An artificial intelligence system for quality level-based prediction of welding parameters for robotic gas metal arc welding

This study addresses the gap between laboratory-focused welding process parameter identification and their practical application in the industry. Unlike traditional input-output mapping reported in the literature, determining acceptable input process parameters in industry hinges on the acceptable threshold for imperfections. The research aims to devise an artificial intelligence system capable of forecasting acceptable parameters for robotic gas metal arc welding, aligning with the quality standards outlined in AWS B4.0:2016 and BS EN ISO 5817:2014. Parameters, including wire feed rate, travel speed, contact tip to work distance, and electrode work angle, are considered in the prediction model. Throat thickness and joint penetration are critical responses for weldments involving 2-mm-, 4-mm-, and 6-mm-thick 4130 steel plates. The fuzzy model achieves effective defuzzification by employing fuzzy expert rules, triangular membership functions, and the centroid area method via the MATLAB fuzzy logic toolbox. The models are rigorously validated through experimental work. The study culminates in the acquisition of accurately predicted and experimentally acceptable input process parameters across varying quality levels (B, C, and D).