RELIABILITY ANALYSIS OF SOME INDUSTRIAL SYSTEMS USING SOFT COMPUTING TECHNIQUES

Abstract

In the design of critical combinations and complex integrations of large engineering systems, day-by-day, their reliability, availability and maintainability (RAM) analysis of the inherent processes in the system and their related equipments riced to be determined. The main feature of these attributes is to increase the designed life of the system by reducing or elinrunating the probability of failureE. For this, to achueve the desired goals of production, the system analyst or plant personnel view reliability and niaintainability as the main pursuit for improving the quality of the system/products so that system should remain active for mnaxinlum possible interval. So it is expected that a production system should remain operative for mnaximuni possible duration to achieve time desired goals of production. But unfortunately failure is an inevitable fact related with technological products and systemS. Over a time, however, a givell system suffers failures and has to be brought ba.k into the serviceable state through appropriate maintenance and repairs. Further, age and undesirable operating conditions of manufacturing/production processes affect each part of the system differently. Generally, a system analysts model and analyze the system behavior through various qualitative and c1uarititative tools/techniques. These techniques require precise knowledge of numerical probabilities and system's components' functional dependencies which may be difficult to be obtained in any large-scale system. As the data are collected or available from the historical records are mostly uncertain, limited and imprecise in nature. Thus in such situation(s) it is difficult, if not impossible, to analyze the behavior and performance of the system Ii up to desired degree of accuracy. If somehow they can be analyzed then they have a. large amount of uncertainties in the analysis. Therefore, it is very hard to construct a precise and comprehensive mathematical model for a.n industria.l system which may be close to real situation. Thus in order to minimize the deleterious effects of system/subsystem failures and to plan/ adopt suitable maintenance strategies, a. throughout analysis of their reliability parameters is desirable by utilizing available resources and uncertain data. The objective of this research work is to analyze the behavior and performance of repairable industrial systems by utilizing available information and uncertain data. For analyzing this a hybridized technique named as Particle swarm optimization based Lambda-Tau (PSOBLT) has been proposed. Fuzzy set theory has been used in the technique for handling the uncertainties in the data and then behavior of the system are analyzed in the form of fuzzy membership functions of various reliability parameters namely systems failure rate, repair time, mean tirrie between failures (MTBF), expected number of failures (ENOF), reliability and availability which affects the system performance. Major advantages of proposed technique is that it gives compressed range of prediction for all computed reliability parameters by utilizing uncertain data. Further using these results, systerni performance has been analyzed by formulating a nonlinear fuzzy optimization problem by considering reliability and cost as objectives. Also the technique has been applied for a. time varying failure and repair rate model instead of constant failure rate model. The present thesis is organized into nine chapters which are briefly summarized as follows: A brief account of the related work of various authors in evaluation of system reliability by using conventional, fuzzy arid optimization techniques is presented in the first chapter. In Chapter 2, the basics and preliminaries related to the reliability analysis and to be used in subsequent chapters are given. U' Chapter 3 presents a hybridized technique nainc.d as particle swarm optimization based larnbda-tau (PSOBLT) technique for analyzing the behavior of an industrial system by utilizing the uncertain.vaguc and limited data. The proposed technique has been applied to various subsystems/units of a paper mill, a complex repairable industrial system, and analyzed their behavior in terms of various reliability parameters/indices in the form of fuzzy membership functions. Computed results are compared with the existing results as obtained by other researchers. Sensitivity and perfoririancc' analysis have also been done on system availability for ranking the sensitive components of each subsystem of the system which helps the system analyst or DM to maintain the behavior of the system as per his preferential order. In Chapter 4, the behavior analysis of the urea fertilizer plant, a complex repairable industrial system, have been dealt by using PSOBLT technique. Using their behavior analysis results, fuzzy multi-objective optimization problem (FMOOP) is forniulated for each subsystems of the plant by taking system's reliability and cost as objectives. Air int,cra.ctive method for solving multi-objective reliability optinniiza tiori problems modeled in crisp and fuzzy enviroinnients is presented. In this. tire conflicting na.tute of the objectives are resolved by defining their fuzzy goals (linear membership functions only) and using the preference of decision makers (DMs) towards the objectives, the problem is reformulated to single objective optimization problem and solved by using PSO algorithm. During the interactive phase, based on the outcomes of previous iteration, the DM has the option to change his or her preferences in view of the importance being given by him to different objectives. Chapter 5 deals with the reliability-redundancy allocation problem (RRAP) of the series system (a Pharmaceutical plant) under the fuzzy environment. In this forniulation, decision variables are treated as uncertain and hence corresponding problem is solved under fuzzy environment. The linear as well as nonlinear (signioidal) membership functions are defined corresponding to their fuzzy goals and then problem is solved by iterative process as described in chapter 4. iv In Chapter 6, performance analysis of repairable industrial systems has been done by incorporating their behavior analysis results. For this FMOOP is fornwlated 40 by considering all the reliability indices as obtained during their behavior analysis. Exponential distribution membership functions have been fitted corresponding to each of the objective function and then based on the preference of DM towards the objective, a problem is reformulated to an equivalent crisp optimization problem and then solved by using PSO. The presented approach has been applied to optimize the performance of an industrial system namely a cattle-feed plant. Chapter 7 describes an approach for reliability and maintainability analysis of an industrial system by using the collected field failure and repair data. The best fit of the failure and repair data between the common theoretical distributions are found by Anderson Darling (AD) goodness-of-fit test. The respective parameters are obtained by optimizing their likelihood functions. Furthermore, the survival and hazard rate models of the entire system are calculated. The reliability, availability and maintainability analysis of the crank - case manufacturing plaint over a period of one year was investigated. In Chapter 8 PSOBLT technique has been used to obtain fuzzy reliability parameters by utilizing the results as obtained in chapter 7. The parameters related to failure and repair rate distributions are tirrie varying arid follow Weibull and Normal distributions respectively instead of constant distribution as discussed in previous chapters. Behavior of the system has been analyzed in the form of crisp and defuzzifled values at +15%, +25% and +50% spreads. Sensitivity analysis has also been done to reflect the effect of components failure and repair rate on system performance and hence, based on their performance as per preferential order, ranking of the components has been presented. Chapter 9 deals with the overall concluding observations of this study and a brief discussion oIl the scope for future work.