机电工程技术2025，Vol.54Issue(12)：33-39,128,8.DOI:10.3969/j.issn.1009-9492.2025.12.003
基于数据-机理驱动的焊接温度场实时重建
Real-time Reconstruction of Temperature Field during Welding by Data-mechanism Driving
摘要
Abstract
The real-time and accurate reconstruction of the welding temperature field is the core challenge for improving the welding quality.Traditional finite element simulations face the problems of low computational efficiency and insufficient dynamic adaptability,while pure data-driven deep learning models have the defect of lacking physical constraints.A real-time reconstruction method of the welding temperature field that integrates the heat transfer mechanism and deep learning is proposed.Through the collaboration of multiple technologies,the efficient modeling and dynamic optimization of the temperature field are achieved.Based on the heat transfer mechanism of pulsed gas tungsten arc welding(GTAW-P),the Abaqus finite element software is used to generate a temperature field dataset containing welding parameters.The surface width is extracted from the real-time molten pool images through image processing technology,and a long short-term memory model(LSTM)is used to establish the dynamic mapping relationship between the molten pool width and the heat source parameters,so as to realize the real-time prediction of the heat source parameters.Furthermore,the DeepONet neural operator model is introduced.The heat conduction partial differential equations(PDEs)are solved through a dual network architecture,and an autoencoder(AE)is used to reduce the dimensionality of high-dimensional data,improving the training efficiency and generalization ability of the model.The experimental results show that the root mean square error(RMSE)of the proposed method in heat source parameter prediction is lower than 0.11,the mean absolute error(MAE)of the temperature field reconstruction is 1.096 K,and the inference time is only 0.11 ms,achieving nearly real-time calculation compared with the traditional finite element method.This method breaks through the computational bottleneck of traditional numerical simulations,establishes a temperature field modeling framework that integrates data-driven and physical constraints,and provides a new paradigm for the real-time optimization of welding processes.关键词
焊接温度场/DeepONet神经算子/LSTM/物理信息深度学习Key words
welding temperature field/DeepONet/long short-term memory(LSTM)/physics-informed deep learning分类
矿业与冶金引用本文复制引用
焦文华,常英杰,董鑫源..基于数据-机理驱动的焊接温度场实时重建[J].机电工程技术,2025,54(12):33-39,128,8.基金项目
国家自然科学基金青年基金(52305375) （52305375）
