Parametric Study and Optimization for Welding Processes Using Machine Learning

Abstract

Optimization facilitates in attainment of maximum strength, efficiency, reliability, productivity and longevity. In this work, data from three material joining processes - Ultrasonic welding of polymers, arc welding as Metal Inert Gas and Tungsten Inert Gas are analysed for establishing quantitative relationship between the process parameters and for prediction of weld features using Multivariate Linear Regression algorithm. The various dependency coefficients and characteristics generated with the ML algorithms are in agreement with the inherent dependency as obtained from experimental data and simulation results. This investigation is a preliminary attempt with a limited set of data to manifest the suitability of machine learning techniques; nevertheless, the results are far from conclusive owing to small data set and hence may be extended to precisely model joining processes with higher number of process parameters, degree of freedom and responses.

Passive machine vision-based defect classification in tungsten inert gas welding on SS304 using AI-based gradient descent algorithm

“…It investigates the weld tensile strength and hardness in three zones: weld metal, HAZ and base metal. 6 Finally, inferred that the hardness and temperature of the base metal and weld zones are heavily influenced by welding speed and current. The strong correlation of weld tensile strength is influenced by the welding current in a greater ratio rather than welding speed.…”