Reliability and Sensitivity Analysis of Structures Using Adaptive Neuro-Fuzzy Systems

In this study, Adaptive Neuro-Fuzzy Inference System (ANFIS) and Monte Carlo simulation are utilized for reliability analysis of structures. The drawback of Monte Carlo Simulation is the amount of computational efforts. ANFIS is capable of approximating structural response for calculating probability of failure, letting the computation burden at much lower cost. In fact, ANFIS derives adaptively an explicit approximation of the implicit limit state functions. For this purpose, a quasi-sensitivity analysis based on ANFIS was developed for determination of dominant design variables, led to the approximation of the structural failure probability. Preparation of ANFIS however, was preceded using a relaxation based method developed by which the optimum number of training samples and epochs was obtained. That was introduced to more efficiently reduce the computational time of ANFIS training. The proposed methodology was considered using some illustrative examples.