基于人工神经网络的多处损伤加筋板剩余强度预测

用BP神经网络算法对多处损伤加筋板的剩余强度数据进行训练学习,将预测值和3种经典分析方法的计算值与实验值进行对比,结果表明,ANN法预测值与实验值吻合得最好,LMC修正法和WSU3修正法次之,Swift塑性区连通法最差。最后用所建立的BP网络对不同主裂纹半长和韧带长度的剩余强度进行了预测,结果发现,在其他参数不变的情况下,不管是双筋条还是三筋条加筋板,剩余强度总是随主裂纹半长的增加而成线性降低,随韧带长度的增加而成线性增加,但双筋条加筋板比三筋条加筋板对主裂纹半长和韧带长度的变化更加敏感。

Prediction model for residual strength of stiffened panels with multiple site damage based on artificial neural network

A prediction model for residual strength of stiffened panels with multiple site damage based on artificial neural network (ANN) is developed, and the results obtained from the trained BP model are compared to the analytical and experimental data available in the literature. The results obtained indicate that the neural network model predictions are in the best agreement with the experimental data than any other methods, and the modified linkup models predict better than the linkup model proposed by Swift. In the end several simulations are carried out to predict the trends with varying input parameters. The results show that the residual strength decreases linearly as the half crack length of lead crack increases and increases linearly as the ligament length increases for both kinds of stiffened panels, but the one bay stiffened panels are more sensitive to the change than the two bay stiffened panels.