Research On Failure Rate Estimation And Multi-objective Optimization Methods For Complex Systems

Engineering systems especially in aerospace industries are increasingly becomingcomplex. As system complexity increases, system failure rates estimation difficulty alsoincreases. Increase system complexity usually generates conflicting design objectives.Designing systems with optimal parameter setting requires efficient optimizationtechniques to handle these multiple objectives design issues. In this thesis the failure rateestimation and multi-objective optimization of complex systems is studied.Usually, analysts apply approaches based on branch reduction or time in attemptto calculate the failure rates of complex systems. In this work, reduction-minimal cuttechnique is proposed for the estimation of complex system failure rates. Series-paralleland complex mixed configuration system structures, comprising of non-identicalcomponents and network system structure comprising of identical components were usedto validate the proposed approach. Approximated percent error of the failure rateestimated using oversimplified approach in relation to the true failure rate based on theproposed technique were estimated and discussed for the three system structuresconsidered.The optimal solution of a multi-objective optimization problem (MOP)corresponds to a Pareto set that is characterized by a tradeoff between objectives. GeneticPareto Set Identification Algorithm (GPSIA) is proposed for multi-objective optimizationproblems. GPSIA is a hybrid technique which combines genetic and heuristic principlesto generate non-dominated solutions. Series–parallel system with active redundancy wasused to validate the proposed algorithm. System reliability and cost were the researchobjective functions subject to system weight constrain. The results reveal evenlydistributed non-dominated set. The distances between successive Pareto points were usedto evaluate the general performance of the method. Plots were also used to show thecomputational results for the type of system studied and the robustness of the technique is discussed in comparison with NSGA-II and SPEA-2.Constraints are usually prevalent in real-world optimization problems. Theoptimization strategy based on GPSIA only considered feasible solutions. That is,solutions, which satisfies all constraints conditions. Infeasible solutions could facilitatefaster convergence to the true Pareto front. In this research a biologically inspiredconstrained GPSIA called GPSIA+DS based on the previously proposed GPSIA, whichconsiders the feasible and infeasible solution in it optimization strategy is furtherproposed. Complex system comprising of mixed configuration, k-out-of-n and redundantsubsystem was used to test GPSIA+DS in comparison with GPSIA and NSGA-II. Theresult showed that GPSIA+DS is efficient.GSPIA+DS is applied to the multi-objective optimization of an aircraft landinggear position indication and actuation system. The efficient of the constrained GPSIA isdemonstrated through the reliability engineering application.