Method for predicting thermal fatigue life of laminated welding spot based on convolutional neural network

The invention provides a method for predicting the thermal fatigue life of a laminated welding spot based on a convolutional neural network. The method comprises the following steps: firstly, determining different form factors influencing the thermal fatigue life of the laminated welding spot, determining the level number of each factor, establishing a Cartesian product combination test group according to the determined factor levels, then establishing a simulation model of the laminated welding spot according to the test group based on ANSYS software, applying a temperature cycle load, and performing finite element simulation to obtain a simulation result. The method comprises the following steps: determining a welding spot with the maximum stress in a laminated welding spot array as a dangerous welding spot, drawing a curve that the strain of the dangerous welding spot changes along with time under temperature circulation, taking the curve that the plastic strain of the dangerous welding spot of the laminat