Research on Fatigue Crack Propagation and Life Prediction Based on GRU Neural Network
摘要
The initiation and propagation of fatigue cracks will seriously affect the safety performance of structures, so it is the key to study the propagation law of fatigue cracks and predict the residual life of structures. Based on crack propagation mechanism, this paper analyzes the key factors affecting the crack propagation of hull plates, and designs a reasonable crack propagation rule and residual life prediction scheme. Based on the joint simulation of ABAQUS and FRANC3D, the fatigue crack propagation of hull plates under different working conditions was analyzed, and the finite element model of hull plates was established in ABAQUS. The finite element model of hull plate was simulated by FRANC3D and the remaining life was analyzed. The GRU neural network model built by Python was used to train the remaining life data set constructed by simulation, and the remaining life of fatigue cracks of hull plates was predicted with a prediction accuracy of more than 90%.