基于反向传播神经网络的疲劳裂纹扩展分析

材料发生疲劳断裂时往往会引起重大安全事故,而基于传统数值模拟方法求解疲劳裂纹扩展问题时模型复杂、计算量大.本文基于包含多隐层的反向传播神经网络分析金属材料疲劳裂纹扩展行为,计算了裂纹扩展过程中的von Mises应力场和位移场,并与数值解和实验解进行对比,误差分析结果表明其求解精度高.并基于该神经网络有效预测了裂纹扩展中裂纹长度及裂纹扩展速率的变化过程,预测精度高.该神经网络分析方法可为材料剩余寿命和疲劳强度预测提供研究基础.

Fatigue Crack Propagation Analysis Based on Back Propagation Neural Network

Fatigue fracture of materials often leads to serious safety accidents. However, when solving fatigue crack propagation problem based on traditional numerical simulation method, the model is complicated and the calculation is expensive. In this paper, the fields of von Mises stress and displacement were computed based on the back-propagation neural network with multi hidden layers, and thus the fatigue crack propagation behavior of metal materials were analyzed. The results were compared with the numerical solution and the experimental solution. The error analysis shows that the proposed method has high accuracy. Based on the neural network, the variation of the crack length and the crack growth rate in the crack propagation process can be predicted effectively, with high prediction accuracy. This neural network analysis method can provide a research basis for predicting the residual life and fatigue strength of materials