摘要
以多层前馈神经网络为基本结构,以误差反向传播算法(BP算法)为网络训练方法,研究40Cr在扭转疲劳载荷下的超低周疲劳断裂问题。训练结果表明:疲劳裂纹的形成寿命随着过渡圆角半径的变大而增长,随着径比的增大而呈缩短趋势,径比越大、过渡圆角半径越小,疲劳裂纹的形成寿命越短。BP神经网络模型训练的结果与试验结果基本吻合,说明此神经网络模型有很好的特性,为超低周疲劳断裂研究提供了一种新方法。
Based on the feed-forward multi-layer neutral network and taking the error back-propagation algorithm as the network training method, the paper researches the problems of the fracture of 40Cr steel in extra-low cycle fatigue torsional loading. The training result shows that fatigue crack＇s forming life becomes longer with increasing the tip radius and shorter with increasing radius ratio （D/d）, vice versa. The training result from BP neutral network is very close to the test result. It means that this neutral network model has a better performance, providing a new method for predicting extra-low cycle fatigue fracture.
出处
《起重运输机械》
2009年第7期80-83,共4页
Hoisting and Conveying Machinery
