GA-BP神经网络的点蚀坑应力集中系数预测

Prediction on Stress Concentration Factor of Corrosion Pit Based on GA-BP Neural Network

摘要: 应用人工神经网络具有高度非线性映射功能，对腐蚀点蚀坑的应力集中系数进行预测。将遗传算法GA(Genetic Algorithm)和反向传播BP(Back Propagation)神经网络相结合建立了预测应力集中系数的GA-BP神经网络模型。运用该模型分别对含半椭球蚀坑圆棒在单轴拉伸和弯曲载荷下的应力集中系数随深径比的变化进行预测，并与BP神经网络预测结果以及有限元计算所得结果相对比。结果表明，GA-BP算法的神经网络预测精度要高于BP网络的，与有限元计算结果吻合度更高(误差在1.5%以内)，说明GA-BP神经网络可以用于预测应力集中系数。

Abstract: The highly nonlinear mapping function of artificial neural network enables it to predict the stress concentration factor of corrosion pits. By combing GA（Genetic Algorithm） with BP（Back Propagation） neural network, a GA-BP neural network model was developed to calculate the stress concentration factors of pits with various ratios of depth and diameter on a round bar under axial tension and bending. The results show that GA-BP agrees well with the finite element results（the error is less than 1.5%） and gives a better prediction than BP, indicating that GA-BP model is able to prediction the stress concentration factor.