Machine Learning Algorithms to Visualize the Weld Quality and Evaluation of Health Issues during the Welding

Abstract

This research explores the potential of machine learning methods to use in the welding sector. Welding operations encounter several challenges that might use machine learning approaches. Welding processes have been found to benefit greatly from the use of machine learning algorithms, which increase both productivity and quality. Industrial robots powered by artificial intelligence may assist in addressing many difficult problems in the manufacturing industry. Many welding procedures depend on human knowledge to choose optimal settings, making them less efficient and more prone to human mistakes. To lessen this reliance, neural networks educate robots and autonomous systems to provide consistent weld quality and increased efficiency. Weld quality is mainly determined by visual examination. As a result, machine learning is also used to provide a visual representation of the welding process. In two dimensions, 12 machine learning pipelines are built: settings for feature engineering and ML techniques. Four feature settings and three machine learning methods were considered. Using data from two active industrial production lines containing 25 welding machines, a machine-learning pipeline was tested. These approaches may also be used to analyze the components using regression analysis to determine which factors relate to certain health concerns.