Reliability analysis using artificial neural networks

A probabilistic analysis approach is developed by extending the Monte Carlo simulation. The Multilayer perceptron with backpropagation learning algorithm is applied in reliability analysis as the substitute of finite element solver. The reliability of a tunnel is analyzed as an example. Through Monte Carlo simulations, the input and output samples of the network are obtained. As comparing to the responses obtained by Monte Carlo simulations with finite element solver, the network performs high accuracy and fast training speed. The results show that the proposed approach is a promising tool for stochastic analysis inasmuch as the error with respect to finite element solver is negligible.