COMPUTATIONAL INTELLIGENCE-BASED RELIABILITY DESIGN OPTIMIZATION FOR STRUCTURAL SYSTEM

The probability constraint is transferred into equivalent determinate constraint by Probabilistic Perturbation method and Edgeworth series method in reliability design optimization with non-normal random parameters. Thus, the initial design point in structural system for reliability design optimization can be obtained rapidly and accurately by Particle Swarm Algorithm. In the case of system with multi-failure modes, the Monte Carlo Stochastic-Wavelet Neural Network (MCS-WNN) method is presented. This approach can implement effectively reliability simulation of structural system. And then, an inverse mapping model of Wavelet Neural Net is presented to accomplish design parameter optimization in structural system. Experimental results show that the aforementioned methods are effective.