COMPUTATIONAL METHODS FOR EFFICIENT STRUCTURAL RELIABILITY AND RELIABILITY SENSITIVITY ANALYSIS

This paper proposes an efficient, adaptive importance sampling (AIS) m ethod that can be used to compute component and system reliability and reliability sensitivities. The AIS approach uses a sampling density t hat is proportional to the joint probability density function of the r andom variables. Starting from an initial approximate failure domain, sampling proceeds adaptively and incrementally to reach a sampling dom ain that is slightly greater than the failure domain to minimize overs ampling in the safe region. Several reliability sensitivity coefficien ts are proposed that can be computed directly and easily from the prev ious AIS-based failure points. These sensitivities can be used to iden tify key random variables and to adjust a design to achieve reliabilit y-based objectives. The proposed methodology is demonstrated using a t urbine blade reliability analysis problem.