Crack Growth Behavior Prediction Based on Bayesian Regularized Neural Networks
Based on the extended finite element method and classical theory,the material parameters in the finite element simulation are calculated,and the 7050 aluminum alloy plates with the center preset cracks are respectively subjected to static load breaking for batch calculation of finite element simulation,the damage propagation characteristics of preset cracks with different lengths and angles under different loading levels are observed.Aiming at the fast assessment of fracture behavior simulation of vehicle struc-tural alloy materials,proposes a rapid prediction method of crack growth behavior based on Bayesian regu-larized neural network model.Construct a neural network model for crack growth prediction,improve the task adaptability of the neural network model,use the Bayesian regularization method to train the model to obtain network weights,and compare the prediction accuracy of various neural network training methods.The simulation result shows that the method proposed is fast and has good prediction accuracy,which can meet the requirements of rapid evaluation of crack propagation characteristics after detecting such cracks in aircraft structures in engineering.
fracturefinite element methodBayesian networkneural networks
NETL
NSTL
万方数据
