焊管2025，Vol.48Issue(3)：23-28,6.DOI:10.19291/j.cnki.1001-3938.2025.03.004
基于注意力机制增强CNN-LSTM的在役管道焊接剩余强度预测研究
Prediction Research of Welding Residual Strength of In-service Pipelines Based on CNN-LSTM-Attention
叶均磊 1陆金桂1
作者信息
- 1. 南京工业大学 机械与动力工程学院,南京 211816
- 折叠
摘要
Abstract
In order to ensure the quality and safety of in-service pipeline welding,an attention mechanism-enhanced Convolutional Neural Network-Long Short-Term Memory network(CNN-LSTM-Attention)prediction model of residual strength for in-service pipeline welding is proposed.The parameter features of in-service pipeline welding data are first extracted by combining the convolutional neural network and long short-term memory network in the model.Then,the key information is highlighted by introducing the attention mechanism to reduce the loss of historical information.Finally,the prediction results are output through the fully connected layer.The results show that the mean absolute percentage error(MAPE),root mean square error(RMSE),and coefficient of determination(R2)of the proposed CNN-LSTM-Attention model are 0.058 9,0.570 2 and 0.991 0,respectively.The prediction effect is significantly improved compared with CNN,LSTM and CNN-LSTM,which can effectively realize the prediction of the remaining strength of pipeline welding in service and provide technical support and decision-making basis for engineers.关键词
卷积神经网络/长短期记忆网络/注意力机制/剩余强度预测/模型训练Key words
CNN/LSTM/attention mechanism/residual strength prediction/model training分类
金属材料引用本文复制引用
叶均磊,陆金桂..基于注意力机制增强CNN-LSTM的在役管道焊接剩余强度预测研究[J].焊管,2025,48(3):23-28,6.
