Application of Artificial Neural Network in Monte CarloSimulation Analysis of a Stochastic Structure

摘要

In analysis of a stochastic structure, the Monte Carlo simulation method can provide the moststraightforward and accurate solutions to various structural problems. However, time-consuming insimulation of lots of random digital samples restricts it in a narrow use. In this paper, use is made of theintelligent tool, Artificial Neural Network, to replace the deterministic FEM solver in MC simulation. Theidea is that FEM is only used to generate the input-output pairs needed for training ANN and that thetrained ANN can map out all the MC samples instantly for the structural responses. Numerical results onanalysis of a bending plate have shown that the proposed MC-ANN method is likely to improve thecomputing efficiency of the MC method by tens of times with little loss in accuracy.