Machine Learning Methods and Techniques for Predicting Weld Defects

Abstract

Welding is a critical process in various industries, and ensuring the quality of welded joints is essential for maintaining structural integrity. However, most of the industrial failures have been weld related. Despite utmost care in design, selection of materials, and fabrication route, degradation of fabricated components can occur in service, leading to failures. In this issue, machine learning techniques can be applied to predict weld defects and improve the overall quality of welded joints. Specifically, this chapter reviews the use of signal detection equipment and machine learning algorithms for real-time monitoring of welding processes. After obtaining images of the welded joints and processing, statistical features of the images are extracted. Furthermore, this chapter discusses the applications of novel models based on learning and the automation methods that have been developed. Machine learning and automation advancements hold great promise for weld quality management.