Neural network method for probabilistic safety assessment of flawed weld structures

摘要

Based on the mapping theorem of multi-layer neural network, an artificial neural network(ANN) model of pressure vessel probabilistic safety assessment (PSA) based on the two criterion analysis method is established with the consideration of the parameters' randomness. Orthogonal test is used to acquire the training samples and testing samples, the trained ANN combined with Monte-Carlo method is used to calculate the failure probability. Comparing the results calculated by this method with the results by direct Monte-Carlo method in engineering application, it can be proven that this method is an efficient approach to realize the intelligent assessment for the PSA of structure integrity.