Publication

Abstract : Gas metal arc welding (GMAW) is widely utilized for depositing corrosion-resistant austenitic stainless steel claddings on low-carbon steel substrates, where the bead geometry directly influences the structural integrity and service life of the component. The objective of this study is to develop a predictive framework for accurately estimating clad bead geometry parameters, namely bead width, penetration depth, reinforcement height, and percentage dilution, based on key GMAW process variables. To achieve this, five supervised machine learning (ML) models—linear regression (LR), K-Nearest Neighbors (KNN), decision tree (DT), random forest (RF), and support vector regression (SVR)— were trained on experimentally obtained datasets and evaluated using performance metrics including the R² score, the mean absolute error (MAE), the mean squared error (MSE), and the root mean squared error (RMSE). Among the models, the DT demonstrated the best predictive performance, achieving an R² score of 0.959, an MAE of 0.134, an MSE of 0.150, and an RMSE of 0.388. The SVR model also performed exceptionally well, with an R² score of 0.952. This study identified the welding gun angle and wire feed rate as the most influential parameters affecting clad bead geometry. The use of these advanced ML models considerably improves the prediction accuracy of clad bead dimensions in GMAW, enabling intelligent process optimization and consistent production of high-quality weld cladding.