Study on the Selection of Optimal Welding Process Variables for Heating Coil Pipelines using Real-time Data and Cluster analysis

Most methods for detecting welding defects are manual, whereby workers directly observe the defects occurring during welding. However, these methods are time-consuming and vary greatly in accuracy depending on the worker's judgment ability. Therefore, a method is required for automatic detection of welding defects. However, welding defects vary in size and shape, presenting limitations to automation. Thus, research on applying machine learning methods and deep learning technology, artificial intelligence (AI) techniques in the 4th Industrial Revolution, is urgently required. In this study, high-speed gas metal arc (GMA) welding was performed using STPG 370 steel to select the optimal welding process parameters for the heating coil pipeline, acquiring real-time data for quality analysis of molecular dynamic (MD) welding. In addition, optimal process variables were selected and verified through thermo-elastoplastic analysis.