Piping reliability prediction using Monte Carlo simulation and artificial neural network

Abstract

To analyze the effect of intergranular stress corrosion cracking (IG-SCC) on the probability of boiling water reactor (BWR) pipe failure, we have introduced some modifications to the IG-SCC model for the piping reliability analysis, including seismic events (PRAISE) code of probabilistic fracture mechanics. The purpose of this article is to evaluate the probability of failure under IG-SCC of several pipe sizes using Monte Carlo simulations (MCS), sensitivity analysis and artificial neural networks (ANN). The MCS generates the reliability data and input parameters for modeling and ANN training. The ANN inputs are the sampled parameters, while the ANN outputs are the reliability estimated by the MCS. The entire database generated by the MCS will be separated into three groups. The data groups are intended for training, testing and validation ANN respectively. The percentage of the entire database for each group should be determined according to specific requirements. Examples are given to demonstrate the proposed method. The retained ANN can be used to efficiently and accurately estimate the reliability of leaks from damaged pipes. Finally, we have observed a strong correlation between the end of life failure probability and the parameter characterizing the damage for each pipe size witch is predicted using a second ANN. This damage parameter can be used to evaluate structural reliability and identify the most effective approaches to improve pipe reliability.