku wynoszącego 45 ton, uwzględniając dodatkowo 20% nadwyŜek dynamicznych. Schemat rozłoŜenia ładunku zewnętrznego w postaci mas skupionych dla obydwu rozwaŜanych konfiguracji ramy przedstawiają rys. 1 i 2. Budowę przestrzennego modelu dyskretnego konstrukcji oparto na elementach typu powłokowego, posiadających po 6 stopni swobody w węźle elementu [1]. W obszarze połączenia przedniej i tylnej części ramy przyjęto rozwiązanie polegające na zastosowaniu podatnych poduszek, powodujących zwiększenie powierzchni współpracy, a tym samym wyeliminowanie niekorzystnych lokalnych nacisków krawędziowych w strefie połączenia. Wzajemne interakcje pomiędzy przednią i tylną częścią ramy odwzorowano poprzez zdefiniowanie oddziaływań kontaktowych pomiędzy współpracującymi elementami. 4. Warunki brzegowe modelu dyskretnego Zasadniczym zagadnieniem w opracowaniu modelu dyskretnego konstrukcji ramy było odwzorowanie warunków pracy zawieszenia pneumatycznego, zapewniającego w trakcie eksploatacji jednakowe reakcje dla kaŜdej osi kół naczepy. Zdecydowano się na odwzorowanie wahacza za pomocą elementów belkowych o określonej sztywności, dla których zdefiniowano zróŜnicowane warunki podparcia w węźle przednim i tylnym modelu wahacza. Połączenie wahacza z ramą w węźle przednim zapewnia sztywne przenoszenie obciąŜenia z zawieszenia na konstrukcję ramy, dopuszczając jedynie jego obrót względem osi kół. Węzeł tylny wahacza połączony jest z ramą za pośrednictwem elementu spręŜystego odwzorowującego poduszkę zawieszenia pneumatycznego. Sztywność elementu modelującego poduszkę została dobrana w sposób zapewniający zbliŜone reakcje w osiach wszystkich kół naczepy. Podparcie pojedynczego koła zrealizowane zostało poprzez odebranie środkowemu węzłowi wahacza (w węźle połączenia elementów belkowych) moŜliwości przemieszczania w kierunku osi X – rys. 3a. a) b) Rys. 3. Warunki brzegowe modelu MES: a) model zawieszenia koła, b) podparcie sworznia ramy oraz warunki symetrii osiowej modelu Dla przedniej część ramy zdefiniowano podparcie sworznia poprzez odebranie moŜliwości przemieszczeń węzłów sworznia w kierunkach osi Y i Z – rys. 3b. Ponadto w kierunku osi X (ruchy sworznia w pionie) wprowadzono podparcie węzłów w postaci reakcji spręŜystej o odpowiednio określonej sztywności. Dodatkowo dla całego modelu zdefiniowano warunki symetrii osiowej (wzdłuŜ osi ramy – oś Y), poprzez odebranie węzłom znajdującym się w osi symetrii modelu moŜliwości przemieszczania w kierunku osi Z oraz obrotu względem osi X i Y. Wyeliminowanie tych stopni swobody zapewniło stabilną pracę modelu, bez powstawania niepoŜądanych przemieszczeń konstrukcji w kierunku poprzecznym do jej osi (na boki) – rys. 3b. 5. Wyniki analizy numerycznej Przeprowadzone obliczenia numeryczne umoŜliwiły ocenę i porównanie stanu odkształcenia i wytęŜenia konstrukcji ramy dla obydwu rozwaŜanych wariantów eksploatacyjnych – ramy zsuniętej oraz maksymalnie rozsuniętej. Otrzymane mapy rozkładów napręŜenia zredukowanego H-M-H w elementach konstrukcji wskazywały obszary, w których napręŜenia zredukowane znacznie przekraczały przyjętą granicę plastyczności materiału Re = 360 MPa. Najbardziej niekorzystne gradienty napręŜenia występowały w części wysuwanej ramy oraz w strefie połączenia obydwu elementów nośnych, zarówno w przypadku konfiguracji zsuniętej, jak i rozsuniętej ramy – rys. 4. a) b) Rys. 4. Rozkład napręŜenia H-M-H w modelu ramy: a) rama rozsunięta – widok ogólny, b) rama zsunięta – zbliŜenie strefy połączenia Występowanie w pewnych obszarach bardzo wysokich poziomów napręŜenia zredukowanego wymagało podjęcia działań zmierzających do ich eliminacji, a tym samym poszukiwania lepszego rozwiązania konstrukcyjnego. W związku z tym wprowadzono modyfikacje szczegółów konstrukcyjnych bieŜącego rozwiązania. Dotyczyło to przede wszystkim zmiany usytuowania otworów technologicznych wykonanych w bocznych ściankach ramy, stanowiących źródło niebezpiecznych gradientów napręŜenia. Otwory te usunięto i zastąpiono je jednym otworem w dolnej ściance belki. Ponadto dodano płaskowniki wzmacniające boczne ścianki ramy w obszarze zagroŜonym oraz dodano Ŝebra wzmacniające (rys. 5). W wyniku takich działań opracowywano kolejne warianty konstrukcji, które poddawano analizie MES. Wprowadzone modyfikacje pozwoliły na eliminację stref newralgicznych ustroju nośnego. Na rys. 6 przedstawiono wyniki obliczeń MES dla ostatecznego wariantu konstrukcji. Rys. 5. Nakładki i Ŝebra wzmacniające przednią część ramy a) b) Rys. 6. Rozkład napręŜenia H-M-H w zmodyfikowanym modelu ramy: a) widok ogólny, b) zbliŜenie części przedniej ramy 6. Podsumowanie Przeprowadzona analiza numeryczna konstrukcji nośnej ramy naczepy umoŜliwia ocenę poprawności opracowanego rozwiązania konstrukcyjnego, pozwalając zidentyfikować obszary newralgiczne, determinujące wytrzymałość całej konstrukcji. Jest to niezwykle istotne w przypadku poszukiwania nowych rozwiązań konstrukcyjnych, gdzie na etapie projektowania skomplikowanych elementów nośnych występuje zbyt wiele nieznanych parametrów konstrukcyjnych. Znajomość rozkładów napręŜenia w elementach krytycznych stanowi zatem zagadnienie o pierwszorzędnym znaczeniu, a zastosowanie metody elementów skończonych pozwala analizować wytęŜenie konstrukcji jeszcze na etapie jej projektowania [6, 9, 10]. Wykorzystanie obliczeń numerycznych MES w procesie projektowania umoŜliwia szybkie i skuteczne wprowadzanie niezbędnych modyfikacji szczegółów konstrukcyjnych, prowadzące do powstawania kolejnych, coraz korzystniejszych wariantów konstrukcyjnych umoŜliwiających w efekcie opracowanie rozwiązania optymalnego. Literatura 1. Abaqus HTML Documentation, 2010. 2. Bathe K J. Finite element procedures. Prentice Hall, 1996. 3. Dacko M, Borkowski W, Dobrociński S, Niezgoda T, Wieczorek M. Metoda elementów skończonych w mechanice konstrukcji. Arkady, Warszawa 1994. 4. Grądzki R. Wprowadzenie do metody elementów skończonych. Politechnika Łódzka, Łódź 2002. 5. Koszałka G., Niewczas A., Dębski H., Golec M., Kaczor M., Taratuta L.: FEM analysis in design of extendable central beam for a semi-trailer. Machine Design, Vol. 3(2011), No.1, pp. 47-50. 6. Osiński J. Obliczenia wytrzymałościowe elementów maszyn z zastosowaniem metody elementów skończonych. Oficyna Wydawnicza PW, Warszawa 1997. 7. Rakowski G, Kacprzyk Z. Metoda Elementów Skończonych w mechanice konstrukcji. Oficyna Wydawnicza PW, Warszawa 2005. 8. Rudawska A., Dębski H. Experimental and numerical analysis of adhesively bonded aluminum alloy sheets joints. Eksploatacja i niezawodnosc – Maintenance and Reliability, 2011, Vol. 49, No. 1, pp. 4-10. 9. Rusiński E, Czmochowski J, Smolnicki T. Zaawansowana metoda elementów skończonych w konstrukcjach nośnych. Oficyna Wydawnicza Politechniki Wrocławskiej, Wrocław 2000. 10. Rusiński E, Kowalczyk M. Zasady przedłuŜania ram samochodów cięŜarowych. Samochody Specjalne, 6/2003, t.7. 11. Rusiński E, Porada D. Komputerowe wspomaganie projektowania ram samochodów cięŜarowych. Samochody Specjalne, 6/2000, t. 4. 12. Tarnowski W. Podstawy projektowania technicznego. WNT, Warszawa 1997. Dr inŜ. Hubert Dębski Wydział Mechaniczny Politechnika Lubelska ul. Nadbystrzycka 36, 20-618 Lublin E-mail: [email protected] Dr inŜ. Grzegorz Koszałka Wydział Mechaniczny Politechnika Lubelska ul. Nadbystrzycka 36, 20-618 Lublin E-mail: [email protected] Dr inŜ. Mirosław Ferdynus Wydział Mechaniczny Politechnika Lubelska ul. Nadbystrzycka 36, 20-618 Lublin E-mail: [email protected] Mgr inż. Łukasz Jedliński Katedra Podstaw Konstrukcji Maszyn Politechnika Lubelska Ul. Nadbystrzycka 36, 20-618 Lublin, Polska E-mail: [email protected] Odszumianie danych rejestrowanych wielokanałowo z użyciem transformaty falkowej Słowa kluczowe: odszumianie danych, transformata falkowa, sztuczne sieci neuronowe, przekładnia stożkowa Streszczenie: W celu uzyskania informacji o interesującym nas zjawisku lub obiekcie najczęściej rejestrowane są wybrane sygnały pomiarowe otrzymane za pośrednictwem czujników. Niestety uzyskane sygnały oprócz pożądanej informacji zawierają również zakłócenia, które są spowodowane m.in. właściwościami toru pomiarowego i procesami towarzyszącymi działaniu obiektu. W wielu przypadkach zachodzi potrzeba pomiaru takiej samej wielkości w różnych miejscach obiektu i/lub kierunkach. Potrzebne są zatem narzędzia do poprawy stosunku sygnału do szumu sygnałów rejestrowanych wielokanałowo. Transformata falkowa jest stosunkowo nową metodą przetwarzania danych, która znalazła zastosowanie w różnych dziedzinach takich jak technika i fizyka. W odniesieniu do sygnałów może być używana do odszumiania, kompresji, wykrywaniu trendu czy nieciągłości sygnału. W pracy tej transformata falkowa została użyta od odszumiania sygnałów drgań zarejestrowanych z dwóch trójosiowych czujników. Obiektem badań była przekładnia zębata stożkowa. Odszumianie sygnałów miało na celu poprawę skuteczności diagnozy uszkodzenia kół zębatych przekładni. 1. Wstęp W diagnostyce wibroakustycznej, jak i w innych dziedzinach nauki, dąży się do poprawy uzyskiwanych rezultatów [2]. W ostatnich latach obserwuje się ciągły rozwój algorytmów wnioskowana diagnostycznego i metod przetwarzania sygnałów [3,11,16]. Sprzęt pomiarowy i komputery pozwalają na pomiar sygnałów ze znacznie większą dokładnością oraz z wielu kanałów pomiarowych jednocześnie. W przypadku badania złożonych obiektów zachodzi potrzeba rejestracji wielu sygnałów. Również na początku badań typuje się więcej punktów pomiarowych, aby nie stracić istotnej informacji i wybrać te optymalne. Uzyskane sygnały pomiarowe zawsze zawierają zakłócenia. W prostym modelu sygnału [1,4,10] zakłada się, że sygnał składa się ze składowej zdeterminowanej zawierającej użyteczną informację oraz składowej losowej czyli szumu. Istnieje wiele metod ekstrakcji użytecznej informacji z sygnału np. poprzez filtrację sygnału, analizę składowych głównych PCA, uśrednianie sygnałów. Wśród tych metod możemy wyróżnić transformatę falkową (WT), która znalazła zastosowanie m.in. w odszumianiu sygnałów [14]. Uogólnieniem odszumiania za pomocą WT dla jednego sygnału jest procedura zaproponowana w artykule [1] dla wielu sygnałów, która została wykorzystana w tej pracy. 2. Stanowisko badawcze Badaną przekładnią jest przekładnia zębata stożkowa jednostopniowa. Na korpusie przekładni zamontowano dwa trójosiowe czujniki przyśpieszeń drgań oznaczone na rysunku 1 liczbą 1 i 2. Oś X czujników jest zgodna z kierunkiem osi wału wejściowego a oś Z z osią wału wyjściowego. Drgania rejestrowano dla przekładni w stanie dobrym oraz z uszkodzoną powierzchnią zębów w wyniku zacierania. Dokładniejszy opis stanowiska badawczego można znaleźć w pracy [9]. Rys. 1. Widok badanej przekładni z zaznaczonym położeniem czujników drgań 3. Dyskretna transformata falkowa (DWT) W przypadku ciągłej transformaty falkowej (CWT) obliczane są współczynniki falkowe dla każdej skali, co generuje dużą ilość danych oraz wymaga długotrwałych obliczeń. Dyskretna transformata falkowa wywodzi się z CWT i umożliwia wybór skal. DWT można zdefiniować jako [17]: −1/2 t − kτ 0 S0j DWT ( j , k ) = S0j x k ψ ∑k j S0 (1) gdzie: S0 > 1, τ0 > 1, ψ – falka podstawowa, x|k| – badany sygnał, j, k – liczby całkowite dodatnie. Do szybkich i wydajnych obliczeń DWT stosowany jest algorytm wprowadzony przez Mallata [12,15] znany pod nazwą multi-resolution analysis (MRA). W wyniku jego działania uzyskujemy wielorozdzielczą reprezentację sygnału w postaci aproksymacji i detali. Kompletne podstawy teoretyczne MRA możemy znaleźć w pracy Mallata [12] natomiast zależność, która go opisuje wyrażona jest wzorem [17]: = x(t ) j −1 ∑ S0 (τ )φ (t − τ ) + ∑∑ d j (τ ) j /2ψ (2 j t − τ ) τ τ j =0 (2) gdzie: τ – współczynnik przesunięcia, S0 – współczynnik skali, dj – współczynnik falkowy dla skali j, ф(t), ψ(t) – funkcja skalująca i falkowa. Detale i aproksymacje obliczane są w wyniku filtracji dwukanałowym zespołem filtrów (zwierciadlanymi filtrami kwadraturowymi [5,6]). Aby otrzymać rozkład sygnału na kilka poziomów operację filtracji wykonujemy w sposób iteracyjny (rys. 2). Z filtra dolnoprzepustowego otrzymujemy aproksymację Aj (składowa sygnału o niskiej częstotliwości) a z filtru górnoprzepustowego detal Dj (składowa sygnału o wysokiej częstotliwości)[7]. Na rysunku 2, przedstawiono przykładowy sygnał, który był rejestrowany z częstotliwością 10 kHz. W wyniku rozkładu, na pierwszym poziomie aproksymacja A1 sygnału, zawiera sygnał o częstotliwości z zakresu 0 – 5 kHz a detal D1 z zakresu 5 – 10 kHz. Jeżeli sygnał oryginalny składał się ze 100 próbek to w wyniku filtracji otrzymaliśmy detal o długości ok. 100 próbek i aproksymację o długości ok. 100 próbek. Suma sygnałów wynikowych jest w przybliżeniu dwa razy większa niż sygnału oryginalnego. Aby uniknąć takiego wzrostu liczby próbek stosuje się decymację usuwając co drugą próbkę z otrzymanej reprezentacji sygnału. Następnie możemy dokonać dalszej dekompozycji. Najczęściej dalszej analizie poddawana jest aproksymacja, w wyniku której otrzymujemy kolejny detal i aproksymację. Dekompozycja sygnału odbywa się na skończoną wartość poziomów ze względu na ograniczoną długość sygnału czy sens fizyczny uzyskanych detali i aproksymacji [18]. Zazwyczaj poziom dekompozycji nie jest większy niż 8. Rys. 2. Dekompozycja sygnału przy użyciu algorytmu multi-resolution analysis [17] Usuwanie zbędnej informacji w postaci szumu z sygnału po dekompozycji falkowej może być realizowane kilkoma metodami. Kolejne aproksymacje sygnału zawierają składową sygnału o coraz niższych częstotliwościach, a więc wybór odpowiedniej aproksymacji jako reprezentacji sygnału będzie metodą usuwania szumu. Inną możliwością jest prosta modyfikacja pierwszego detalu lub pierwszych detali poprzez zmianę wartości współczynników falkowych na zero. Przedstawione metody odszumiania sygnału są metodami niezbyt dokładnymi. Bardziej wyrafinowane algorytmy odszumiania sygnału bazują na zerowaniu współczynników falkowych detali na podstawie kryterium obliczanego dla każdego detalu. Oprócz wyboru odpowiedniego kryterium obliczania progu, poniżej którego, współczynniki falkowe są zerowane, należy także wybrać sposób egzekwowania progu. Metoda hard thresholding realizuje proces odszumiania ustawiając wartość zero dla elementów o wartości absolutnej poniżej wartości progu, pozostałe elementy nie są zmieniane. Odmianą metody hard thresholding jest metoda soft thresholding, która różni się od poprzedniej metody tym, że zmienia elementy, które nie są zerowane. Dzięki temu nie powstaje nieciągłość w miejscu, w którym elementy mają wartość równą wartości progu [13]. Rekonstrukcja sygnału odbywa się w podobny sposób jak dekompozycja. Należy nadpróbkować detale i aproksymacje przed syntezą w filtrach, których wybór jest krytyczny do pełnego odtworzenia oryginalnego sygnału. 3.1 Procedura odszumiania Procedura odszumiania danych wielowymiarowych jest uogólnieniem procedury odszumiania danych jednowymiarowych. Rozważania są zgodne z pracą Aminghafari i in. [1]. Załóżmy następujący p-wymiarowy model sygnału: X(t) = f(t) + ε(t), t = 1,..., n (3) gdzie: X(t), f(t), ε(t) są wymiaru 1xp, f(t) – sygnał, który chcemy odszumić, ε(t) – szum Gaussowski z nieznaną macierzą kowariancji E(ε(t)T ε(t))=Σε. Każda składowa X(t) jest postaci dla 1 ≤ i ≤ p: Xi(t) = fi(t) + εi(t), t = 1,..., n gdzie: fi – należy do pewnej przestrzeni funkcyjnej (najczęściej przestrzeni L2 lub Besova). (4) Macierz kowariancji Σε, która powinna być dodatnio określona, uchwyca stochastyczną zależność pomiędzy składowymi X(t) i modelami przestrzennej korelacji. Procedura odszumiania może być wyrażona za pomocą trzech kroków[1] dla macierzy X, która jest wymiaru n x p i składa się z p sygnałów (kolumn macierzy X) tak, że n>>p: • • Dla każdej kolumny macierzy X wykonaj dekompozycję falkową stopnia J. W tym kroku uzyskuje się J+1 macierzy D1,...,DJ zawierających współczynniki detali stopnia od 1 do J z p sygnałów i współczynniki aproksymacji AJ z p sygnałów. Macierze Dj i AJ są wymiaru n2-j x p oraz n2-J x p; Wyznacz estymator 𝛴� ε macierzy kowariancji szumu a następnie dokonaj dekompozycji SVD (singular value decomposition) macierzy 𝛴� ε za pomocą macierzy ortogonalnej V takiej, że 𝛴� ε = VAVT. Następnie zmień bazę przestrzeni za pomocą macierzy przejścia V (dokładnie obliczając Dj V, 1≤ i ≤ p) i zastosuj t = 2 λ log( n ) i i filtrowanie jednowymiarowe za pomocą progu dla i-tej kolumny • macierzy Dj V; Dokonaj rekonstrukcji odszumionej macierzy 𝑋� za pomocą uproszczonych macierzy detali i aproksymacji poprzez zmianę bazy przestrzeni za pomocą macierzy VT i odwrotnego przekształcenia falkowego. 4. Wyniki badań Sygnały drgań zarejestrowane w trakcie badań zostały poddane odszumianiu zgodnie z przedstawioną powyżej procedurą. Odszumianie przeprowadzono dla sygnałów z przekładni w stanie dobrym i uszkodzonym dla sześciu rejestrowanych kanałów. Na rys. 3 przedstawiono wykres współczynników falkowych obrazujących zmianę częstotliwości sygnału drgań w czasie. Rys. 3. Wykres 3D sygnału drgań Wyniki odszumiania dla sygnałów w dziedzinie czasu dla przekładni uszkodzonej przedstawiono na rysunku 4. Na podstawie wykresów dla osi czujników zgodnych z kierunkiem osi wału wyjściowego można zauważyć, że sygnały te zawierają najwięcej szumu. 400 Original signal De-noised signal 1 300 600 400 500 200 2 Original signal De-noised signal 2 0 -500 1 100 0 4 200 5 x 10 4 0 -500 1 2 3 4 0 5 x 10 4 500 4 -200 0 -500 1 -300 2 3 4 5 x 10 500 4 1000 -400 500 5 -400 3 500 3 -100 -200 2 0 6 0 -500 -500 1 2 3 4 5 x 10 0.5 1 2 1 3 5 4 x 10 4 1.5 2 2.5 -600 4 3 3.5 4 4.5 5 0.5 4 1 1.5 2 2.5 3 3.5 4 4.5 600 400 x 10 Original signal De-noised signal 3 5 4 x 10 800 Original signal De-noised signal 4 600 400 200 200 0 0 -200 -200 -400 -400 -600 -600 -800 -800 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.5 1 1.5 2 2.5 3 3.5 4 4.5 4 400 x 10 Original signal 1000 De-noised signal 800 5 5 4 x 10 Original signal De-noised signal 6 600 200 400 0 200 0 -200 -200 -400 -400 -600 -600 -800 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0.5 5 1 1.5 2 2.5 3 3.5 4 4.5 5 4 4 x 10 x 10 Rys. 4. Sygnały drgań dla przekładni uszkodzonej dla sześciu kanałów pomiarowych przed i po odszumianiu Poziom dekompozycji sygnałów wynosił trzy a na falkę bazową wybrano Coiflet 1 (rys. 5) 2 1 0 -1 0 1 2 3 4 5 Rys. 5. Falka Coiflet 1 Dla tego samego kryterium obliczania progów dla detali sygnału odszumianie zrealizowano dla metody soft i hard thresholding. Następnie obliczono cechy sygnałów i wybrano osiem cech z dwudziestu za pomocą algorytmu przedstawionego w pracy [8]. Wybrane cechy to: • wartość średnia, • wartość skuteczna, • wartość szczytowa, • współczynnik szczytu, • współczynnik luzu, • odchylenie standardowe, • energy ratio, • FM0. Aby porównać skuteczność odszumiania sygnałów obliczono także takie same cechy dla sygnałów bez przetwarzania. Klasyfikacja stanu została wykonana za pomocą sztucznej sieci neuronowej multi layer perceptron (MLP). Tabela 1. Wyniki klasyfikacji stanu dla sygnału bez przetwarzania Jakość Jakość Jakość Wszystkie Nazwa sieci (uczenie) (testowanie) (walidacja) próby MLP 8-4-2 89,88 91,66 91,66 90,41 Tabela 2. Wyniki klasyfikacji stanu dla sygnału odszumionego (soft thresholding) Jakość Jakość Jakość Wszystkie Nazwa sieci (uczenie) (testowanie) (walidacja) próby MLP 8-4-2 95,53 95,83 95,83 95,62 Tabela 3. Wyniki klasyfikacji stanu dla sygnału odszumionego (hard thresholding) Jakość Jakość Jakość Wszystkie Nazwa sieci (uczenie) (testowanie) (walidacja) próby MLP 8-9-2 97,02 94,44 94,44 96,24 Rezultaty klasyfikacji sieci podano w tabelach. Skuteczność klasyfikacji dla sygnału bez przetwarzania wyniosła około 90% a dla sygnału po odszumianiu odnotowano około 5% poprawę wyników. Różnice między metodą hard i soft thresholding są nieznaczne. 5. Podsumowanie W pracy przedstawiono metodę odszumiania sygnałów przyśpieszeń drgań rejestrowanych jednocześnie dla tego samego obiektu. Porównując sygnały przed i po odszumianiu można zauważyć, że dla obu czujników, na tym samym kierunku (zgodnym z osią wału wyjściowego) rejestracji drgań, występuje najwięcej zakłóceń. Odszumianie sygnałów wykonano dwoma metodami soft i hard thresholding. Następnie sieć neuronowa miała za zadanie rozróżnić stan przekładni. Dla obu metod uzyskano bardzo zbliżone i zadowalające wyniki. Jakość klasyfikacji wyniosła ok. 96%. Aby zweryfikować skuteczność metody wykonano taką samą procedurę dla sygnału oryginalnego. Poprawność klasyfikacji była mniejsza w przybliżeniu o 5%. Literatura 1. 2. 3. 4. Aminghafari M, Cheze N, Poggi J M. Multivariate denoising using wavelets and principal component analysis. Computational Statistics & Data Analysis 2006; 50: 2381-2398. Antoni J, Bonnardot F, Raad A, El Badaoui M. Cyclostationary modelling of rotating machine vibration signals. Mechanical Systems and Signal Processing 2004; 18: 1285-1314. Antoni J, Randall R B. Unsupervised noise cancellation for vibration signals: part I-evaluation of adaptive algorithms. Mechanical Systems and Signal Processing 2004; 18: 89–101. Barszcz T. Decomposition of vibration signals into deterministic and nondeterministic components and its capabilities of fault detection and identification. Int. J. Appl. Math. Comput. Sci. 2009; 19(2): 327-335. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. Batko W, Dąbrowski Z, Engel Z, Kiciński J, Weyna S. Nowoczesne metody procesów wibroakustycznych. Radom: Wydawnictwo Instytutu Technologii Eksploatacji, 2005. Białasiewicz J T. Falki i aproksymacje. Warszawa: WNT, 2004. Bilgin S, Colak O H, Koklukaya E, Ari N. Efficient solution for frequency band decomposition problem using wavelet packet in HRV. Digital Signal Processing 2008; 18: 892-899. Jedliński Ł, Jonak J. Optimum choice of signals’ features used in toothed gears’ diagnosis. Diagnostyka 2010; 55: 9-12. Jedliński Ł, Jonak J. Quality evaluation of the bevel gear assembly based on analysis of the vibration signal. Diagnostyka 2010; 53: 23-26. Krukowski W, Józefczyk I. Transformacja falkowa w diagnostyce urządzeń mechanicznych. Diagnostyka 2008; 46: 75-82. Łazarz B, Wojnar G, Czech P. Early fault detection of toothed gear in exploitation conditions. Eksploatacja i Niezawodność – Maintenance and Reliability 2011; 1: 68-77. Mallat S G. A theory for multiresolution signal decomposition: the wavelet representation. IEEE Transactions on Pattern Analysis and Machine Intelligence 1989; 7: 674-693. Misiti M, Misiti Y, Oppenheim G, Poggi J M. Wavelet toolboxTM 4. User’s guide. The MathWorks, Inc. 2010. Peng Z K, Chu F L. Application of the wavelet transform in machine condition monitoring and fault diagnostics: a review with bibliography. Mechanical Systems and Signal Processing 2004; 18: 199-221. Seker S, Ayaz E. Feature extraction related to bearing damage in electric motors by wavelet analysis. Journal of the Franklin Institute 2003; 340: 125-134. Wang Y, Zuo M, Lei Y, Fan X. Improvement of local mean approximation in empirical mode decomposition for gear fault detection. Eksploatacja i Niezawodność – Maintenance and Reliability 2010; 2: 59-66. Wu J D, Kuo J M. An automotive generator fault diagnosis system using discrete wavelet transform and artificial neural network. Expert Systems with Applications 2009; 36: 9776-9783. Zimroz R. Zastosowanie analizy falkowej w diagnostyce uszkodzeń lokalnych układów napędowych maszyn górniczych. Diagnostyka 2009; 49: 113-122. Dr inŜ. Andrzej Tomporowski Wydział InŜynierii Mechanicznej UTP w Bydgoszczy Ul. Prof. S. Kaliskiego 7/ bud. 3.2.pok 202 85- 788 Bydgoszcz e-mail: [email protected] Strumień wydajności wielotarczowego rozdrabniania ziaren ryŜu Słowa kluczowe: rozdrabnianie, biomateriały, wydajność Streszczenie: Poszukiwania rozwiązań konstrukcyjnych zespołów rozdrabniających ziarna zbóŜ, prowadzące do energooszczędnych procesów produkcyjnych uzasadniają podjęcie badań nad doskonaleniem teorii i konstrukcji rozdrabniaczy. Istotny wpływ na wydajność, funkcjonalność i sprawność procesu rozdrabniania ziarna ryŜu mają trudne do opisania zjawiska zachodzące w przestrzeni roboczej rozdrabniacza. Pomimo, Ŝe dostępne są opracowania na temat podstaw funkcjonalności i badań rozdrabniaczy wielotarczowych [m.in.1,2,4], jak dotychczas nie podejmowano próby opisu wpływu cech i zachowań rozdrabnianego materiału na wydajność, sprawność i uŜyteczność procesu rozdrabniania biomateriałów. Podstawą do poprawy funkcjonalności działania maszyny rozdrabniającej jest przeprowadzone rozpoznanie moŜliwości istniejących rozwiązań oraz określenie zakresu, skutecznych w przetwórstwie, cech konstrukcyjnych zespołu roboczego z obszaru dopuszczalnego. Spełnienie kryteriów moŜe być osiągnięte między innymi na drodze celowego sterowania cechami konstrukcyjnymi zespołu wielotarczowego. Aby jednak działania te przyniosły planowane korzyści, konieczne staje się poznanie zaleŜności oraz opracowanie opisu matematycznego przepływu rozdrabnianego ziarna przez przestrzeń roboczą rozdrabniacza wielotarczowego, jako zmiennej wynikowej konstrukcji i działania zespołu roboczego. 1. Wprowadzenie Innowacyjne podejście w badaniach konstrukcyjnych polega na projektowaniu celów, środków lub technik działania, które nie weszły jeszcze w skład praktyki, a tym bardziej w skład wiedzy inŜynierskiej. Zabiegi weryfikacyjne zaś mogą dotyczyć takich problemów rozdrabniania, jak dynamika zmian długości początkowej ziaren, transport materiału rozdrabnianego (ziaren o róŜnej długości) z jednoczesnym quasi-ścinaniem pomiędzy kolejnymi tarczami, strumień cząstek opuszczających maszynę z daną intensywnością i w określonych warunkach. W ten sposób wiedza o konstrukcji rozdrabniaczy, jako nauka praktyczna i empiryczna bada nie tylko to co jest, lecz takŜe to co mogłoby być, a co na razie istnieje tylko w sferze koncepcji [1, 2, 3, 5]. Cel badań. Podstawowym celem badań było określenie czynników, cech konstrukcyjnych i uŜytkowych oraz ich wpływu na dynamikę oraz wydajność procesu rozdrabniania biosurowców ziarnistych (ryŜ długoziarnisty). Analizy dokonano na przykładzie rozdrabniacza wielotarczowego wykonanego wg patentu UTP w Bydgoszczy [3]. Dodatkowo, celem pracy było przybliŜenie złoŜoności zjawisk, procesów i relacji rozdrabniania wielotarczowego. 2. Model rozdrabnianych ziaren ryŜu Materiałem wsadowym w prowadzonych badaniach wydajności (objętościowej, masowej) są ziarna ryŜu długoziarnistego o stabilizowanych parametrach zarówno wilgotnościowych, oraz konfekcjonowanej wielkości. ZałoŜono więc, Ŝe dla potrzeb niniejszego rozwaŜania wyjściowym modelem wsadu, rozdrabnianego ziarna będzie substancja rozdrabniana w jednym wymiarze o powtarzalnej, stałej wielkości pojedynczego ziarna. Przyjmuje się, Ŝe ziarna ryŜu ustawiają się w otworach tarcz rozdrabniających, wzdłuŜ swojej dłuŜszej osi, prostopadle do międzytarczowej płaszczyzny cięcia. Granulometryczny stan rozdrabnianego medium ρ nm w przestrzeni roboczej rozdrabniacza wielotarczowego, wielootworowego opisuje rozkład prawdopodobieństwa długości ziarna. PoniewaŜ materiał w otworach tej samej tarczy podlega takiemu samemu procesowi cięcia w kaŜdym otworze, jego stan indeksowany jest numerem tarczy (n) oraz numerem cięcia (m): ρ nm : (0, l max ] → [0,1], ∫ l max 0 ρ nm dl = 1. (1) Przemieszczenie materiału do następnej tarczy wiąŜe się z przejściem przez proces cięcia, zatem n ≤ m. Początkowy stan granulometryczny materiału w otworze pierwszej tarczy (wszystkie ziarna są całe, nie rozdrobnione) przed pierwszym cięciem wynosi ρ 00 i jest wejściowym stanem materiału, który w dalszej części opracowania zostanie opisany pewną funkcją skupioną wokół wartości długości lmax. Rys.1. Technika (sposób i środek konstrukcyjny) ustawienia tarcz w rozdrabniaczu przy quasi – ścinaniu materiału ziarnistego [2]; 1- tarcza poprzedzająca, 2 - tarcza następująca, 3 - wejściowy stan materiału, 4 - następny stan materiału, di - średnica otworu, t01, t01, tr1 - podziałka , h1, h2 - grubość tarcz, s – szczelina, v1, v1 – prędkość liniowa Wyjściowy stan materiału podczas rozdrabniania zmienia się na skutek dwóch czynników, mechanizmów (Rys.1): rozdrabniania (quasi-ścinania) oraz usuwania ziaren o rozmiarze poŜądanym i mniejszym, z przestrzeni międzytarczowych zespołu rozdrabniającego urządzenia. 3. Proces rozdrabniania W momencie, w którym spotykają sie otwory z dwóch sąsiednich tarcz (Rys.2), a ich wzajemny przekrój zaczyna sie powiększać (0, max), następuje wypełnianie następującego otworu częścią materiału rozdrabnianego z otworu poprzedzającego. śeby proces cięcia zachodził wydajnie, efektywnie, otwór w tarczy poprzedzającej musi być całkowicie wypełniony w momencie, gdy zaczyna sie zmniejszać wzajemny przekrój otworów (max, 0). W tym celu powinny być spełnione trzy warunki: 1. Ilość materiału w otworze poprzedzającym musi być zawsze większa od dostępnego miejsca w otworze następującym. W przeciwnym wypadku moŜna sobie wyobrazić przesypywanie się materiału z otworu tarczy poprzedzającej do otworu tarczy następującej bez jego cięcia na całej długości urządzenia. PoniewaŜ część materiału jest usuwana do przestrzeni pomiędzy tarczami, objętość otworów w tarczy następującej powinna być zawsze mniejsza niŜ w tarczy poprzedzającej (realizuje się to poprzez mniejszy przekrój otworu lub mniejszą grubość tarczy, tu jednak ograniczeniem są względy konstrukcyjne, napręŜenia, które tarcza musi przenosić). Grubość (h) n-tej tarczy, dla potrzeb obliczeń, oznaczono przez yn, natomiast przez ~yn( k ) oznaczono wysokość, do której materiał wypełnia otwór w n-tej tarczy przed k-tym cięciem. 2. KaŜdy punkt przekroju otworu następującego powinien w pewnym momencie ruchu względnego tarcz znaleźć się na obszarze przekroju otworu z tarczy poprzedzającej. 3. W momencie, gdy następuje cięcie, otwór w tarczy następującej jest zamknięty na przejście kolejną tarczą (nie powstają tzw. korytarze otworowe). Rys.2. Pole przekroju i efektywna powierzchnia rozdrabniania dla dwóch tarcz rozdrabniania między krawędziami otworów rozdrabniających [2], FR – międzyotworowe pole powierzchni cięcia (np. F1, F2) Wypełnienie otworu w warstwie zmienia sie następująco (Rys.3). Z rozwaŜań wynika, Ŝe k-te cięcie za tarczą (n−1)-szą następuje we wcześniejszym momencie niŜ k-te cięcie za tarczą n-tą. Przy tak przyjętej numeracji cięcia, quasi-ścinania (na kaŜdej tarczy numeracja zaczyna sie od pierwszego cięcia), ziarno moŜe na kaŜdej granicy poszczególnych tarcz podlegać cięciu o tym samym numerze. Po wypełnieniu otworu zaczyna się zmniejszanie części wspólnej przekrojów otworów, proces cięcia (quasi-ścinania). Wstępnie załoŜono, Ŝe kaŜde ziarno w przekroju otworów rozdrabniających podlega procesowi cięcia. PołoŜenie ziarna względem płaszczyzny, w której odbywa sie cięcie jest losowe z rozkładem równomiernym. Ziarno kaŜdej długości rozpadnie (rozdrobni) się z jednakowym prawdopodobieństwem na dwie mniejsze cząstki, o sumie długości będącej długością (wymiarem) przed cięciem. Cięcie, a w zasadzie quasi-ścinanie następuje zawsze w materiale, który przed zasypaniem był w poprzedzającej tarczy. Rozkład długości ziarna, przy cięciu, w materiale, który uzupełnił pustą przestrzeń w tarczy (n+1)-szej zmienia się według zaleŜności: lmax m x 1 ρ (x ) + ρ~nm+1 ( x ) = An,m ρ nm = 1 − ρ nm (l )dl , (2) m n m ∫x ~ ~ y − y y − y n +1 n +1 n +1 n +1 natomiast w materiale, który pozostał w tarczy n-tej: x 1 lmax ~ ρ~nm+1 (x ) = Bn,m ρ nm = 1 − ~ m ρ nm (x ) + ~ m ∫ ρ n (l )dl , (3) yn yn x Funkcje, które otrzymano są nieujemne, jako suma dwóch nieujemnych składników. Całkując od 0 do l moŜna łatwo sprawdzić, Ŝe są to rozkłady prawdopodobieństwa: lmax x 1 lmax lmax x 1 lmax x ~ m+1 ∫0 ρ n (x )dx = 1 − ~ynm + ~ynm ∫0 ∫x ρ n (l )dldx = 1 − ~ynm + ~ynm ∫0 ∫0 ρ n (l )dxdl = (4) 1 lmax x 1 − ~ m + ~ m ∫ xρ n (l )dx = 1 yn yn 0 ~ i tak samo dla rozkładu ρ~ m . Zatem operatory A i B są poprawnie określonymi n +1 n ,m n ,m operatorami stochastycznymi. Rys.3. Schemat transportu materiału rozdrabnianego z quasi-ścinaniem pomiędzy kolejnymi tarczami; gdzie ↗ oznacza fazę cięcia Dla uproszczenia załoŜono, Ŝe po cięciu rozkład długości ziarna w tarczy (n+1)-szej będzie jednorodny (frakcja cięta i ta która była w otworze przed cięciem się wymieszają), będzie ona zatem średnią waŜoną z ρ nk+1 i ρ nk : ~ ym y −~ y nm+1 ρ nm+1 (x ) = n+1 ρ nm+−11 + n An,m ρ nm ( x ) (5) y n+1 y n+1 4. Proces usuwania Po rozcięciu dwie warstwy materiału przesuwają się względem siebie zgodnie z kierunkiem obrotów sąsiednich tarcz i gradientem wzajemnych prędkości. Drobiny materiału są usuwane z otworu poprzedzającego (działa na nie siła składowa siły grawitacji o kierunku prostopadłym do szczeliny międzytarczowej), natomiast nie są usuwane z otworu następującego (poniewaŜ składowa prostopadła siły grawitacji jest w kierunku od szczeliny). Po cięciu rozkład długości będzie zatem wynosił: ρ~nm+1 ( x ) = Bn ,m ρ n ,m ρ~ m+1 ( x ) lmax ρ~ m +1 ( x ) ∫l = n min n 0 −1 x〉 l max x〈lmin (6) Poziom materiału po m-tym cięciu (przed m+1 cięciem) w n-tej szczelinie, ~y nm +1 wynosi: lmin ~ Bn ,m ρ nm ( x )xdx ∫ k ~y m+1 = ( y − y + ~ 0 (7) y n +1 )1 − lmax ~ n n n +1 m ∫0 Bn,m ρ n (x )xdx śeby uzyskać rozkład w całej przestrzeni otworu przed (m+1)-szym cięciem (po ponownym zasypaniu), naleŜy wziąć średnią waŜoną: ~ y −~ y nm+1 y m+1 ρ nm+1 ( x ) = n An−1,m ρ nm−1 + n Bn ,m ρ nm (8) yn yn Operator Bn ,m nie jest juŜ operatorem liniowym jak An ,m , poniewaŜ zaleŜy od poziomu materiału, który pozostał w tarczy n-tej po m-tym cięciu: ~y m +1 , i jest funkcją n rozkładu prawdopodobieństwa w materiale (który wpływa na to, jaka część materiału będzie usuwana z maszyny przy cięciu) [6]. śeby móc traktować Bn ,m jako operatory liniowe, wielkości y nm naleŜy traktować, w kaŜdym kroku procedury, jako z góry zadane i iteracyjnie uzgadniane z wynikami badań, doświadczeń. 5. Strumień cząstek opuszczających maszynę Przy cięciu m-tym, strumień cząstek opuszczających maszynę przez szczelinę (na przerwie) pomiędzy tarczą n-tą i (n+1)-szą jest dany rozkładem prawdopodobieństwa: lmax ρ (l )dl lmin lmax ρ (l )dldx ∫0 ∫x m n n sn ( x ) = ∫x 0, −1 l 〈l min l 〉 l min (9) a jego objętość jest równa: lmin ( V = y n − y n +1 + ~ y nk+1 )∫ ∫ 0 lmax 0 ~ B n ,m ρ nm ( x )xdx ~ Bn ,m ρ nm ( x )xdx (10) 6. Przemiany rozkładów długości w tarczach Rozkład długości ziaren w n-tej tarczy po m-tym cięciu wyraŜono oznaczeniem ρ nm . Operatory przeprowadzające rozkład ρ nm w ρ nm+1 i w ρ nm +1 są oznaczone odpowiednio jako An ,m i Bn ,m . Zmiany rozkładów ilustruje diagram wydajności (Rys.4). ↘ ↘ ↘ B0,0 → B0,0 → B0,0 → ↘ ↘ ↘ B1,0 → B1,1 → B1,2 → ↘ ↘ ↘ B2,0 → B2,1 → B2,2 → B0,0 → B1,3 → B2,3 → Rys.4. Wydajnościowy diagram rozkładów długości ziaren, produktów quasi-ścinania, w poszczególnych tarczach Stan ρ 00 jest dany (rozkład długości ziaren w materiale wejściowym). Stan ρ nm uzyskano z działania sum iloczynów operatorów A i B na stanie ρ 00 . Iloczyny reprezentują wszystkie drogi, którymi moŜna dojść do tego stanu ze stanu ρ 00 . Dla przykładu: ρ 23 = (A1, 2 A0,1 B0,0 + A1, 2 B1,1 A0,0 + B2, 2 A1,1 A0.0 )ρ 00 (11) Do stanu ρ nm moŜna dojść stosując n razy operator A i m razy operator B 1. Drogę określa jednoznacznie podzbiór n-elementowy zbioru kroków w całej drodze {1, . . . ,m}, kroków, w których działa operator A. Ilość takich podzbiorów, zatem iloczynów w sumie m wynosi . W ogólności operatory A i B nie komutują. Dla dalszych rozwaŜań kluczowe n będzie obliczanie takich wyraŜeń. Stany ρ są dodatnimi elementami przestrzeni Banacha1 L (0, 1), a operatory A i B są endomorfizmami tej przestrzeni. Z przyczyn praktycznych moŜna przybliŜyć stany ρ dodatnimi elementami R d , a operatory A i B macierzami d×d. Fizycznie odpowiada to podziałowi cząstki o długości l na d niepodzielnych juŜ dalej części o długości l/d – odpowiadającej liniowej mierze rozdrobnienia. 1 precyzyjniej, średnie waŜone wyniku działania tego operatora 7. Przepływ energii Przy załoŜeniu nieskończenie szybkiego rozchodzenia sie napręŜeń, w momencie, gdy zaczyna sie zamykanie obszaru wspólnego dwóch otworów następuje w kaŜdym ziarnie równomierne zgęszczanie materiału, aŜ do przekroczenia progowego zgęszczenia, po którym następuje pęknięcie ziarna. Po pęknięciu następuje przesuwanie sie warstw względem siebie, przy czym siła tarcia jest stała na jednostkę sąsiadujących powierzchni dwóch warstw. Oprócz tego dochodzi stałe tarcie materiału o powierzchnie tarcz, w miejscach, gdzie w tarczy następującej nie ma otworu. 8. Wnioski Zjawiska, procesy i relacje rozdrabniania wielotarczowego, pomimo swojej złoŜoności, naleŜą do dość podatnych na opisy formalne. Uzyskanie odpowiedzi na pytanie o czynniki procesowe (czynności i sposoby), o cechy konstrukcyjne (środki, urządzenia i instalacje), warunki uŜytkowe i ich wpływ na dynamikę oraz wydajność rozdrabniania ziaren ryŜu, na przykładzie rozdrabniacza wielotarczowego, było moŜliwe, przy załoŜeniu nieskończenie szybkiego rozchodzenia sie napręŜeń quasi-ścinania, w zakresie: – rozkładu prawdopodobieństwa długości ziarna, – rozkładu prawdopodobieństwa cząstek w strumieniu opuszczającym maszynę przez szczelinę między tarczami. „Praca naukowa finansowana ze środków Narodowego Centrum Badań i Rozwoju w latach 2010/2013 jako projekt rozwojowy” Literatura 1. Detyna J. Analysis of nonequilibrium stases in the sieve separation process. Eksploatacja i Niezawodnosc – Maintenance and Reliability 1(49)2011: 78-85 2. Flizikowski J. Konstrukcja rozdrabniaczy Ŝywności. Wydawnictwo Uczelniane Akademii Techniczno-Rolniczej w Bydgoszczy, Bydgoszcz 2005 3. Flizikowski J., Bieliński K., Bieliński M. PodwyŜszenie energetycznej efektywności wielotarczowego rozdrabniania nasion zbóŜ na paszę. Wydawnictwo ATR- OPO, Bydgoszcz 1994 4. Flizikowski J., Bieliński M. Rozdrabniacz wielotarczowy zwłaszcza do materiałów ziarnistych. Patent RP-144 566 5. Kaleta A., Wojdalski J. Przetwórstwo rolno-spoŜywcze. Wybrane zagadnienia inŜynieryjno-produkcyjne i energetyczne. Warszawa 2007 6. Razavi S.M.A., Farahmandrar R. Effect of hulling and milling on the physical properties of rice grains. International Agrophysics, 2008, 22, 353-359 Prof. Zdzisław Chłopek Instytut Transportu Samochodowego w Warszawie 03–301 Warszawa, ul. Jagiellońska 80 E–mail: [email protected] Badania zagrożenia środowiska cząstkami stałymi podczas eksploatacji pojazdów samochodowych Słowa kluczowe: pyły, cząstki stałe, PM10, PM2.5, PM1, pojazdy samochodowe Streszczenie. W pracy przedstawiono wyniki badań imisji frakcji pyłów PM10, PM2.5 i PM1. Do modelowania imisji frakcji cząstek stałych PM2.5 i PM1 wykorzystano wyniki badań empirycznych, przeprowadzonych na stacjach nadzorowania jakości powietrza w aglomeracji czeskiego miasta Brna. Wyniki modelowania imisji frakcji cząstek stałych PM2.5 i PM1 nie umożliwiły sformułowania jednoznacznych wniosków, co dowodzi konieczności statystycznego potraktowania badanego problemu. Stwierdzono jednak istotną zależność modeli imisji frakcji cząstek stałych od źródeł emisji pyłów i warunków ich rozprzestrzenia. 1. Wstęp Zagrożenie środowiska przez pyły jest powszechnie znane. Szkodliwość pyłów dla zdrowia ludzi jest tematem wielu prac, dotyczących zarówno aspektów zdrowotnych [1, 10, 16, 18, 28, 31, 34, 35], jak i oceniających czynniki, mające wpływ na emisję pyłów [2, 3, 5–9, 11–15, 17, 19, 21, 22, 24, 27, 29]. Źródłami emisji pyłów są zarówno zjawiska naturalne, jak i działania cywilizacyjne. Do najważniejszych naturalnych źródeł emisji pyłów zalicza się przede wszystkim: wybuchy wulkanów, materiały osadowe, aerozole morskie, roślinne i zwierzęce, a także pożary lasów. W skali globalnej naturalne źródła emisji pyłów są dominujące, jednak w obszarach o szczególnie intensywnej działalności ludzi największy wpływ na zanieczyszczenie środowiska pyłami mają ich źródła antropogenne. Antropogennymi źródłami pyłów są właściwie wszystkie procesy produkcyjne i procesy spalania paliw. Istotną rolę w zagrożeniu środowiska pyłami, szczególnie w centrach wielkich aglomeracji miejskich, odgrywa motoryzacja. Szkodliwość pyłów dla zdrowia ludzi jest zależna od składu chemicznego i mineralogicznego oraz budowy fizycznej cząstek pyłu, a także od wymiarów cząstek pyłów [2, 3, 6, 7, 17, 21, 22, 35]. W zależności od umownych wymiarów cząstek pyłu wyróżnia się [2, 3, 6, 7, 17, 21, 22, 35]: pył całkowity TSP (total suspended particles) – mieszaninę małych cząstek stałych o umownych wymiarach mniejszych niż 300 m, zawieszonych w powietrzu (faza rozproszona układu dwufazowego ciało stałe – gaz), pył zawieszony PM10 – o umownych wymiarach mniejszych od 10 m, pył drobny PM2.5 – o umownych wymiarach mniejszych od 2,5 m, tzw. nanocząstki PM1 – o umownych wymiarach mniejszych od 1 m, stanowiące pył praktycznie niewidzialny [24, 29]. Cząstki stałe o umownych średnicach większych niż 10 m są zatrzymywane głównie w górnych drogach oddechowych, skąd są w znacznej części wydalane, cząstki PM10 (z wykluczeniem cząstek PM2.5) przenikają nawet do płuc i choć w płucach się nie kumulują, to gromadzą się w górnych drogach oddechowych. Cząstki stałe PM2.5 przenikają do najgłębszych partii płuc, gdzie są kumulowane. Cząstki stałe PM1 przenikają nawet do układu krwionośnego. Do szczególnie toksycznych cząstek stałych zalicza się pyły, zawierające związki metali ciężkich oraz wielopierścieniowe związki organiczne – wiele z tych związków ma właściwości rakotwórcze [31, 35] Oprócz negatywnego oddziaływania pyłów na zdrowie ludzi i zwierząt stwierdza się również szkodliwe wpływanie pyłów na rośliny oraz na glebę i wody. Wspólnie z dwutlenkiem siarki, tlenkiem węgla i innymi związkami pyły przyczyniają się do powstawania zjawiska smogu londyńskiego [2, 11]. Pyły mają również wpływ osłabiający na zjawisko cieplarniane w atmosferze [2, 11]. Należy także zwrócić uwagę na fakt, że pyły ograniczają widoczność, co ma wpływ na bezpieczeństwo ruchu drogowego. Szkodliwość zanieczyszczeń powietrza ocenia się na podstawie imisji zanieczyszczeń – stężenia zanieczyszczeń rozproszonych w powietrzu, mierzonego na wysokości 1,5 m nad powierzchnią podłoża [26]. Przekroczenia dopuszczalnych imisji cząstek stałych PM10 są w państwach rozwiniętych gospodarczo jednymi z najczęściej występujących powodów podejmowania przez władze działań naprawczych, dotyczących jakości środowiska. Od 2009 r. w Unii Europejskiej obowiązuje również limitowanie imisji cząstek stałych PM2.5. W przyszłości planuje się także objęcie kontrolą imisji cząstek stałych PM1. Ocena poszczególnych źródeł emisji pyłów ze względu na szkodliwość dla środowiska jest bardzo trudna, jako że badania jakości powietrza w poszczególnych miejscach obejmują wpływ wszystkich występujących źródeł. Ponadto na jakość powietrza mają wpływ również warunki rozprzestrzeniania zanieczyszczeń. Z tego powodu jest celowe prowadzenie badań porównawczych w miejscach, charakteryzujących się różnym udziałem źródeł emisji zanieczyszczeń i ich rozprzestrzeniania. Na podstawie analiz wyników takich badań istnieje możliwość wnioskowania o wpływie poszczególnych źródeł emisji zanieczyszczeń na stan jakości powietrza. Podstawową trudnością w tego rodzaju badaniach jest stosunkowo uboga sieć stacji nadzorowania jakości powietrza, prowadzących ciągłe pomiary imisji kompletnych frakcji cząstek stałych, czyli – zgodnie z obecnym stanem – PM10, PM2.5 oraz PM1. W celu oceny wpływu poszczególnych źródeł emisji pyłów na imisję frakcji cząstek stałych jest dodatkowo konieczne wykonywanie pomiarów z częstotliwością, umożliwiającą identyfikację właściwości dynamicznych procesów opisujących zjawiska, będące przyczyną emisji pyłów, np. ruchu pojazdów samochodowych. Ocenia się, że do takich celów jest konieczne wykonywanie pomiarów z odstępem czasu co najwyżej 1 h. Również wymagania są stawiane czasowi badań, jest bowiem celowe uwzględnianie zmienności procesów determinujących emisję antopogenną pyłów, związanej z cyklem tygodniowym, a także zmienności wynikającej z pór roku. Dodatkowo badania długoterminowe mogą skutecznie przyczynić się do zmniejszenia wpływu na wyniki badań zakłóceń, związanych z czynnikami przypadkowymi, do których można zaliczyć warunki pogodowe. Jak wiadomo, fluktuacje czynników pogodowych mają w danym obszarze charakter normalny, w związku z czym ich wartość oczekiwana dąży do zera wraz z wydłużaniem się czasu obserwacji. W związku z tym jest celowe, aby pomiary były prowadzone w czasie co najmniej jednego roku. Spośród licznych stacji nadzorowania jakości powietrza, jakie znajdują się w Europie, istnieją takie aglomeracje miejskie, w których znajduje się kilka stacji, umieszczonych w miejscach o zróżnicowanym charakterze źródeł emisji i rozprzestrzeniania zanieczyszczeń. Dodatkowo na stacjach tych jest prowadzone badanie kompletnych frakcji cząstek stałych oraz zanieczyszczeń, takich jak tlenki azotu i tlenek węgla, których imisję postuluje się jako związaną z imisją frakcji cząstek stałych [5–9, 12]. Taki zbiór stacji nadzorowania jakości powietrza znajduje się m.in. w aglomeracji czeskiego miasta Brna. W niniejszej pracy są wykorzystywane do analiz imisji frakcji cząstek stałych wyniki z trzech stacji nadzorowania jakości powietrza w Brnie. Brno jest miastem na Morawach w Republice Czeskiej. Jest położone w południowo– wschodniej równinnej części kraju u zbiegu rzek Svratki i Svitavy. Liczba mieszkańców Brna wynosi ponad 400000 (2008 r.). Brno zajmuje powierzchnię o polu 230 km2. Stacje nadzorowania jakości powietrza są własnością Wydziału Ochrony Środowiska Magistratu miasta Brna. W badaniach wykorzystano wyniki pomiarów ze stacji: Svatoplukova, Zvonařka i Lány. Stacje te różnią się charakterem obszaru, na którym się znajdują. Stacje Svatoplukova i Zvonařka znajdują się przy dużych arteriach komunikacyjnych, szczególnie stacja Zvonařka, położona bezpośrednio przy ulicy. Stacja Lány jest położona w miejscu odległym od ruchliwych ulic. Na stacjach tych są dokonywane pomiary imisji m.in. cząstek stałych PM10, PM2.5 i PM1 oraz tlenków azotu i tlenku węgla co 1 h. W niniejszym opracowaniu nie są wykorzystywane wyniki badań imisji tlenków azotu i tlenku węgla, stosowane do opracowywania behawiorystycznych [4] modeli imisji cząstek stałych PM10 [6–9], przekraczałoby to bowiem możliwości objętościowe publikacji. Wybór stacji jednak celowo uwzględniał możliwość rejestracji imisji tlenków azotu i tlenku węgla, dzięki czemu można uzyskać kompletne materiały do modelowania imisji poszczególnych frakcji cząstek stałych. 2. Modelowanie imisji cząstek stałych PM2.5 i PM1 Stan zagrożenia środowiska można oceniać albo na podstawie bezpośrednich pomiarów imisji zanieczyszczeń, jednak ocena dotyczy w takim wypadku jedynie miejsca i czasu dokonywania pomiarów, a uogólnianie wyników badań nie zawsze jest dostatecznie uprawnione. Często wyniki długotrwałych badań w miejscach o typowych warunkach emisji zanieczyszczeń i ich rozprzestrzeniana uprawniają jednak do uogólniania wniosków. W takich wypadkach podstawą oceny zanieczyszczenia środowiska są wyniki badań oraz modelowania imisji zanieczyszczeń. W innych wypadkach zagrożenie środowiska ocenia się na podstawie znajomości emisji zanieczyszczeń i modelowania ich rozprzestrzeniania. Znajomość emisji zanieczyszczeń jest możliwa dzięki wynikom pomiarów i w tym wypadku występują te same ograniczenia jak w wypadku pomiarów imisji. Zupełnie inna sytuacja jest dla mobilnych źródeł emisji, np. samochodów. W tym wypadku jest możliwe tylko modelowanie emisji zanieczyszczeń. Ponieważ modelowanie emisji zanieczyszczeń stanowi podstawowe narzędzie w ocenie zagrożenia stanu środowiska w miejscach najbardziej narażonych, tzn. w centrach miast, tradycyjnie wszystkie rodzaje modelowania, związanego z emisją zanieczyszczeń, nazywa się modelowaniem emisji zanieczyszczeń, choć w wielu wypadkach formalnie modelowanie dotyczy imisji. Takie uproszczenie znajduje uzasadnienie w możliwości zwięzłego formułowania sądów, mimo że formalnie jest nieścisłe. Modelowanie emisji cząstek stałych PM10 nie jest tematem niniejszej pracy, jednak jest nierozerwalnie związane z modelowaniem imisji cząstek stałych PM2.5 i PM1. Modelowanie emisji cząstek stałych PM10 ma bogatą literaturę [6–9, 12–15, 21, 22, 27]. Stosowane są dwie metody badań: modelowanie emisji cząstek stałych PM10 na podstawie wiedzy o ruchu i właściwościach pojazdów i dróg – modele tworzone na zasadzie podobieństwa strukturalnego [4], modelowanie imisji cząstek stałych PM10 na podstawie imisji tlenków azotu i tlenku węgla – modele tworzone na zasadzie podobieństwa funkcjonalnego (behawiorystyczne) [4]. Modele tworzone na zasadzie podobieństwa strukturalnego uwzględniają następujące źródła emisji cząstek stałych PM10 [6, 7, 13–15, 27]: samochody, nawierzchnia dróg, zanieczyszczenia stałe znajdujące się na drodze – w postaci wzniecania pyłów. Źródłami pyłów, pochodzących z samochodu, są [2, 3, 5–9, 13–15, 27]: silnik spalinowy – cząstki stałe są zawarte w spalinach [11, 23, 29], pary trące – przede wszystkim w układzie hamulcowym [2, 3] oraz w sprzęgle, ogumienie kół jezdnych, inne części samochodu, ulegające zużyciu eksploatacyjnemu. Modele behawiorystyczne nie uwzględniają w jawny sposób źródeł emisji cząstek stałych, zarówno tych, które są związane z motoryzacją, jak i pozostałych. W modelach behawiorystycznych wykorzystuje się istotną statystyczną zależność imisji cząstek stałych i imisji innych zanieczyszczeń, przy czym w teorii i praktyce techniki motoryzacyjnej istnieje – przynajmniej częściowo – uzasadnienie takiej zależności, jak np. jednoczesne zwiększanie się emisji cząstek stałych z silników spalinowych i innych źródeł samochodu oraz emisji tlenku węgla i tlenków azotu w miarę zwiększania się prędkości samochodu i – w związku z tym – obciążenia silnika. W modelach behawiorystycznych postuluje się zazwyczaj liniową zależność imisji cząstek stałych i innych zanieczyszczeń. W ogólności wyniki analizy modeli zbudowanych na zasadzie podobieństwa strukturalnego nie mogą być porównywane z wynikami analizy modeli zbudowanych na zasadzie podobieństwa funkcjonalnego, jako że w modelach strukturalnych nie uwzględnia się w jawny sposób innych źródeł emisji pyłów niż związanych z ruchem pojazdów samochodowych. W rzeczywistości duża skala dowolności przyjmowania współczynników modeli strukturalnych, trudnych zazwyczaj do identyfikacji, powoduje, że to jest istotny powód nieporównywalności wyników analizy modeli strukturalnych i funkcjonalnych. Frakcję cząstek stałych PM2.5 można traktować jako podzbiór frakcji PM10. W związku z tym postuluje się liniową zależność imisji cząstek stałych PM2,5 – IPM2,5 od imisji cząstek stałych PM10 – IPM10: IPM 2,5 k 2.510 IPM10 (1) gdzie: kPM2.5–10 – współczynnik modelu emisji cząstek stałych PM2.5; k PM 2.510 0;1 Podobnie jak w wypadku modelowania imisji cząstek stałych PM2.5, cząstki stałe PM1 można potraktować jako podzbiór cząstek stałych PM10 i cząstek stałych PM2.5. Imisję cząstek stałych PM1 – IPM1 można zatem modelować jako liniowo zależną od imisji cząstek stałych PM10: IPM1 k PM110 IPM10 (2) gdzie: kPM1–10 – współczynnik modelu emisji cząstek stałych PM1; k PM110 0;1 oraz jako liniowo zależną od imisji cząstek stałych PM2.5: IPM1 k PM1 2.5 IPM2.5 (3) gdzie: kPM1–2.5 – współczynnik modelu emisji cząstek stałych PM1; k PM1 2.5 0;1 . Identyfikacja modeli imisji cząstek stałych PM2.5 (1) i imisji cząstek stałych PM1 (2 i 3) polega na wyznaczeniu współczynników modeli kPM2.5–10, kPM1–10 i kPM1–2.5 na podstawie wyników badań empirycznych imisji frakcji cząstek stałych PM10, PM2.5 i PM1. Wyniki identyfikacji są w ogólności zależne od warunków emisji zanieczyszczeń i ich rozprzestrzeniania, a także od okresu dokonywania pomiarów. 3. Badania imisji cząstek stałych PM10, PM2.5 i PM1 na wybranych stacjach monitorowania jakości powietrza Badania na stacjach nadzorowania jakości powietrza w Brnie, wykorzystywane w niniejszej pracy, były dokonywane od 1 stycznia do 31 grudnia 2010 r. z odstępem próbkowania wynoszącym 1 h. Na wykresach 1–3 przedstawiono przebiegi imisji frakcji cząstek stałych dla wartości uśrednionych w okresie 1 tygodnia w dziedzinie czasu t, oznaczonego numerami dni – d i miesięcy – m. 120 IPM10 IPM2.5 IPM1 IPM10, IPM2.5, IPM1 [g/m3] 100 80 60 40 20 0 1-01 29-01 26-02 26-03 23-04 21-05 18-06 16-07 13-08 10-09 8-10 5-11 3-12 31-12 t [d-m] Rys. 1. Przebieg imisji I cząstek stałych PM10, PM2.5 i PM1 na stacji nadzorowania jakości powietrza Brno–Svatoplukova 100 IPM10 IPM2.5 IPM1 90 IPM10, IPM2.5, IPM1 [g/m3] 80 70 60 50 40 30 20 10 0 1-01 29-01 26-02 26-03 23-04 21-05 18-06 16-07 13-08 10-09 8-10 5-11 3-12 31-12 t [d-m] Rys. 2. Przebieg imisji I cząstek stałych PM10, PM2.5 i PM1 na stacji nadzorowania jakości powietrza Brno–Zvonařka 80 IPM10, IPM2.5, IPM1 [g/m3] 70 60 50 40 30 20 10 IPM10 IPM2.5 IPM1 0 1-01 29-01 26-02 26-03 23-04 21-05 18-06 16-07 13-08 10-09 8-10 5-11 3-12 31-12 t [d-m] Rys. 3. Przebieg imisji I cząstek stałych PM10, PM2.5 i PM1 na stacji nadzorowania jakości powietrza Brno–Lány Przebiegi imisji frakcji cząstek stałych wskazują na silną zależność imisji od pór roku: znacznie większa jest imisja w zimnych porach roku. Również zauważa się zależności imisji frakcji cząstek stałych od dni tygodnia, co wskazuje na silny wpływ na tę imisję czynników cywilizacyjnych. Wyraźny jest wzajemny związek imisji poszczególnych frakcji, co uzasadnia przyjęcie modeli liniowych (1–3). Na rysunkach 4–6 przedstawiono charakterystykę statystyczną badanych zbiorów imisji frakcji cząstek stałych1: wartość minimalną, wartość maksymalną, wartość średnią, odchylenie standardowe i rozstęp. 1 W statystyce, szczególnie w komercyjnych zastosowaniach, stosuje się mało formalne nazewnictwo, nie zawsze będące w zgodzie ze sformalizowaną matematyką. W związku z tym określenia typu „wartość maksymalna” należy traktować jako „wartość największą”, a wartość minimalna” jako „wartość najmniejszą”, nie są to bowiem wartości ekstremalne w sensie pojęć występujących w analizie matematycznej. Jednak ze względu na rozpowszechnienie tego typu nazw i możliwość uzyskania dzięki temu zwartości wypowiedzi w niniejszej pracy określenia te są w opisach używane. min, max, AV, D, D [g/m3] 100 Svatoplukova Zvonařka Lány 80 60 40 20 0 min max AV D DD Rys. 4. Charakterystyka statystyczna stężenia cząstek stałych PM10: min – wartość minimalna, max – wartość maksymalna, AV – wartość średnia; D – odchylenie standardowe, D – rozstęp min, max, AV, D, D [g/m3] 100 Svatoplukova Zvonařka Lány 80 60 40 20 0 min max AV D D D Rys. 5. Charakterystyka statystyczna stężenia cząstek stałych PM2.5: min – wartość minimalna, max – wartość maksymalna, AV – wartość średnia; D – odchylenie standardowe, D – rozstęp min, max, AV, D, D [g/m3] 100 Svatoplukova Zvonařka Lány 80 60 40 20 0 min max AV D DD Rys. 6. Charakterystyka statystyczna stężenia cząstek stałych PM1: min – wartość minimalna, max – wartość maksymalna, AV – wartość średnia; D – odchylenie standardowe, D – rozstęp Istnieją duże różnice ekstremalnych wartości imisji poszczególnych frakcji cząstek stałych. Najmniejsze wartości wielkości: wartości maksymalnej, średniej i minimalnej imisji poszczególnych frakcji są dla stacji Lány (oprócz wartości średniej imisji cząstek stałych PM1 i wartości minimalnej imisji cząstek stałych PM10, których wartości są bardzo zbliżone do wartości dla stacji Zvonařka). Interesującym wynikiem jest fakt, że największe wartości maksymalna i średnia imisji wszystkich frakcji występują na stacji Svatoplukowa, położonej w obszarze o mniej intensywnym ruchu drogowym niż stacja Zvonařka. Z przebiegów imisji frakcji cząstek stałych wynika ich silna korelacja. Potwierdza to analiza korelacji. Analizę przeprowadzono z zastosowaniem teorii korelacji liniowej Pearsona [30] oraz metod nieparametrycznych [33]: korelacji rang Spearmana [32], korelacji tau Kendalla [20] i korelacji gamma Kruskala [25]. Na rysunkach 7–9 przedstawiono współczynniki korelacji Pearsona r, Spearmana R, Kendalla tau i Kruskala gamma między badanymi zbiorami. Svatoplukova Zvonařka Lány r, R, gamma, tau 1 0,8 0,6 0,4 0,2 0 r R gamma tau Rys. 7. Współczynniki korelacji Pearsona r, Spearmana R, Kendalla tau i Kruskala gamma między zbiorami imisji cząstek stałych PM10 i PM2.5 Svatoplukova Zvonařka Lány r, R, gamma, tau 1 0,8 0,6 0,4 0,2 0 r R gamma tau Rys. 8. Współczynniki korelacji Pearsona r, Spearmana R, Kendalla tau i Kruskala gamma między zbiorami imisji cząstek stałych PM10 i PM1 Svatoplukova Zvonařka Lány r, R, gamma, tau 1 0,8 0,6 0,4 0,2 0 r R gamma tau Rys. 9. Współczynniki korelacji Pearsona r, Spearmana R, Kendalla tau i Kruskala gamma między zbiorami imisji cząstek stałych PM1 i PM2.5 Prawdopodobieństwo nieodrzucenia hipotezy o braku korelacji między badanymi zbiorami jest we wszystkich wypadkach mniejsze od 1∙10–6. Wyniki analizy korelacji zbiorów imisji frakcji wymiarowych cząstek stałych w pełni uprawniają do formułowania sądu o silnej korelacji badanych zbiorów. Wartości współczynnika korelacji Pearsona dla poszczególnych zbiorów oraz prawdopodobieństwa nieodrzucenia hipotezy o braku korelacji wg Pearsona między badanymi zbiorami uzasadniają przyjęcie postulatu liniowych modeli imisji cząstek stałych PM2.5 i PM1. Na podstawie wyników badań empirycznych zidentyfikowano parametry modeli imisji cząstek stałych PM2.5 i PM1. Na wykresach 10–12 przedstawiono przebiegi współczynnika modeli imisji cząstek stałych PM2.5 i PM1 oraz wartość średnią tych współczynników w okresie badań. 1 kPM2.5-10, kPM1-2.5, kPM1-10 0,8 0,6 0,4 0,2 kPM2.5-10 kPM1-2.5 kPM1-10 AV[kPM2.5-10] AV[kPM1-2.5] AV[kPM1-10] 0 1-01 29-01 26-02 26-03 23-04 21-05 18-06 16-07 13-08 10-09 8-10 5-11 3-12 31-12 t [d-m] Rys. 10. Przebieg oraz wartość średnia AV współczynników k modeli emisji cząstek stałych PM2.5 i PM1 na stacji nadzorowania jakości powietrza Brno–Svatoplukova 1 kPM2.5-10, kPM1-2.5, kPM1-10 0,8 0,6 0,4 0,2 kPM2.5-10 kPM1-2.5 kPM1-10 AV[kPM2.5-10] AV[kPM1-2.5] AV[kPM1-10] 0 1-01 29-01 26-02 26-03 23-04 21-05 18-06 16-07 13-08 10-09 8-10 5-11 3-12 31-12 t [d-m] Rys. 11. Przebieg oraz wartość średnia AV współczynników k modeli emisji cząstek stałych PM2.5 i PM1 na stacji nadzorowania jakości powietrza Brno–Zvonařka 1 kPM2.5-10, kPM1-2.5, kPM1-10 0,8 0,6 0,4 kPM2.5-10 kPM1-2.5 kPM1-10 AV[kPM2.5-10] AV[kPM1-2.5] AV[kPM1-10] 0,2 0 1-01 29-01 26-02 26-03 23-04 21-05 18-06 16-07 13-08 10-09 8-10 5-11 3-12 31-12 t [d-m] Rys. 12. Przebieg oraz wartość średnia AV współczynników k modeli emisji cząstek stałych PM2.5 i PM1 na stacji nadzorowania jakości powietrza Brno–Lány Istnieje wyraźna regularność polegająca na tym, że w miesiącach zimnych współczynniki modeli imisji frakcji cząstek stałych PM2.5 i PM1 są większe niż w miesiącach ciepłych, co oznacza większy udział cząstek drobnych w miesiącach zimnych. Na rysunkach 13–15 przedstawiono charakterystykę statystyczną parametrów modeli imisji frakcji cząstek stałych. 1 Svatoplukova Zvonařka Lány kPM2.5-10 0,8 0,6 0,4 0,2 0 min max AV D W D D rr Rys. 13. Charakterystyka statystyczna współczynników modelu emisji cząstek stałych PM2.5: min – wartość minimalna, max – wartość maksymalna, AV – wartość średnia; D – odchylenie standardowe, W – współczynnik zmienności, D – rozstęp, r – stosunek rozstępu i wartości średniej 1 Svatoplukova Zvonařka Lány kPM1-10 0,8 0,6 0,4 0,2 0 min max AV D W D D rr Rys. 14. Charakterystyka statystyczna współczynników modelu (2) emisji cząstek stałych PM1: min – wartość minimalna, max – wartość maksymalna, AV – wartość średnia; D – odchylenie standardowe, W – współczynnik zmienności, D – rozstęp, r – stosunek rozstępu i wartości średniej 1 Svatoplukova Zvonařka Lány kPM1-2.5 0,8 0,6 0,4 0,2 0 min max AV D W D D rr Rys. 15. Charakterystyka statystyczna współczynników modelu (3) emisji cząstek stałych PM1: min – wartość minimalna, max – wartość maksymalna, AV – wartość średnia; D – odchylenie standardowe, W – współczynnik zmienności, D – rozstęp, r – stosunek rozstępu i wartości średniej Współczynnik zmienności oraz stosunek rozstępu i wartości średniej dla współczynników modeli jest znacznie mniejszy niż w wypadku zbiorów imisji. Współczynnik zmienności dla współczynników modeli jest rzędu (5 ÷ 20)%. Porównano wartości średnie parametrów modeli imisji frakcji cząstek stałych – rysunki 16–18. 1 0,831 AV[kPM2.5-10] 0,8 0,844 0,719 0,6 0,4 0,2 0 Svatoplukova Zvonařka Lány Rys. 16. Wartość średnia AV współczynników k modelu emisji cząstek stałych PM2.5 1 0,888 0,899 0,921 AV[kPM1-2.5] 0,8 0,6 0,4 0,2 0 Svatoplukova Zvonařka Lány Rys. 17. Wartość średnia AV współczynników k modelu (2) emisji cząstek stałych PM1 1 0,781 0,749 0,8 AV[kPM1-10] 0,642 0,6 0,4 0,2 0 Svatoplukova Zvonařka Lány Rys. 18. Wartość średnia AV współczynników k modelu (3) emisji cząstek stałych PM1 Wyznaczone wartości średnie współczynników modeli imisji cząstek stałych PM2.5 i PM1 mieszczą się w granicach spotykanych w literaturze [21, 22]. Wyniki identyfikacji modeli imisji cząstek stałych PM2.5 i PM1 nie mają jednoznacznej interpretacji, a można nawet powiedzieć, że są zastanawiające. Otóż wartości współczynników modeli dla stacji Zvonařka i Lány mają zbliżone wartości, szczególnie dla modeli imisji cząstek stałych PM2.5 oraz modelu (3) imisji cząstek stałych PM1. W wypadku modelu (2) imisji cząstek stałych PM1 różnica współczynnika modelu dla stacji Lány i Svatoplukova jest większa niż dla stacji Zvonařka i Lány. Należało oczekiwać, że zbliżone wartości współczynników modeli powinny być dla stacji Zvonařka i Svatoplukova, albo dla stacji Svatoplukova i Lány, co wynika z warunków usytuowania tych stacji, szczególnie ze względu na ruch drogowy na ulicach położonych w pobliżu. 4. Podsumowanie Pyły stanowią jedno z najpoważniejszych zagrożeń środowiska, szczególnie w centrach wielkich aglomeracji miejskich. Do oceny imisji poszczególnych frakcji pyłów wykorzystuje się wyniki badań empirycznych, prowadzonych na stacjach nadzorowania jakości powietrza, oraz wyniki modelowania imisji zanieczyszczeń. Do badania imisji poszczególnych frakcji pyłów wykorzystuje się modele emisji cząstek stałych PM10, zbudowane na zasadzie kryterium podobieństwa strukturalnego, i modele rozprzestrzeniania zanieczyszczeń oraz modele imisji cząstek stałych PM2.5 i PM1, zbudowane na zasadzie podobieństwa funkcjonalnego. Identyfikacja funkcjonalnych modeli imisji cząstek stałych PM2.5 i PM1, wykonana na podstawie wyników pomiarów imisji frakcji cząstek stałych PM10, PM2.5 i PM1 w 2010 r. na trzech stacjach nadzorowania jakości powietrza w Brnie o zróżnicowanym charakterze źródeł emisji zanieczyszczeń i ich rozprzestrzeniania, umożliwiła sformułowanie następujących wniosków: 1. Zauważa się silną korelację między zbiorami imisji poszczególnych frakcji cząstek stałych na wszystkich stacjach. 2. Występuje silna zależność imisji frakcji cząstek stałych od pór roku: znacznie większa jest imisja w zimnych porach roku. 3. Również zauważa się zależności imisji frakcji cząstek stałych od dni tygodnia, co wskazuje na silny wpływ na tę imisję czynników cywilizacyjnych. 4. Wyraźny jest wzajemny związek imisji poszczególnych frakcji, co uzasadnia przyjęcie modeli liniowych imisji cząstek stałych PM2.5 i PM1. 5. Istnieją duże różnice ekstremalnych wartości imisji poszczególnych frakcji cząstek stałych, co jest potwierdzone przez duże wartości współczynnika zmienności oraz stosunku rozstępu i wartości średniej. 6. Wyznaczone wartości średnie współczynników modeli imisji cząstek stałych PM2.5 i PM1 mieszczą się w granicach spotykanych w literaturze [21, 22]. 7. Istnieje wyraźna regularność polegająca na tym, że w miesiącach zimnych współczynniki modeli imisji frakcji cząstek stałych PM2.5 i PM1 są większe niż w miesiącach ciepłych, co oznacza większy udział cząstek drobnych w miesiącach zimnych. 8. Wyniki identyfikacji modeli imisji cząstek stałych PM2.5 i PM1 nie mają jednoznacznej interpretacji. Nie uzyskano wyników jednoznacznie wskazujących na wpływ ruchu drogowego na skład frakcji wymiarowych cząstek stałych. Niejednoznaczność uzyskanych wyników identyfikacji modeli imisji cząstek stałych PM2.5 i PM1 wskazuje na konieczność szerszego potraktowania rozpatrywanego zagadnienia. Jest uzasadnione oczekiwanie, że na podstawie większego zbioru wyników badań empirycznych, obejmujących również wyniki ze stacji położonych w innych obszarach, istnieje możliwość sformułowania bardziej jednoznacznych i ogólnych wniosków. Mimo częściowo krytycznej oceny wyników badań modeli imisji cząstek stałych PM2.5 i PM1 można stwierdzić, że modelowanie frakcji wymiarowych cząstek stałych zgodnie z kryterium podobieństwa funkcjonalnego jest praktycznie jedyną skuteczną metodą badania zagrożenia środowiska przez pyły. Piśmiennictwo 1. Canagaratna M.: Chase studies of particulate emissions from in–use New York City vehicles. Aerosol Science and Technology 2004; 38(6): 555–573. 2. Chłopek Z., Jakubowski A.: A study of the particulate matter emission from the braking systems of motor vehicles. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2009; 4: 45–52. 3. Chłopek Z., Jakubowski A.: The examination of the reduction of particulate matter emission from motor vehicle braking systems. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2010; 4: 29–36. 4. Chłopek Z., Piaseczny L.: Remarks about the modelling in science researches. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2001; 4: 47–57. 5. Chłopek Z., Skibiński F.: Wprowadzenie do tematyki emisji cząstek stałych PM2,5 z transportu samochodowego. Transport Samochodowy 2010; 3: 73–87. 6. Chłopek Z., Żegota M.: Badania emisji cząstek stałych PM10. Rozdział monografii „Edukacja ekologiczna. Podstawy działań naprawczych w środowisku”. Nałęczów: Polskie Towarzystwo Inżynierii Ekologicznej, 2004: 114–120. 7. Chłopek Z., Żegota M.: The emission of particulate matter PM10 from vehicles. Eksploatacja Niezawodnosc – Maintenance and Reliability 2004; 1: 3–13. 8. Chłopek Z.: Modele behawiorystyczne emisji cząstek stałych PM10 ze źródeł transportu drogowego. Zeszyty Naukowe Instytutu Pojazdów Politechniki Warszawskiej 2011; 1: 111–118. 9. Chłopek Z.: Ocena stanu zagrożenia środowiska przez cząstki stałe PM2,5 ze źródeł transportu drogowego. Zeszyty Naukowe Instytutu Pojazdów Politechniki Warszawskiej 2011; 1: 101–110. 10. Christoforou C. S.: Trends in fine particle concentration and chemical composition in Southern California. Journal of the Air & Waste Management Association 2000; 50(1): 43–53. 11. Eastwood P.: Particulate emissions from motor vehicles. Chichester: John Wiley, 2008. 12. Environment Australia: Emission estimation technique manual for aggregated emission from motor vehicles. Version 1.0. Canberra, 2000. 13. EPA: Addendum to Emission Factor Documentation for AP–42, Section 11.2.5 and 11.2.6 (now 13.2.1). Paved Roads. Final Report. September 1997. 14. EPA: Compilation of Air Pollutant Emission Factors AP–42, 5. Edition, Vol. I, Chapter 13, Section 13.2.1. Miscellaneous sources. October 1997. 15. EPA: Compilation of Air Pollution Emission Factors – AP 42, Emission Factor Documentation, Section 13.2.1, Paved Roads. Appendix C.1: Procedures for Sampling Surface Dust Loading and Appendix C.2: Procedures for Laboratory Analysis of Surface Loading Samples. 1993. 16. Forsberg B. et al: Comparative health impact assessment of local and regional particulate air pollutants in Scandinavia. Journal of the Human Environment 2005; 34(1).: 11–19. 17. Harrison R. M., Jones A. M., Lawrence R. G.: Major component composition of PM10 and PM2.5 from roadside and urban background sites. Atmos. Environ 2004; 38: 4531– 4538. 18. Hoek G. et al.: Association between mortality and indicators of traffic–related air pollution in the Netherlands: a cohort study. Lancet 2002; 360:1203–1209. 19. Houthujis D. et al.: PM10 and PM2.5 concentrations in central and eastern Europe: Results from the CESAR study. Atmospheric Environment 2001; 35: 2757–2771. 20. Kendall M.G.: A New measure of rank correlation. Biometrika 1938; 30: 81–89. 21. Keogh D. U., Ferreira L., Morawska L.: Development of a particle number and particle mass vehicle emissions inventory for an urban fleet. Environmental Modelling & Software 2009; Vol. 24, Issue 11: 2009 1323–1331. 22. Ketzel M. et. al.: Estimation and validation of PM2.5/PM10 exhaust and non–exhaust emission factors for practical street pollution modeling. Atmospheric Environment 2007; Vol. 41, Issue 40: 9370–9385. 23. Kinney P.L. et al.: Airborne concentration of PM2.5 and diesel exhaust particles on Harlem sidewalks. A community – based pilot study. Environmental Health Perspectives 2000; 108: 213÷218. 24. Kittelson D. B.: Engines and nanoparticles: A review. J. Aerosol Sci. 1998; Vol. 29, No. 5/6. Elsevier Science Ltd.: 575–588. 25. Kruskal H., Wallis W.A.: Use of ranks in one–criterion variance analysis. Journal of the American Statistical Association 1952; 47(260): 583–621. 26. Leksykon ekoinżynierii. Pod red. Gabriela Borowskiego. Warszawa: Polskie Towarzystwo Inżynierii Ekologicznej, 2010. 27. Lohmeyer A., Düring I.: Validierung von PM10–Immissionsberechnungen im Nahbereich von Straßen und Quantifizierung der Staubbildung von Straßen, Lützner Straße in Leipzig. Sächsisches Landesamt für Umwelt und Geologie. Dresden über Staatliche Umweltbetriebsgesellschaft. Radebeul: Februar 2001. 28. MacNee W., Donaldson K.: Exacerbations of COPD – environmental mechanism. Chest 2000; 117: 390–397. 29. Mayer A. et al.: Nanoparticle–emissions of EURO 4 and EURO 5 HDV compared to EURO 3 with and without DPF. SAE Emissions Measurement and Testing 2007: 335–43. 30. Pearson K.: On the theory of contingency and its relation to association and normal correlation. Drapers’ Company Research Memoirs. Biometric 1904; Ser. I. 31. Schwartz J., Laden F., Zanobetti A.: The concentration–response relation between PM2.5 and daily deaths. Environmental Health Perspectives 2002; Vol. 110, No. 10: 1025–1029. 32. Spearman Ch.: The proof and measurement of association between two things. Americal Journal of Psychology 1904; 15: 72–101. 33. Wasserman L.: All of nonparametric statistics. Springer, 2007. 34. Wilkinson P. et al.: Case–control study of hospital admission with asthma in children aged 5–14 years: relation with road traffic in north west London. Thorax 1999; 54: 1070–1074. 35. Yuh–Shen Wu et al.: The measurements of ambient particulates (TSP, PM2,5, PM2.5–10), chemical component concentration variation, and mutagenicity study during 1998–2001 in central Taiwan. Journal of Environmental Science and Health, Part C Environmental Carcinogenesis and Ecotoxicology Reviews 2002; Vol. 20, Issue 1: 45–59. Prof. dr hab. inż. Monika Hardygóra,1,2 mgr inż. Henryk Komander1, dr inż. Ryszard Błażej1, dr hab. inż. Leszek Jurdziak1, Metoda prognozowania trwałości zmęczeniowej złączy wieloprzekładkowych taśm przenośnikowych Streszczenie. Przedstawiono metodę badań rozkładu naprężeń w spoinie klejowej zakładkowych złączy taśm przenośnikowych wieloprzekładkowych. Określono wpływ właściwości wytrzymałościowych taśm i gumy klejowej na wielkość naprężeń i na trwałość zmęczeniową złączy. Podano zależności pozwalające na obliczenie wielkości naprężeń i prognozowanie trwałości złączy taśm przenośnikowych. Słowa kluczowe: przenośniki taśmowe, taśma przenośnikowa, złącza taśm przenośnikowych 1. Wstęp Wieloprzekładkowe taśmy przenośnikowe produkowane w odcinkach o określonej długości są łączone w dłuższe odcinki lub pętle według schematu pokazanego na rysunku 1. Z konstrukcji złącza wynika, że w przekrojach styków przekładek jest o jedną przekładkę mniej niż w łączonych taśmach. Strata wytrzymałości taśmy na obszarze złącza jest odwrotnie proporcjonalna do liczby przekładek taśmy. Rys. 1. Schemat wymiarowy złącza taśm 4 przekładkowych Fig. 1 Dimensional diagram of the 4 plies conveyor belt splice. Wyniki badań statycznej wytrzymałości złączy pokazują, że strata wytrzymałości jest większa od tej jaka wynika ze straty jednej przekładki. Jest to spowodowane działaniem naprężeń stycznych w spoinie klejowej, które są największe na stykach przekładek i wzmacniają działanie karbu w tych miejscach złącza. 1 INSTYTUT GÓRNICTWA Politechniki Wrocławskiej, Zakład Systemów Maszynowych, tel.71 320 68 60, E-,mail :[email protected], [email protected] [email protected]; [email protected]; 2 KGHM CUPRUM Ltd. – Centrum Badawczo-Rozwojowe, Wroclaw tel. tel. 071 781 22 01 W próbie statycznego rozciągania prawidłowo wykonanego złącza ulega ono zniszczeniu wskutek zerwania przekładek na styku pierwszego lub ostatniego stopnia. W praktyce eksploatacyjnej często dochodzi do rozklejania się styków złącza, które jest początkiem jego zniszczenia. Wynika z tego, iż pod wpływem obciążeń zmęczeniowych wcześniejszemu zniszczeniu ulega spoina klejowa [1,2,3]. Zagadnienia związane z ustaleniem jakie właściwości taśm przenośnikowych i materiałów wulkanizacyjnych mają zasadniczy wpływ na wielkość naprężeń w spoinie klejowej oraz na ich trwałość zmęczeniową były przedmiotem prac prowadzonych w Laboratorium Transportu Taśmowego Instytutu Górnictwa Politechniki Wrocławskiej [4,5], których wyniki przedstawiono poniżej. 2. Badania naprężeń w spoinie klejowej złącza taśm przenośnikowych. Wielkość naprężeń w spoinie klejowej złącza badano poprzez pomiary kąta odkształcenia postaciowego spoiny γ. Kąt γ definiowany jest jako iloraz bezwzględnego odkształcenia postaciowego ∆S i odległości g pomiędzy przemieszczającymi się krawędziami spoiny pod wpływem działania siły rozciągającej F (rys.1). tgγ = ∆S g (1) Z uwagi na właściwości przekładek tkaninowych i gumy spoiny klejowej kąt odkształcenia postaciowego γ nie jest jednakowy na całej długości poszczególnych stopni złącza (rys.2) w związku z czym jego pomiar musi się odbywać w wielu punktach na długości spoiny. F F γ max l st Rys. 2. Schemat odkształceń spoiny klejowej na długości stopnia lst złącza Fig. 2 Strain diagram of the adhesive joint along the length of the 1st step of a belt splice Pomiarów kątów γ dokonywano na pełnowymiarowych złączach taśm 4-przekładkowych. Aby uniezależnić wyniki badań od wpływu ewentualnych wad wykonawczych złącza wykonywano w ten sposób, że poszczególne stopnie uzyskiwano poprzez przecinanie przykładek taśmy w określonych miejscach odpowiadających długości stopni. Próbki do badań o szerokości 50 mm wycinano wzdłuż osi taśmy i obciążano je siłą wywołująca naprężenia o wielkości 15% wytrzymałości taśmy. Na podstawie zmierzonych kątów γ sporządzano wykresy zależności γ=f(lx), gdzie lx to odległość punktu pomiarowego od miejsca styku przeciętych przekładek. Przykładowe wyniki pomiarów kąta γ po aproksymacji liniami trendu pokazano na rysunku 3. 70 EP 1000/4 60 PP 1000/4 Angle γ [degrees] 50 40 30 20 10 0 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 Splice length [mm] Rys. 3. Rozkład kątów odkształcenia postaciowego γ w spoinie klejowej na długości złączy EP1000/4 i PP1000/4 Fig. 3 Distribution of non-dilatational strain angles γ in adhesive joint along the length of belt EP1000/4 and PP1000/4 splices Wykresy zależności γ=f(lx) aproksymowano liniami trendu i obliczano wydłużenia względne spoiny ε z zależności (2): ε = 1 − cos γ − ε t ⋅ν cos γ + ν ⋅ sin γ (2) gdzie: εt – wydłużenie względne taśmy przy naprężeniu 15% wytrzymałości taśmy, ν - współczynnik Poissona gumy klejowej. Powyższą zależność określono rozpatrując schemat odkształceń spoiny klejowej przedstawiony na rys 4, przyjmując, że wydłużenie względne spoiny ε=(g1-g)/g i uwzględniając wpływ wydłużenia taśmy oraz współczynnika Poissona gumy klejowej. F g g1 γ F Rys. 4. Schemat odkształceń wymiarowych elementu spoiny klejowej złącza. Fig. 4 Dimensional deformations diagram of adhesive joint component of belt splice. W wyniku obliczeń uzyskiwano wykresy rozkładu wydłużeń spoiny klejowej na długości złącza ε=f(lx). Aby uzyskać obraz rozkładu naprężeń w spoinie klejowej na długości złącza przeprowadzono badania gumy klejowej celem uzyskania zależności τ=f(ε), co pozwoliło na przeliczenie uzyskanych wyników na zależność τ=f(lx). Przykładowe wyniki obliczeń rozkładu naprężeń w spoinie złączy EP1000/4 i PP1000/4 pokazano na rysunku 5. 2 EP 1000/4 PP 1000/4 Strain [MPa] 1,5 1 0,5 0 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 Splice length [mm] Rys. 5. Rozkład naprężeń w spoinie klejowej na długości złączy EP1000/4 i PP1000/4 Fig. 5 Stress distribution in the adhesive joint along the length of splices EP1000 / 4 and PP1000 / 4 Wyniki pomiarów podstawowych właściwości łączonych taśm, właściwości gumy klejowej oraz wyniki badań wielkości naprężeń na zewnętrznych stykach złączy przedstawiono w tabeli 1. Wytrzymałość taśmy na rozciąganie Rt oraz wydłużeń εt oznaczono według EN ISO 283:2008, moduł sprężystości taśmy Mt według EN ISO 9856:2005, wytrzymałość na rozciąganie gumy klejowej według ISO 252:2007, moduł gumy klejowej wg ISO 252:2007 przy wydłużeniu równym 100% osiągniętym po 3 godzinach obciążania próbek ze stała siłą. Tabela 1. Zestawienie wyników badań właściwości wytrzymałościowych taśm oraz naprężeń w spoinie klejowej złączy. Właściwości gumy klejowej Właściwości łączonych taśm Lp 1 2 3 4 5 6 7 8 Rodzaj materiału przekładek PP PP EP EP PP EP EP PP Moduł Wytrzymałość sprężystości na rozciąganie taśmy Rt [kN/m] Mt [kN/m] 1356 1258 1050 997 2359 1771 1942 1752 4639 4375 12665 14717 11030 26577 18683 8697 Wydłużenie Wytrzymałość względne Moduł gumy na rozciaganie przy 15% Rt Mg [MPa] TS [Mpa] εt [%] 6,28 5,45 1,11 1,07 4,27 1,13 1,97 6,17 6,2 10,9 12,4 15,2 7,0 6,2 7,2 9,6 1,3 1,7 0,8 1,2 1,4 1,3 0,8 0,9 Maksymalne naprężenia w spoinie klejowej τ [MPa] 1,70 1,65 0,43 0,79 1,95 0,74 0,62 1,00 W przeprowadzonych badaniach zarówno kąty γ spoiny klejowej jak i moduły gumy spoiny oznaczano po 3 godzinach od momentu ich naprężenia z uwagi na duży wpływ zjawiska pełzania tych materiałów na ich wydłużalność w początkowym okresie działania sił rozciągających. Wyniki uzyskanych badań zestawionych w tabeli 1 poddano analizie mającej na celu przedstawienie zależności naprężeń τ w funkcji zmiennych Mt/Rt i Mg. Założono następującą postać tej zależności (3) : M τ = C( t )x M g y (3) Rt gdzie: C – stała, x, y – poszukiwane wykładniki potęg. Powyższą funkcję sprowadzono do liniowej logarytmując obydwie strony równania przez co można było zastosować regresję liniową. Poszukiwane wielkości C, x i y wyznaczono w programie STATGRAPHICS Centurion XV (v.15.2.06, edition Professional) firmy StatPoint Inc., którego licencję akademicką posiada Instytut Górnictwa Politechniki Wrocławskiej. W wyniku przeprowadzonych analiz regresji liniowej wielorakiej z transformacją logarytmiczną dobrano następującą postać poszukiwanej zależności: Rt 0,525 0.99 ) Mg (4) Mt Współczynnik korelacji tej funkcji w odniesieniu do wyników badań wyniósł 2 R =90,6% , a jego skorygowana wartość 86.9%. Z uwagi na to, że uzyskana wartość P w tabeli analizy wariancji ANOVA była mniejsza niż 0.05 (P-value=0.0027), stwierdzono statystyczną zależność pomiędzy zmiennymi na poziomie ufności 95.0%. Wykres funkcji (4) przedstawiono na rysunku 6. τ = 2,434( 12 11 10 9 8 7 Tał 6 5 4 3 2 5 4.4 1 3.8 3.2 0 2.6 Rt/Mt 2 1.4 0.8 0.2 0.8 0.6 1 1.2 1.4 1.6 1.8 2 Mg Rys 6. Zależność zmiennej zależnej τ od zmiennych niezależnych Mt/Rt i Mg (4). Fig. 6. The relationship of the dependent variable τ from independent variables Mt/Rt and Mg (4). Przeprowadzone badania wykazały, ze właściwości wytrzymałościowe łączonych taśm i gumy klejowej mają istotny wpływ na wielkość naprężeń w spoinie klejowej. Maksymalne naprężenia występujące na zewnętrznych stykach przekładek złącza w zależności od modułu sprężystości taśmy i modułu gumy mogą różnić się nawet kilkakrotnie jak to widać na przykładzie wykresów przedstawionych na rysunku 5, dla złączy taśm EP100/4 i PP1000/4 Przybliżoną wartość maksymalnych naprężeń w spoinie klejowej można oszacować formułą (5): R τ = 2,4 t M g (5) Mt 3. Badania trwałości zmęczeniowej złączy W spoinie klejowej złącza poddanego cyklicznym obciążeniom zmęczeniowym kumuluje się ciepło w wyniku czego wzrasta jej temperatura. Wzrost temperatury zależy od wielkości i częstotliwości obciążeń oraz od rodzaju gumy spoiny i właściwości przekładek tkaninowych taśmy. Nawet przy takich samych warunkach obciążeniowych różnica temperatur poszczególnych złączy może dochodzić do kilkudziesięciu stopni. W takim wypadku porównywanie wyników badań pomiędzy poszczególnymi rodzajami złączy nie jest możliwe. W przeprowadzonych badaniach zmęczeniowych założono, że temperatura spoiny klejowej badanych złączy nie może przekraczać zakresu 23oC ±2oC. Eksperymentalnie ustalono, że powyższy zakresu temperatur jest możliwy do uzyskania przy następujących warunkach badań: - Zakres obciążeń 5% do 20% wytrzymałości taśmy - Częstotliwość obciążeń 0,3 Hz - Sinusoidalna charakterystyka obciążeń - Temperatura otoczenia 18oC Temperaturę spoiny kontrolowano przy pomocy pirometru. Badania wykonywano na próbkach o kształcie wiosełka o szerokości 50mm w części badawczej i szerokości 100 mm w części uchwytowej. Długość całkowita próbek wynosiła 1150mm. Na podstawie analizy wyników badań wielkości naprężeń w spoinach klejowych stopnie środkowe złączy skrócono do 150mm. Stopnie zewnętrzne miały długość standardową. Badania wykonywano na maszynie do badań dynamicznych firmy Zwick-Amster typu HC-25. Jako kryterium oceny trwałości zmęczeniowej złączy przyjęto liczbę cykli zmęczeniowych powodujących rozwarstwienie spoiny o wielkości równej 3mm. Podstawowe parametry badanych złączy oraz uzyskane wyniki liczby cykli zmęczeniowych podano w tabeli 2. W sumie poddano badaniom 12 próbek złączy. Podana w tabeli liczba cykli zmęczeniowych to średnia z dwóch badanych próbek. W tabeli 2 zamieszczono też właściwości złączy, które na podstawie przeprowadzonej analizy miały istotny wpływ na trwałość zmęczeniową. Wytrzymałość adhezyjną T między łączonymi przekładkami oznaczono według EN ISO 252:2008. , Tabela 2. Zestawienie wyników badań zmęczeniowych złączy. Lp 1 2 3 4 5 6 Rodzaj Wytrzymałość na materiału rozciąganie taśmy przekładek Rt [kN/m] PP EP EP EP EP PP Moduł sprężystości taśmy Mt [kN/m] Wytrzymałość adhezyjna spoiny klejowej T [N/mm] Liczba cykli zmęczeniowych po których rozpoczyna się rozwarstwianie złącza LC 4639 12665 14717 26577 18683 8697 6,3 11,4 14,6 9,2 7,5 12,1 440 212 000 359 000 145 000 8600 1900 1356 1050 997 1771 1942 1752 Wyniki uzyskane z badań zmęczeniowych zestawionych w tabeli 2 poddano analizie mającej na celu przedstawienie zależności liczby cykli w funkcji zmiennych Rt, Mt, Mg i T. Założono, że liczba cykli może zależeć od 3 zmiennych: Rt/Mt, Mg i T i w programie STATGRAPHICS zastosowano dla nich procedurę selekcji modelu regresji wielorakiej. Okazało się, że najlepszym pod względem skorygowanej wartości R2 jest model oparty na dwóch zmiennych Rt/Mt, i T (R2=95.04%, a jego wartość skorygowana 91.74%). W wyniku przeprowadzonej analizy nieliniowej regresji wielorakiej uzyskano następującą zależność na liczbę cykli LC : LC = 12,3362( M t 1,90881 1.92878 ) T Rt (6) Uzyskane wskaźniki korelacji dla zależności (6) wynoszą: R2=94,7% i jego wartość skorygowana 91,2%. Na rys. 7 przedstawiono zależność opisaną równaniem (6). Tysiące 600 550 500 450 400 350 300 LC 250 200 150 100 50 3 4 5 6 7 Mt/Rt 8 16 15 14 13 12 11 10 9 0 3 4 5 6 7 8 9 10 11 12 13 14 15 16 T Rys 7. Zależność liczby cykli LC od zmiennych niezależnych Mt/Rt i T (6). Fig. 7 Dependence of the number of cycles LC of the independent Mt/Rt and T(6). Do celów praktycznych zależność (6) można uprościć do postaci (7): LC = 8,05( Mt 2 2 ) T Rt (7) Uzyskane wskaźniki korelacji dla zależności (7) wynoszą: R2=94,6% i jego wartość skorygowana 94,6%. 4. Podsumowanie Przedstawione wyniki badań pokazały jak kształtują się naprężenia w spoinie klejowej na długości poszczególnych stopni złączy wieloprzekładkowych taśm przenośnikowych. Stwierdzono, że ekstremalne wielkości tych naprężeń występujące na stykach złącza zależą głównie od modułu sprężystości łączonej taśmy, jej wytrzymałości oraz od modułu gumy klejowej. Na podstawie otrzymanej zależności (4 lub 5) producenci taśm przenośnikowych i materiałów do ich łączenia mogą tak dobrać ich parametry, aby przy obciążeniu taśmy wynoszącym 15% jej wytrzymałości maksymalne naprężenia w spoinie klejowej były rzędu 0,5 MPa. Badania zmęczeniowe złączy wykazały, że istotny wpływ na ich trwałość ma jednostkowy moduł sprężystości łączonych taśm Rt/Mt, oraz wytrzymałość adhezyjna T gumy klejowej do łączonych przekładek. Na podstawie badań określono zależność (6, 7) przy pomocy której można obliczyć trwałość zmęczeniową złącza mierzoną liczbą cykli zmęczeniowych powodujących początek rozwarstwiania się styków złącza pod wpływem cyklicznych obciążeń w zakresie od 5% do 20% wytrzymałości taśmy. Stwierdzono, że w zależności od właściwości wytrzymałościowych łączonych materiałów graniczna liczba cykli zmęczeniowych może wynosić od kilkaset do kilkaset tysięcy, co między innymi wyjaśnia przyczynę rozklejania się wielu złączy w praktyce eksploatacyjnej. Uzyskane wzory na liczbę cykli zmęczeniowych LC pozwalają w prosty sposób tak dobrać właściwości taśm i materiałów do ich łączenia aby uzyskać wysoką trwałość zmęczeniową złączy a przez to ich większą niezawodność. Przeprowadzone badania wskazują na istotne parametry wpływające na wytrzymałość i trwałość złączy taśm przekładkowych . Wydaje się celowe aby opracowana metodą testować kolejne złącza o zróżnicowanych właściwościach i budowie. Pozwoli to na ewentualne uściślenie wzorów przedstawionych w artykule i bardziej precyzyjne prognozowanie trwałości zmęczeniowej złączy. Prace badawcze finansowane ze środków Narodowego Centrum Badań i Rozwoju w latach 2009-2012 jako projekt rozwojowy Nr R09 001906 Literatura [1] Błażej Ryszard., Hardygóra Monika. Modelling of shear stresses in multiply belt splices. Bulk Solids Handling. 2003, 23/4, 234-241. [2] Kozhushko G.G., Kopnov V.A. Fatigue strength functions in shear loading of fabric conveyor belts. International Journal of Fatigue 1995, 17/8, 539-544 [3] Mazurkiewicz Dariusz “Problems of numerical simulation of stress and strain in the area of the adhesive-bonded joint of a conveyor belt”, Archives of Civil and Mechanical Engineering, 2009, IX/2, 75-91. [4] Projekt badawczy nr 4 T 12A 06630 pt: Opracowanie zasad doboru parametrów wytrzymałościowych taśm przenośnikowych i materiałów do ich łączenia celem uzyskania optymalnej wytrzymałości i trwałości połączeń. Instytut Górnictwa Politechniki Wrocławskiej, Wrocław 2009r. [5] Projekt badawczy nr N R09 0019 06 pt: Przenośnik taśmowy o zwiększonej efektywności ekonomicznej i energetycznej zbudowany i eksploatowany wg zasad zrównoważonego rozwoju. Instytut Górnictwa Politechniki Wrocławskiej, Wrocław 2012r. dr inŜ. Krzysztof Parczewski dr inŜ. Henryk Wnęk Katedra Silników Spalinowych i Pojazdów Akademia Techniczno-Humanistyczna Ul. Willowa 2, 43-300 Bielsko-Biała, Polska e-mail: [email protected], [email protected] Wykorzystanie przyczepności podczas hamowania pojazdu Słowa kluczowe: przyczepność, badania pojazdu, hamowanie pojazdu, tor krzywoliniowy, poślizg kół, ABS Streszczenie. W publikacji przedstawiono zagadnienie wykorzystania przyczepności opony do nawierzchni jezdni podczas hamowania. Wyniki prezentowane w publikacji oparto na badaniach drogowych pojazdu osobowego wyposaŜonego w układ zapobiegający blokowaniu kół. Przeprowadzono dwa rodzaje prób stosowanych do badania wykorzystania przyczepności - badania pojazdu wykonującego manewr hamowania na prostoliniowym odcinku drogi oraz na łuku drogi. Na podstawie badań określono siły hamowania oraz współczynniki przyczepności dla poszczególnych kół pojazdu, w tym wartości graniczne współczynnika przyczepności. 1. Wstęp Podczas ruchu pojazdu za przekazywanie sił z pojazdu na podłoŜe odpowiedzialne są opony oraz stan nawierzchni drogi. Wartości sił przenoszonych na nawierzchnię drogi zaleŜą od parametrów pojazdu i jego ruchu. Są one ograniczone przyczepnością na styku opona jezdnia. Zjawisko przyczepności zachodzi na styku koła z jezdnią i obejmuje wszystkie warunki oraz mechanizmy występujące podczas tej współpracy. Współczesne pojazdy są wyposaŜone w układy zapobiegające blokowaniu kół podczas hamowania, co ogranicza zakres zmian pracy układu hamulcowego do obszaru ograniczonych poślizgów. W pracy przedstawiono zagadnienia współpracy koła z jezdnią podczas hamowania pojazdów na drodze prostoliniowej i na łuku drogi, wyznaczone graniczne wartości współczynników przyczepności oraz przebiegi tych współczynników uzyskane z pomiarów. 2. Wykorzystanie przyczepności Podczas ruchu pojazdu mechanizm powstawania sił na styku opony z jezdnią wynika ze stycznych i normalnych napręŜeń w obszarze tego styku. Naciski na jezdnię są róŜne w kaŜdym miejscu styku z oponą i zmieniają się zarówno w kierunku wzdłuŜnym, jak i poprzecznym. Podczas toczenia dochodzi jeszcze problem niesymetryczności śladu współpracy. KaŜdy element opony, znajdujący się w obszarze kontaktu, jest odpowiedzialny za przenoszenie sił wzdłuŜnych i poprzecznych przez oponę. Są dwa główne mechanizmy [1, 2, 3, 4, 5, 6, 10] odpowiedzialne za powstawanie sił tarcia między oponą a jezdnią: tarcie odkształceniowe i adhezja. Na powierzchni przylegania (adhezja) powstaje siła w wyniku wiązania międzycząsteczkowego między gumą bieŜnika i powierzchnią drobiny asfaltu. Oddziaływanie to zmniejsza się wraz z obecnością zanieczyszczeń lub wody w miejscu styku. Mechanizm tarcia odkształceniowego powstaje w wyniku straty energii podczas odkształcenia gumy na drobinie asfaltu. Tarcie powstające w następstwie tego mechanizmu nie jest „tak czułe” na zanieczyszczenia i obecność wody. Rys. 1. Mechanizmy odpowiedzialne za powstawanie sił tarcia między oponą a nawierzchnią jezdni Ruch samochodu moŜe przebiegać zgodnie z osią wzdłuŜną pojazdu oraz w kierunku prostopadłym do tej osi. Przyczepność między oponą a jezdnią moŜe być opisana za pomocą współczynnika przyczepności (stosunku sił przyczepności do siły nacisku koła na nawierzchnię drogi) [2]. Współczynnik przyczepności rozumiany jest, jako stosunek maksymalnej stycznej siły wypadkowej przenoszonej przez koło do siły nacisku działającej na to koło. W trakcie przeprowadzonych analiz wyznaczono chwilowe współczynniki przyczepności wykorzystując zaleŜność: W µ= FZ Przyczepność obejmuje łącznie siłę przenoszoną na powierzchnię drogi w kierunkach obwodowym XK i bocznym YK. Siła wypadkowa W jest ograniczona przyczepnością koła do nawierzchni Fµ. W = X K2 + YK2 oraz W ≤ Fµ Rozpatrując poszczególne koła oddzielnie, moŜna wyznaczyć graniczne wartości sił przyczepności, które mogą być przez nie przenoszone. Podczas ruchu pojazdu na prostym, poziomym odcinku drogi powyŜsza zaleŜność uprości się ze względu na moŜliwość pominięcia siły poprzecznej. W takim przypadku cała siła przyczepności moŜe być wykorzystana na hamowanie. W przypadku ruchu pojazdu na łuku drogi, wpływ siły bocznej jest tym większy im mniejszy jest jego promień. W przypadku, gdy na koło działa jednocześnie siła wzdłuŜna i poprzeczna z pewnym uproszczeniem moŜna zapisać zaleŜność określającą współczynnik przyczepności [9]. µ = µ x2 + µ y2 gdzie: µx – współczynnik przyczepności wzdłuŜnej, µy – współczynnik przyczepności poprzecznej. Podczas hamowania na łuku drogi na pojazd działa siła dośrodkowa zaleŜna od v2 prędkości jazdy FR = m ⋅ , która ogranicza, moŜliwą do wykorzystania, siłę przyczepności R w kierunku wzdłuŜnym. Z tego względu tylko część siły przyczepności moŜe być wykorzystana na hamowanie pojazdu. Analizując ruch pojazdu poruszającego się po torze kołowym moŜna wyznaczyć, jaką część współczynnika przyczepności moŜna wykorzystać na hamowanie. v2 µ h = µ m − g ⋅ R gdzie: µh – część współczynnika przyczepności wykorzystywana na hamowanie pojazdu, µm – współczynnik przyczepności przylgowej (maksymalna wartość względnej siły przyczepności, jaka moŜe być uzyskana w danych warunkach), v – prędkość jazdy, R – promień skrętu, g – przyspieszenie ziemskie. 2 Podczas manewrów rozpędzania czy jazdy ze stałą prędkością (po drodze płaskiej) stosunkowo rzadko wykorzystywana jest cała przyczepność kół. W przypadkach manewrów hamowania na drodze prostoliniowej lub na łuku drogi znacznie częściej dochodzi do pełnego wykorzystania przyczepności, szczególnie podczas wykonywania manewrów na drodze mokrej, pokrytej śniegiem lub oblodzonej. Zastosowanie mechanizmów zapobiegających blokowaniu kół podczas hamowania (ABS) ogranicza poślizgi kół. Będzie to powodowało ograniczenie sił hamujących działających na poszczególne koła pojazdu. PoniŜej przedstawiono badania wykorzystania przyczepności pojazdu podczas hamowania. 3. Badania przyczepności pojazdu 3.1. ZałoŜenia do badań Przeprowadzono dwa rodzaje prób stosowanych do badania wykorzystania przyczepności - badania pojazdu wykonującego manewr hamowania na prostoliniowym odcinku drogi oraz na łuku drogi. a) b) R=2 0m ∗ ∗- początek manewru hamowania Rys. 2. Tory poszczególnych prób: a) hamowanie na prostoliniowym odcinku drogi, b) hamowanie na łuku drogi Ze względów bezpieczeństwa badania wykonywano na suchej i czystej nawierzchni asfaltowej. Do badań wykorzystano czujniki ciśnienia zamontowane w układzie hamulcowym, głowicę do pomiaru prędkości wzdłuŜnej i poprzecznej, czujnik siły na pedale hamulca, czujniki pozwalające na pomiar przyspieszeń pojazdu w kierunkach X oraz Y, czujniki do pomiaru prędkości kątowych ruchu nadwozia samochodu oraz czujniki do pomiaru kąta obrotu i momentu na kole kierownicy [7, 8]. Masa pojazdu wynikała z jego masy własnej, masy aparatury pomiarowej i masy kierowcy. W trakcie hamowania następuje zmiana sił nacisku poszczególnych kół na jezdnię. Powoduje to zmianę granicznych sił przyczepności i w efekcie zastosowania układu ABS, ograniczenie sił hamujących generowanych przez hamulce poszczególnych kół. ObciąŜenia poszczególnych kół pojazdu wyznaczano na podstawie pomiarów połoŜenia środka masy, oraz sił wzdłuŜnych i poprzecznych wynikających z warunków ruchu. Przy wyznaczaniu sił nacisku na poszczególne koła nie uwzględniono zmiany połoŜenia środka masy wynikającej z pochylenia pojazdu. Na podstawie zmierzonych ciśnień w układzie hamulcowym oraz parametrów geometrycznych hamulców wyznaczono siły hamujące dla poszczególnych kół. Uwzględniono wpływ bezwładności kół na siły hamujące wywołane zmianą ich prędkości obrotowych. 3.2. Próba hamowania na prostoliniowym odcinku drogi Pierwsza próba była przeprowadzana na prostoliniowym odcinku drogi. Kierowca utrzymywał prostoliniowy kierunek jazdy. Po uzyskaniu odpowiedniej prędkości naciskał na pedał hamulca. Siła nacisku na pedał zapewniała działanie układu zapobiegającego blokowaniu kół. Rys. 3. Przebieg prędkości jazdy pojazdu podczas próby hamowania na prostoliniowym odcinku drogi Rys. 4. Przebieg obciąŜeń kół pojazdu podczas próby hamowania na prostoliniowym odcinku drogi Rys. 5. Przebieg sił hamujących działających na poszczególne koła pojazdu Na wykresach przedstawiono widoczne dociąŜenie kół przednich podczas hamowania oraz wyraźną róŜnicę wielkości sił hamujących osi przedniej i tylnej. Pewne przesunięcia działania hamulców kół prawych i lewych wynikają z niejednorodności podłoŜa oraz pewnej niesymetrii obciąŜenia pojazdu. 3.3. Próba hamowania na łuku drogi Druga próba była przeprowadzana na łukowym odcinku drogi. Kierowca utrzymywał kierownicę w taki sposób by pojazd poruszał się po torze kołowym. Po pokonaniu drogi około 15 m na torze kołowym, kierowca naciskał na pedał hamulca. Siła nacisku na pedał zapewniała działanie układu ABS. Na rysunku 6 przedstawiono tor ruchu pojazdu uzyskany na podstawie wyników badań. Na następnych wykresach kolejno pokazano przebiegi zmian obciąŜeń kół pojazdu (rys. 7) oraz sił hamujących działających na poszczególne koła pojazdu (rys. 8). Rys. 6. Tor ruchu pojazdu podczas próby hamowania na łuku drogi Rys. 7. Przebieg obciąŜeń kół pojazdu podczas próby hamowania na łuku drogi Rys. 8. Przebieg sił hamujących działających na poszczególne koła pojazdu Na rysunkach przedstawiono widoczne zmiany obciąŜeń kół prawych i lewych podczas hamowania na torze krzywoliniowym oraz wyraźną róŜnicę wielkości sił hamujących osi przedniej i tylnej, skorygowane ze względu na rozkład nacisków oraz siłę dośrodkową działającą na pojazd. MoŜna zauwaŜyć, Ŝe w początkowej fazie hamowania nacisk koła tylnego lewego jest bliski zeru, co skutkuje ograniczeniem ciśnienia przez układ zapobiegający blokowaniu kół w obwodach hamulców kół tylnych i tym samym spadkiem sił hamowania do niewielkich wartości. 4. Analiza pomiarów pod kątem wykorzystania przyczepności kół Na podstawie przedstawionych powyŜej analiz oraz wyników badań wyznaczono wykorzystanie przyczepności kół pojazdu podczas prób drogowych hamowania na prostoliniowym odcinku oraz podczas hamowania na łuku drogi. Na rysunkach 9 i 10 przedstawiono wartości wyznaczonych, wykorzystywanych współczynników przyczepności oraz graniczne wartości tych współczynników wynikające z warunków ruchu. Rys. 9. Współczynniki przyczepności podczas hamowania na prostoliniowym odcinku drogi MoŜna zauwaŜyć, Ŝe w pierwszym przypadku maksymalna wartość wykorzystanego współczynnika przyczepności dla kół przednich oscyluje wokół wartości 0,75, a dla kół tylnych jest większa i oscyluje wokół wartości 0,8. Rys. 10. Współczynniki przyczepności podczas hamowania na łuku drogi W przypadku hamowania na łuku drogi wielkości wykorzystanego współczynnika przyczepności wzrastają od początku hamowania do wartości maksymalnej wraz ze zmniejszaniem się prędkości jazdy. Dla przedniego prawego koła (dociąŜanego) wartości współczynnika przyczepności są większe niŜ dla lewego. Po zmniejszeniu prędkości jazdy wartości współczynników się stabilizują. Współczynniki przyczepności kół tylnych są wyraźnie mniejsze w początkowym stadium hamowania, po czym wzrastają do wartości maksymalnych. RóŜnice między wartościami współczynników dla kół przednich, wynikają z niedokładności ich oszacowania spowodowanego pominięciem wpływu pochylenia bocznego samochodu oraz ze znacznej róŜnicy obciąŜeń prawej i lewej strony pojazdu. Badanie hamowania pojazdu na łuku drogi pozwoliło na wyznaczenie granicznego sumarycznego współczynnika przyczepności (rys. 11) wyznaczonego na podstawie elipsy przyczepności. Rys. 11. Graniczne wartości współczynników przyczepności uzyskane podczas próby hamowania na łuku drogi (linią grubą oznaczono elipsę przyczepności) 5. Podsumowanie i wnioski Badania przyczepności kół do nawierzchni jezdni pokazały, Ŝe podczas awaryjnego hamowania na drodze prostoliniowej przyczepność jest wykorzystana w pełni od początkowej chwili hamowania, aŜ do zatrzymania pojazdu. DociąŜenie osi przedniej i odciąŜenie osi tylnej skutkuje wyraźnym zróŜnicowaniem ciśnień w obwodach hamulcowych, co pozwala na wykorzystanie przyczepności kół. Niewielkie róŜnicę pomiędzy poszczególnymi kołami jednej osi, wynikają z miejscowych warunków przyczepności i są wywołane przez drobne nierówności i zabrudzenia. W przypadku hamowania na łuku drogi ograniczenie przyczepności wzdłuŜnej wynika z występowania siły dośrodkowej. Układ ABS, zapobiegający blokowaniu kół podczas hamowania, nie pozwala na osiąganie duŜych sił wzdłuŜnych, zapewniając odpowiednie warunki na rozwijanie sił poprzecznych i tym samym zachowanie stateczności ruchu pojazdu. Wraz ze zmniejszaniem się prędkości jazdy korekta wynikająca z ruchu po łuku drogi jest coraz mniejsza. Wyraźne zróŜnicowanie nacisków poszczególnych kół, szczególnie stronami - prawą i lewą, wynika z działania siły dośrodkowej. Niesymetria wywołana przez obciąŜenie pojazdu tylko kierowcą dodatkowo wpływa na wielkości poszczególnych nacisków i rozwijanych sił hamujących. Podobnie jak podczas hamowania na prostym odcinku drogi, występują znaczne róŜnice ciśnień w obwodach hamulcowych kół przednich i tylnych. Przyczepność kół przednich jest wykorzystana w pełni, natomiast w przypadku kół tylnych, z których tylne lewe koło chwilowo traci kontakt z jezdnią a tylne prawe jest częściowo obciąŜone, pełne wykorzystanie przyczepności następuje dopiero przy niŜszych prędkościach jazdy. Po przekroczeniu granicznej prędkości, przy której układ ABS się wyłącza, następuje wyraźny wzrost siły hamującej na kołach tylnych, co powoduje znaczne zwiększenie sił hamujących. Stan ten pokazano na rysunku 11, na którym równieŜ graniczne wartości współczynnika przyczepności wyznaczone z elipsy przyczepności i uzyskane z badań drogowych. 6. Literatura 1. Andrzejewski R. Dynamika pneumatycznego koła jezdnego. Warszawa: WNT 2010. 2. Arczyński St. Mechanika ruchu samochodu. Warszawa: WNT 1993. 3. Fundowicz P. Droga hamowania na łuku drogi. Zeszyty Instytutu Pojazdów, 1(77)/2010, Politechnika Warszawska, 2010; s. 103-110. 4. Gillespie T. D. Fundamentals of vehicle dynamics. Warrendale: SAE Inc. 1992. 5. GrzegoŜek W. Modelowanie dynamiki samochodu przy stabilizującym sterowaniu siłami hamowania. Kraków: Zeszyty Naukowe Politechniki Krakowskiej, Seria Mechanika, monografia 275, 2000. 6. Pacejka Hans B. Tire and vehicle dynamics. Warrendale: SAE 2006. 7. Parczewski K. Wnęk H., Wpływ niesprawności zawieszeń na stateczność ruchu pojazdu - porównanie badań symulacyjnych i pomiarów. Archiwum Motoryzacji, 2006; nr 2, s.159-169. 8. Parczewski K., Wnęk H. Wykorzystanie modelu samochodu do analizy ruchu pojazdu po torze krzywoliniowym. Lublin: Eksploatacja i Niezawodność 2010; nr 4(48), s. 37-46. 9. Prochowski L., Unarski J., Wach W., Wicher J. Podstawy rekonstrukcji wypadków drogowych. Warszawa: WKŁ 2008. 10. Smith R.H. Analyzing friction in the design of rubber products and their paired surfaces. CRC Press 2008. Dr inż. Daniel Pieniak Faculty of Transport and Computer Science University of Economics and Innovation in Lublin Mełgiewska 7-9 str., 20-209 Lublin e-mail: [email protected] Dr n. med. Agata M. Niewczas Department of Conservative Dentistry Medical University of Lublin, Karmelicka 7 str., 20-081 Lublin, e-mail: [email protected] Dr inż. Paweł Kordos Faculty of Transport and Computer Science University of Economics and Innovation in Lublin Mełgiewska 7-9 str., 20-209 Lublin e-mail: [email protected] Wpływ zmęczenia cieplnego oraz starzenia na mikrotwardość kompozytów polimerowo – ceramicznych do zastosowań biomedycznych Słowa kluczowe: zmęczenie cieplne, starzenie, mikrotwardość, kompozytowe wypełnienia stomatologiczne Streszczenie: Badania prezentowane w niniejszej publikacji dotyczyły kompozytów polimerowoceramicznych stosowanych w stomatologii zachowawczej. Celem pracy była porównawcza ocena wpływu długotrwałego oddziaływania wilgotnego środowiska oraz cyklicznych obciążeń cieplnych na mikrotwardość nowego kompozytu bazującego na siloranach oraz dwóch tradycyjnych kompozytów bazujących na związkach metakrylanu. Próbki z kompozytów poddano oddziaływaniu środowiska soli fizjologicznej o cyklicznie zmiennych temperaturach (5°C i 65°C) wykorzystując specjalny symulator szoków termicznych. Wykonywano pomiary mikrotwardości metodą Vickersa przed rozpoczęciem testu zmęczenia cieplnego oraz po serii 4000 cykli termicznych. Wykazano, że w przeciwieństwie do tradycyjnych kompozytów stomatologicznych mikrotwardość kompozytu bazującego na siloranach nie zmniejsza się pod wpływem cyklicznego oddziaływania szoków termicznych odpowiadających warunkom fizjologicznym jamy ustnej. W warunkach przeprowadzonych badań stwierdzono nieznaczny wzrost tej mikrotwardości. Przeprowadzono również badania starzeniowe polegające na ocenie mikrotwardości próbek kompozytów przez okres 6 miesięcy. W tym okresie czasu próbki przechowywano w soli fizjologicznej. Pomiary mikrotwardości wykonywano po każdym miesiącu ekspozycji. Wykazano, że długotrwałe oddziaływanie środowiska soli fizjologicznej w warunkach stałej temperatury nie zmienia mikrotwardości żadnego z badanych materiałów. 1 1. Wstęp Kompozyty polimerowo-ceramiczne są wykorzystywane na szeroką skalę jako materiały do wypełnień stomatologicznych. Stosuje się je ze względu na ich właściwości estetyczne (dostępna jest bardzo szeroka gama barw) oraz dobre właściwości mechaniczne. Polimerowe materiały kompozytowe, oparte na związkach metakrylanowych składają się osnowy, którą stanowi żywica fotopolimeryzująca (do 40% objętości struktury) oraz z makrolub mikro- lub nanowypełniaczy nieorganicznych bazujących najczęściej na związkach krzemu (ponad 60% objętości struktury) [2,11,17]. Dodatkowo w skład kompozytu wchodzą fotoinicjatory i środki preadhezyjne. Krótką charakterystykę związków wchodzących w skład struktury kompozytu polimerowego przedstawiono w tabeli 1. Tab. 1. Cechy niektórych składników kompozytów polimerowo-ceramicznych stosowanych w stomatologii. Oznaczenie Nazwa składnika Cechy składnika Faza organiczna kompozytu (polimerowa) Bis-GMA dimetakrylan eteru diglicydowego bisfenolu TEGMA dimetakrylan glikolu tri etylenowego UDMA PCDMA dimetakrylan uretanowy dimetakrylan polikarbonowy Monomery Bis-GMA mają dużą masę cząsteczkową oraz ulegają wolnorodnikowej polimeryzacji addycyjnej, tworząc polimer bogaty w wiązania krzyżowe oraz inicjatory. Są to monomery rozpuszczające, czynią materiał kompozytowy mniej lepkim, ich zawartość wpływa również na poziom skurczu polimeryzacyjnego. - Faza nieorganiczna kompozytu - W skład tej fazy wchodzą m.in.: krzemiany glinowo-litowe, kwarc krystaliczny, szkło barowo-glinowo-borowo-krzemowe, szkło strontowo-barowo-aluminiowofluorosilikatowe, krzemionka, silinizowane wapno Ze względu na wielkość cząsteczek fazę nieorganiczną dzielimy na makrowypełniacze (wielkość cząsteczek od 1-30 μm) z kwarcu, szkła, ceramiki lub mikrowypełniacze (wielkość cząsteczek od 0,007 do 0,04 μm) preparowane z dwutlenku krzemu lub uzyskiwane w inny sposób – prepolimeryzowane w wyniku postępowania technologicznego (1-200 μm), aglomerowane (1-25 μm) lub sferoidalne (20-30 μm) Fotoinicjator CQ kamforochinon Fotoinicjatory służą polimeryzacji poprzez wytworzenie wolnych rodników (w przypadku fotopolimeryzacji światłem o dł. fali ok. 400 – 500 nm, światło niebieskie lub fioletowe W ostatnich latach w praktyce stomatologicznej coraz częściej wykorzystuje się materiały kompozytowe, w których osnową jest żywica bazująca na siloranach. Materiały kompozytowe posiadające matrycę siloranową charakteryzują się redukcją skurczu polimeryzacyjnego. Matryca siloranowa ma postać pierścieni otwierających się podczas polimeryzacji. Cząsteczki siloranu są opisywane jako hybryda złożona po połowie z cząstek siloxanów i oxiranów [15]. Materiały kompozytowe z matrycą siloranową dają nowe możliwości ograniczenia zjawiska nieszczelności brzeżnej wypełnień stomatologicznych [5,15,19]. W literaturze [14] podaje się również, że polimery bazujące na siloranach posiadają bardzo korzystne właściwości mechaniczne: wysoką udarność i odporność na zginanie a także zadowalającą twardość. Natomiast mało znana jest kwestia stabilności tych parametrów w warunkach długotrwałego oddziaływania wilgotnego środowiska oraz cyklicznych obciążeń cieplnych odpowiadających warunkom użytkowania w jamie ustnej. Dlatego autorzy niniejszej pracy podjęli badania mające na celu porównanie wpływu starzenia i zmęczenia cieplnego na mikrotwardość nowego kompozytu bazującego na siloranach oraz własności dwóch tradycyjnych kompozytów stomatologicznych bazujących na metakrylanach. Mikrotwardość jest właściwością kompozytu korelującą z odpornością na zużycie, także w przypadku zmęczenia cieplnego [3]. Badanie mikrotwardości pozwala na ocenę 2 właściwości mechanicznych kompozytu. Jak wykazano w badaniach [13] istnieją silne korelacje pomiędzy mikrotwardością kompozytu a wartościami modułu sprężystości, głębokością fotopolimeryzacji oraz najsilniejsze z poziomem skurczu polimeryzacyjnego. W publikacji dotyczącej relacji pomiędzy właściwościami fizyko-mechanicznymi tego typu kompozytów polimerowych z ich zużyciem [9,16] wykazano ścisłą zależność między mikrotwardością kompozytu a poziomem jego zużycia w warunkach symulacji In vitro. Ponadto, wykazano również korelację z poziomem konwersji wypełniacza kompozytu [6]. Badanie mikrotwardości może również służyć ocenie lokalnego gradientu fotopolimeryzacji, tzn. jednorodności właściwości kompozytu w obszarze oddziaływania widma światła lampy [8,18], wpływu czasu polimeryzacji i rodzaju światła lampy. Może także posłużyć jako miara rezydualnych właściwości mechanicznych w badaniach starzeniowych i zmęczeniowych. Podejmując badania symulacyjne in vitro zmęczenia cieplnego systemu mechanicznego ząb – wypełnienie kompozytowe należy zapewnić warunki obciążeń odpowiadające warunkom fizjologicznym jamy ustnej człowieka. Parametryzowane powinny być cechy środowiska symulacji takie jak: temperatura cieczy roboczej (sztucznej śliny lub soli fizjologicznej), czas przetrzymania cieczy roboczej w pojemniku z próbkami, bądź badanej próbki w pojemniku z cieczą roboczą oraz liczba cykli obciążeniowych (szoków termicznych) W dotychczas prowadzonych badaniach przyjmowano zróżnicowane założenia co do parametrów eksperymentu. Dolna temperatura cieczy roboczych wykorzystanych w eksperymentach wynosiła od 2 do 24ºC [7], natomiast temperatury cieczy podgrzanej wahały się od 45ºC [4] do 60 ºC [20]. Czas przetrzymania kondycjonowanej cieczy w pojemniku z próbkami wynosił od 15 do nawet 180 sekund, zaś liczba cykli wahała się od 25 do 1 mln cykli termicznych. Obecnie najczęściej zakłada się następujące parametry eksperymentu [1,7]: temperatura cieczy roboczej chłodzonej wynosi 5 ºC, temperatura cieczy roboczej podgrzanej wynosi od 55ºC do 65 ºC, czas przetrzymania cieczy roboczej w pojemniku z próbkami – 30 sekund, liczba cykli termicznych – od kilku tysięcy do stu tysięcy cykli. 2. Materiał i metoda badań W obydwu prowadzonych testach, zmęczenia cieplnego i starzenia, badano te same kompozyty stosowane w stomatologii. Były to tradycyjne materiały komercyjne oparte na związkach metakrylanowych o nazwach handlowych: Ice (SDI), Venus (Heraeus) oraz nowy kompozyt z matrycą siloranową o nazwie handlowej Filtek Silorane (3M ESPE) – tabela 2. Tab. 2. Wykaz badanych kompozytów Materiał Typ Ice Nanohybryda Producent SDI Rodzaj żywicy ester metakrylowy Venus Mikrohybryda Heraeus-Kulzer bisphenol-A glycidyl methacrylate (Bis-GMA) trietylen glycol dimethacrylate (TEGMA) Filtek Silorane Siloran 3M ESPE silorany Rodzaj i zawartość wypełniacza (wt%) 77,5 wypełniacz nieorganiczny 80 szkło barowe, wypełniacz nieorganiczny 76 wypełniacz nieorganiczny Rozmiar cząsteczek wypełniacza 0,04 – 1,5 µm 0,04 -1 µm 0,04 – 1,7 µm 3 Przykładowe wyniki analizy SEM badanych materiałów zaprezentowano na rysunku 1. Na obrazach mikroskopowych widoczne są cząsteczki fazy polimerowej kompozytu (większe). Widoczne są również cząsteczki wypełniacza nieorganicznego (mniejsze). Są to cząsteczki o bardziej regularnym kształcie, często zbliżonym do sferycznego, wielkości ziaren tej fazy są zbliżone. a) b) c) Rys. 1. Obrazy SEM przełomów badanych materiałów: a) Filtek Silorane, b) Ice, c) Venus Z wymienionych materiałów wykonano próbki w kształcie krążków o średnicy 14 mm i grubości 1 mm. Proces fotopolimeryzacji prowadzono za pomocą lampy halogenowej w sposób przedstawiony na rysunku 2. Czas naświetlania próbek przyjęto wg wskazań producentów materiałów i ustalono na 40s. Rys. 2. Sposób wykonywania próbek z kompozytów światło utwardzalnych: 1 – lampa, 2 – naświetlana próbka kompozytowa, 3 i 5 – płytki pleksiglasowe, 4 – forma stalowa W badaniach wykorzystano symulator szoków termicznych, skonstruowany w celu badania materiałów stomatologicznych. Urządzenie zostało wykonane przez pracowników Wydziału Mechanicznego Politechniki Lubelskiej we współpracy z Uniwersytetem Medycznym w Lublinie. Urządzenie do realizacji symulacji szoków termicznych (rys. 3) składa się z mikroprocesorowego układu sterowania oraz układu hydraulicznego. Urządzenie umożliwia wywoływanie szoków termicznych w próbkach umieszczonych w naczyniu pomiarowym zainstalowanym w symulatorze. Działanie urządzenia polega na okresowym pompowaniu i wypompowywaniu cieczy roboczej z naczynia pomiarowego. Naczynie pomiarowe jest naprzemiennie napełniane ogrzaną (temp. 65°C) lub ochłodzoną cieczą (temp. 5°C) roboczą z dwóch niezależnych układów kondycjonowania temperatury [12]. 4 Rys. 3. Stanowisko do badań trwałościowych z wyszczególnieniem podzespołów modułu szoków termicznych: 1 – mikroprocesorowy moduł sterujący, 2 – zawory sterujące, 3 – symulator żucia, 4 – pompa perystaltyczna, 5 – termostat chłodzący, 6 – ultratermostat grzejący Czas wykonywania poszczególnych procedur w ramach każdego cyklu szoków termicznych był zaprogramowany i powtarzalny. Czas przetrzymania cieczy chłodzonej i podgrzanej wynosił po 30s, czas wypompowania i wpompowania cieczy roboczej wynosił po 10s (rys. 4). Rys. 4. Algorytm szoku cieplnego z jednokrotnym pompowaniem cieczy roboczej Badanie mikrotwardości prowadzono metodą Vickersa na urządzeniu Futertech FM 700 (Future-tech Corp. Japan), pod obciążeniem 50g. Czas penetracji wgłębnika ustalono na 15s. Pomiary wykonano w dziesięciu punktach na powierzchniach próbek. Współrzędne pomiarów ustalono tak, aby obejmowały możliwe cały obszar powierzchni próbki. Były one takie same dla wszystkich próbek. Badanie prowadzono na naświetlanej (lc) i nienaświetlanej (nlc) płaszczyźnie próbek. 5 Analizy mikroskopowe kompozytów prowadzono w Laboratorium Mikroskopii Elektronowej Katolickiego Uniwersytetu Lubelskiego na skaningowym mikroskopie elektronowym (SEM) firmy Zeiss. Przeprowadzono również badania starzeniowe (przez okres 6 miesięcy), polegające na ocenie mikrotwardości kompozytów w funkcji czasu ekspozycji w soli fizjologicznej w warunkach stałej temperatury 36°C. Jednym z celów tych badań było uzyskanie próby referencyjnej do porównania z wynikami uzyskanymi w próbie zmęczenia cieplnego. 3. Wyniki badań Wyniki badań zmęczeniowych realizowanych z wykorzystaniem symulatora szoków cieplnych przedstawiono w tabeli 3 oraz na rysunku 5. Natomiast, wyniki badań starzeniowych przedstawiono na wykresach ramkowych na rysunku 6. Tab. 3. Statystyki opisowe wyników badań mikrotwardości. TC – cykle termiczne, lc – strona naświetlana, nlc – strona nienaświetlana Group Filtek Silorane 0TC lc Filtek Silorane 0TC nlc Filtek Silorane 4kTC lc Filtek Silorane 4kTC nlc Ice 0TC lc Ice 0TC nlc Ice 4kTC lc Ice 4kTC nlc Venus 0TC lc Venus 0TC nlc Venus 4kTC lc Venus 4kTC nlc Valid N 40 40 40 40 40 40 40 40 40 40 40 40 Mean 51,97275 49,38750 54,80075 50,15750 51,95350 49,96100 47,23800 44,58550 48,60775 48,88650 37,78125 35,14625 Median 52,20000 49,71000 55,24500 50,02500 51,86500 49,84000 47,22500 44,99500 48,39500 49,11000 37,88000 34,96000 Minimum 48,80000 42,76000 47,42000 42,09000 45,53000 45,18000 41,92000 35,28000 43,94000 42,90000 33,82000 30,37000 Maximum 57,00000 55,48000 61,31000 56,24000 57,71000 54,28000 51,75000 52,56000 55,73000 54,27000 41,77000 41,45000 Std.Dev. 1,791623 2,742934 3,102902 3,091503 2,736424 2,700346 2,737509 4,141693 2,613053 2,974151 1,973859 2,322400 Coef.Var. 3,447235 5,553903 5,662153 6,163590 5,267064 5,404909 5,795141 9,289327 5,375794 6,083787 5,224440 6,607817 6 a) b) c) Rys. 5. Szeregi rozdzielcze mikrotwardości badanych materiałów po testach zmęczeniowych: a) Filtek Siloran, b) Ice , c) Venus 7 a) b) c) Rys. 6. Zależność mikrotwardości badanych materiałów od czasu starzenia w soli fizjologicznej: a) Filtek Siloran, b) Ice , c) Venus 8 4. Dyskusja Wykazano wpływ szoków termicznych na spadek mikrotwardości tradycyjnych materiałów Venus oraz Ice. Największy spadek mikrotwardości po realizacji 4000 cykli termicznych (4kTC) odnotowano dla materiału Venus, co jest widoczne na wykresie (rys. 5c) oraz potwierdza się w wynikach testu t-Studenta (tab. 4). Wartości parametru t dla tego materiału były w większości porównań najwyższe. Wykazano również wpływ zmęczenia cieplnego na szybsze osłabienie powierzchni nienaświetlanej (nlc), niż powierzchni naświetlanej (lc) materiału Venus. Także badania materiału Ice potwierdziły wpływ szoków termicznych na spadek mikrotwardości. Nie był on tak duży jak to miało miejsce w przypadku materiału Venus, był jednak zauważalny. Zależność tą potwierdzono w teście t-Studenta. Najwyższe wartości parametru t uzyskano w porównaniach wyników uzyskanych przed i po realizacji testu zmęczenia cieplnego. Tab. 4. Wyniki testu t-Studenta istotności różnic mikrotwardości kompozytów. TC – cykle termiczne, lc – strona naświetlana, nlc – strona nienaświetlana. No 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Gr. 1 vs Gr. 2 FSi 0TC lc vs. Fsi 0TC nlc FSi 0TC lc vs. Fsi 4kTC nlc FSi 4kTC lc vs. Fsi 0TC lc FSi 4kTC lc vs. Fsi 0TC nlc FSi 4kTC lc vs. Fsi 4kTC nlc FSi 4kTC nlc vs. Fsi 0TC nlc Ice 0TC lc vs. Ice 0TC nlc Ice 0TC lc vs. Ice 4kTC lc Ice 0TC lc vs. Ice 4kTC nlc Ice 0TC nlc vs. Ice 4kTC lc Ice 0TC nlc vs. Ice 4kTC nlc Ice 4kTC lc vs. Ice 4kTC nlc Venus 0TC lc vs. Venus 4kTC lc Venus 0TC lc vs. Venus 4kTC nlc Venus 0TC nlc vs. Venus 0TC lc Venus 0TC nlc vs. Venus 4kTC lc Venus 0TC nlc vs. Venus 4kTC nlc Venus 4kTC lc vs. Venus 4kTC nlc t 4,99 3,21 4,99 8,27 6,7 1,18 3,28 7,71 9,39 4,48 6,88 3,38 20,91 24,35 0,45 19,68 23,03 5,47 df 78 78 78 78 78 78 78 78 78 78 78 78 78 78 78 78 78 78 p 0,000004 0,00191 0,000004 0 0 0,242252 0,001563 0 0 0,000025 0 0,001138 0 0 0,657331 0 0 0,000001 W wypadku materiału FSi (Filtek Silorane) nie odnotowano spadku mikrotwardości po realizacji 4000 cykli termicznych (4kTC). Stwierdzono natomiast jej niewielki wzrost. Statystycznie istotny wzrost mikrotwardości na powierzchni naświetlanej potwierdzają również wyniki testu t-Studenta. W przypadku tego materiału wpływ oddziaływań termicznych powodował skutki odmienne od obserwowanych dla pozostałych dwóch materiałów. Fakt ten należy wiązać z odmiennym składem fazowym kompozytu Filtek Siloran, w którym osnowa bazuje na związkach siloranu. Podobne spostrzeżenia zaprezentowano w pracy [14]. Analizując wyniki badań starzeniowych należy stwierdzić, że pomimo niewielkich fluktuacji mikrotwardości w okresie próby (rys. 6), nie wykazano jednoznacznego wpływu ekspozycji w soli fizjologicznej na zmianę mikrotwardości badanych kompozytów. 9 5. Wnioski: 1. Wykazano, że szoki termiczne odpowiadające warunkom jamy ustnej nie mają wpływu na spadek mikrotwardości materiału FSi (Filtek Siloran) bazującego na związkach siloranu. 2. W przypadku tradycyjnych materiałów Venus i Ice (bazujących na związkach metakrylanowych) stwierdzono spadek mikrotwardości pod wpływem zmęczenia cieplnego. 3. Nie wykazano wpływu starzenia badanych materiałów w środowisku soli fizjologicznej na spadek ich mikrotwardości. 4. Zastosowana metoda badań może być przydatna do wstępnej kwalifikacji przydatności eksploatacyjnej kompozytowych materiałów stomatologicznych Literatura: 1. Achilias D S., Karabela M M., Sideridou I D. Thermal degradation of light-cured dimethacrylate resins Part I. Isoconversional kinetic analysis. Thermochemica Acta 2008; 472: 74–83. 2. Andrzejczuk M., Lewandowska M., Kurzydłowski K J. Właściwości mechaniczne światłoutwardzalnych kompozytów zbrojonych mikro- i nanocząstkami. Kompozyty (Composites) 2005; (5)1: 75-79. 3. Assuncao W G., Gomes E A., Barao V A R., Barbosa D B., Delbenc J A., Tabata L F. Effect of storage in artificial saliva and thermal cycling on Knoop hardness of resin denture teeth. Journal of Prosthodontic Research 2010; 54: 123–127. 4. Chan K C., Swift E J. Marginal seal of new generation dental bonding agents. Journal of Prosthet. Dentistry 1994; 72: 420–423. 5. Eick J D., Kotha S P., Chappelow C C., Kilway K V., Giese G J., Glaros A G. Properties of silorane-based dental resins and composites containing a stress-reducing monomer. Dental Materials 2007; 23: 1011-1017 6. Ferracane J L. Correlation between hardness and degree of conversion during the setting reaction of unfilled dental restorative resins. Dental Materials 1985; 1: 11—14. 7. Gale M S., Darvell B W. Thermal cycling procedures for laboratory testing of dental restorations. Journal of Dentistry 1999; 27: 89–99. 8. Geis-Gerstorfer J. In vitro corrosion measurements of dental alloys. Journal of Dentistry 1994; 22: 247-51. 9. Heintze S D., Zellweger G., Zappini G. The relationship between physical parameters and wear of dental composites. Wear 2007; 263; 1138–1146. 10. Hill T., Lewicki P. Statistics: methods and applications: a comprehensive reference for science, industry and data mining. wyd. StatSoft 2006. 10 11. Joyston-Bechal A., Kidd E., Joyston-Bechal S. Essentials of dental caries: the disease and its management. 2nd ed. wyd. Oxford University Press, Oxford 1998. 12. Kordos P., Hunicz J., Niewczas A. The station designed for accelerated fatigue tests of dental materials. Ekspoloatacja i Niezawodnosc. Maintenance and Reliability 2009; 1(41): 63-69. 13. Li J., Li H., Foka A S L., Watts D.C. Multiple correlations of material parameters of light-cured dental composites. Dental Materials 2009; 25: 829–836. 14. Lien W., Vandewalle K S. Physical properties of a new silorane-based restorative system. Dental Materials 2010; 26: 337-344 15. Moszner N., Salz U. New development of polymeric dental composites. Prog Polym Sci 2001; 26: 535-536 16. Sanders-Tavares da Cunha Mello F., Feilzer A J., de Gee A J., Davidson C L. Sealing ability of eight resin bonding systems in a Class II restoration after mechanical fatiguing. Dental Materials 1997; 13: 372-376. 17. Szafran M., Rokicki G., Bobryk E., Szczęsna B. Effect of filler’s surface treatment on mechanical properties of ceramic-polymer composites used in dentistry. Kompozyty (Composites) 2006; 6(3): 78-82. 18. Versluis A., Tantbirojn D., Douglas W H. Do dental composites always shrink toward the light? Journal of Dental Research 1998; 77: 1435—45 19. Weinmann W., Thalacker C., Guggenberger R. Siloranes in dental composites. Dental Materials 2005; 21: 68-74 20. Yoshida K., Matsumura H., Atsuta M. Monomer composition and bond strength of lightcured 4-META opaque resin. Journal of Dental Restoration 1990; 69: 849–851. 11