Machine Learning Model to Predict Optimal Welding Parameters for Dissimilar Welding Between SS316L and SS430

• Version
• Download 8
• File Size 0.00 KB
• File Count 1
• Create Date 21 April, 2026
• Last Updated 21 April, 2026

Machine Learning Model to Predict Optimal Welding Parameters for Dissimilar Welding Between SS316L and SS430

Suhani Kanoje, Krushnakant Gobade, Pragati Meshram, Vishal Ganthade, Prof. Chetan Tembhurkar

DOI :  10.46335/IJIES.2026.11.4.4

Abstract – Dissimilar welding of stainless steels is widely used to combine the advantages of different materials; however, it introduces challenges due to differences in thermal and metallurgical properties. These differences often lead to defects such as poor penetration, reduced mechanical strength, and non-uniform microstructure. The present study aims to improve weld quality by developing a machine learning–based approach for predicting and optimizing Metal Inert Gas (MIG) welding parameters. Experimental work is carried out on SS316L and SS430 stainless steel plates using ER309L filler wire. Key welding parameters, including voltage, current, travel speed, and wire feed rate, are varied systematically. The welded specimens are tested to evaluate tensile strength using a Universal Testing Machine, and penetration depth is measured to assess weld quality. The collected data is used to develop a Random Forest regression model capable of predicting multiple output responses. The results indicate that the model can accurately estimate tensile strength and penetration depth while identifying an optimal heat input range of 0.25–0.35 kJ/mm for improved weld performance. A Streamlit-based interface is also developed to provide real-time predictions and assist in parameter selection. The study demonstrates that the use of machine learning reduces experimental effort, improves consistency, and enhances the overall efficiency of the welding process.