Abstract: To address the issue that metal magnetic memory testing of pressure vessels cannot achieve quantitative analysis and identification, a three-dimensional force-magnetic coupling model was established using ANSYS software based on the energy conservation relationship and electromagnetic field theory. Based on defect types such as circular holes, pits, rectangular grooves, and pores, the variation of metal magnetic memory signals with defect depth was analyzed, and the magnetic memory signals were then integrated. Results showed that as the defect depth increased, the magnetic field energy amplitude near pit defects and rectangular groove defects exhibited linear changes. Meanwhile, the study also found that the integral method was effective for studying stress concentration and macroscopic defects in ferromagnetic materials. Furthermore, after introducing a BP neural network to the proposed method, intelligent classification of defect types was achieved, enabling both defect localization and qualitative classification.