Efficient methods for reliability-based design optimization

This dissertation develops efficient methods for reliability-based design optimization (RBDO). Accurate and efficient methods are developed for reliability analysis at both component and system levels. Efficient optimization approaches are developed to enable the accurate reliability methods to be adopted by RBDO.; Two reliability analysis methods are developed. The first method combines analytical and simulation-based methods through adaptive response surfaces, and the second method performs Monte Carlo simulation based on the response surface of the failure indicator function. The developed methods are applied to solve an automotive door problem in terms of satisfying two quality issues. A general framework of the proposed method for reliability analysis of problems with a large number of random variables at both component and system level is constructed.; A decoupling method for efficient RBDO approach is developed next, based on direct reliability analysis, as opposed to existing decoupling methods that depend on inverse first-order reliability analysis. Since the proposed approach does not depend on first-order reliability method for reliability analysis, it can take advantage of more accurate reliability analysis methods, leading to higher accuracy and actual feasibility of the RBDO solution. A decoupled RBDO method using efficient simulation-based techniques for reliability assessment is also presented. The use of simulation enables system-level reliability to be included in the decoupled RBDO formulation. The proposed decoupled RBDO methods are applied to a vehicle side impact problem.; The satisfaction of the two automotive quality issues are probabilistically evaluated and treated as two objectives in the optimization formulation. A multi-objective RBDO method is developed to generate a series of solutions, under different scenarios, which can be utilized to facilitate decision-making. The proposed method is general and can be applied to solve a wide variety of RBDO problems with competing objectives and reliability constraints at both component and system levels, including reliability-based robust design (RBRD) in which the mean and variation of the objective function are of interest.