Research on Fatigue Stress Reconstruction of Major Equipment Based on Neural Network

Abstract

Abstract This article first proposes a stress/strain reconstruction method based on neural networks. The construction of the dataset and the setting of the network structure are introduced around this method. Standard component strain reconstruction experiments are carried out to verify the method, and the error between the reconstructed values and the measured values is within 15%. This article further takes the tunnel boring machine (TBM) disk cutter shaft as the research object, extracts the coordinate load stress dataset of dangerous positions through finite element method, trains and tests the stress reconstruction model, and the error between the reconstruction value and the simulation value is within 10%. Finally, this article utilized a stress reconstruction model to reconstruct the stress time history of the dangerous position of the TBM rolling cutter shaft and evaluated the fatigue life of the cutter shaft based on various life criteria.