基于卷积神经网络的焊接接头疲劳寿命预测
Fatigue Life Prediction of Welded Joints Based on Convolutional Neural Networks
邱广宇 1耿健 1史少聪 1李昱1
作者信息
- 1. 大连交通大学 机车车辆工程学院,大连 116028
- 折叠
摘要
针对利用最小二乘法寻找应力-疲劳寿命曲线受制于特定加载条件与材料假设的问题,考虑影响疲劳寿命的不确定因素,提出一种基于卷积神经网络的疲劳寿命预测方法.测试结果表明,对于不同接头形式和材料类型的高低周疲劳寿命,卷积神经网络模型均能够进行有效预测,且其预测准确性和稳定性比反向传播(Back Propagation,BP)神经网络和最小二乘法方法更高.
Abstract
In order to find the problem that the stress-fatigue life curve is subject to specific loading conditions and material assumptions by using the least square method,a fatigue life prediction method based on convolutional neural network is proposed considering the uncertain factors affecting the fatigue life.The test results show that the convolutional neural network model can effectively predict the high-low cycle fatigue life of different joint forms and material types,and its prediction accuracy and stability are higher than that of Back Propagation(BP)neural network and least square method.
关键词
焊接接头/疲劳寿命/预测/卷积神经网络Key words
welded joint/fatigue life/prediction/convolutional neural networks引用本文复制引用
出版年
2024
NETL
NSTL
万方数据