Abstract

Under uncertain conditions the field of reliability-based optimization provides a rational framework for structural optimization which explicitly accounts for the uncertainties. Reliability-based design formulations require advanced and efficient tools for structural modeling, reliability estimation, reliability sensitivity analysis, and mathematical programming. This is of particular significance when dealing with involved stochastic finite element models. The solution of this type of problems can be obtained by a number of techniques such as deterministic optimization schemes or simulation based algorithms. Among these techniques efficient first-order schemes, where the number of function evaluations (solution of finite element models) is minimized during the design process, are especially attractive. One of the key aspects in the implementation of these algorithms is the ability to perform efficient reliability sensitivity analyses. In this context, it is the objective of this talk to present several schemes for efficient reliability sensitivity analysis. The approaches are basically simple post-processing of an advanced sampling-based reliability analysis, namely subset simulation. The influence of design variables modeled as deterministic or random vectors into the system reliability is considered explicitly in the analyses. The performance of the proposed methodologies is illustrated by a number of numerical examples involving finite element models under stochastic excitation. Future research directions as well as new challenges and opportunities are finally discussed.