Machine Learning Approach for Corrosion Analysis in Copper–Stainless Steel Welds Produced by Cold Gas Tungsten Arc Welding

Abstract

The corrosion behavior of galvanically coupled copper–AISI 304 stainless-steel welds (Cu-Fe immiscible system) fabricated using the novel cold gas tungsten arc welding (CGTAW) process was investigated. The controlling factors and mechanisms governing corrosion in three types of anode-dominated weld microstructures were discussed. In a contaminated circulating water environment, corrosion rates were associated with the degree of mixing between the Cu-matrix and Fe-based phases (uniform or over-mixed distributions) and with the presence of defects. A hierarchical Random Forest (RF)-based machine learning framework was developed to model the process–structure–property relationship, in which process parameters influence corrosion rate through intermediate microstructural evolution and defect formation.