Identification of SMAW Surface Weld Defects Using Machine Learning

Published in Recent Advances in Materials Processing and Characterization, 2022

This paper presents a machine learning approach for identifying surface defects in shielded metal arc welding (SMAW) using high-resolution images. Applied various image processing techniques to extract geometrical features of welds, then built CNN and ResNet50 models to classify acceptable beads and surface defects, achieving over 98% accuracy for both models.

Recommended citation: Ramesh, K., E. V. Ramana, L. Srikanth, C. Sri Harsha, and N. Kiran Kumar. "Identification of SMAW Surface Weld Defects Using Machine Learning." In Recent Advances in Materials Processing and Characterization: Select Proceedings of ICMPC 2021, pp. 339-350. Singapore: Springer Nature Singapore, 2022.
Download Paper