Weld joint forming quality real-time prediction method based on Res-LSTM

• Main Authors: ZHANG CHENG, DING MENGJIA, XIA HAO, JING YAN, CHEN YUSHAN, LI YAO, TIAN ZHEN, ZHANG YINGFENG, KANG CHENGFEI, LI YUNXI
• Format: Patent
• Language: Chinese; English
• Published: 29.03.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; PHYSICS; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

In order to solve the technical problem that an existing method for evaluating and predicting the welding seam forming quality through a neural network is low in prediction precision, the invention provides a real-time prediction method for the welding seam forming quality based on Res-LSTM. According to the method, source domain data and target domain data are constructed by collecting data, feature extraction is performed on the source domain data by using a CNN-based welding seam forming transfer learning model, and convolutional layer parameter correction is performed on a ResNet-based welding seam forming diagnosis model according to the target domain data; building a weld forming quality prediction model based on LSTM (Long Short Term Memory) to predict a weld time sequence feature vector at the next welding moment, inputting the weld time sequence feature vector into a full connection layer of the trained weld forming diagnosis model, and outputting a welding label by the full connection layer of the w