A safety assessment approach to pressure vessels based on machine learning

Abstract

The safety assessment of a pressure vessel with a surface crack is an important part of the safety assessment of engineering equipment. However, the existing methods are mostly based on the assumption of plane specimens and the K criterion applicable to brittle fracture, which may lead to unacceptable errors when applied to a fracture problem in an elastoplastic pressure vessel. In this article, based on the finite element method (FEM) and artificial neural network (ANN), the elastic-plastic three-dimensional J-integral of a crack tip in a pressure vessel with an axial semi-elliptic crack on the surface under the loading of internal pressure is studied. First, the influence of the vessel geometry, the crack size, and internal pressure on the three-dimensional J-integral is analyzed. Second, the machine learning dataset is constructed based on the results of 1,200 cases of FEM calculation; then ANNs are used to discover the potential relationship between multiple parameters and the three-dimensional J-integral. The results show that the neural network constructed in this article can well predict the elastoplastic three-dimensional J-integral of a pressure vessel surface crack.