Continuum Sensitivity Method for Reliability-Based Structural Design and Optimization*

Abstract

The computational cost of reliability analysis and optimization of medium/large structural systems is prohibitively high. One of the reasons for this situation is the use of inefficient methods, such as finite differences, for design sensitivity analysis. A significant increase in the efficiency of reliability computations is expected from the integration of reliability methods with the discrete and continuum methods of design sensitivity analysis. So far, only the discrete method has been used in the so-called stochastic finite element method. Integration of reliability computations with the continuum method of design sensitivity analysis is proposed in this paper as a new, powerful approach to the probabilistic design of large structural systems. Basic formulations and solution methods of reliability analysis, design sensitivity analysis of reliability indices, reliability-based optimization, and multiobjective reliability-based optimization are reviewed with focus on the sensitivity information necessary to carry out the reliability computations. The most time-consuming sensitivity computations, related to the dependence of structural performance measures on design parameters, are proposed to be performed using the continuum method of design sensitivity analysis. Efficiency of the methodology is demonstrated on an example of reliability-based optimization of a hydraulic shear machine. Sine the approach allows the integration of already existing CADICAE computational resources with state-of-the-art design sensitivity and reliability techniques, it has the potential to become a fundamental strategy for reliability-based design and optimization of large structural systems.