Machine Learning Optimization of IN625 Coating Properties in Cold Spray Process

Cold spray technology presents a promising solution for repairing components in the nuclear power industry due to its low processing temperature, which minimizes thermal effects on the substrate. This advantage allows for the deposition of thick parts with strong interlayer bonding, making it suitable for robust repairs. In this study, we focus on developing effective repair procedures by optimizing key processing parameters such as gas pressure, gas temperature, and traverse speed. To achieve this, a combination of machine learning and experimental testing was employed on cold-sprayed Ni-based superalloy (IN625). The prepared samples were assessed for microhardness, adhesion strength, and porosity, and these experimental results were subsequently used to train a machine learning model. This model predicts material properties under varying process conditions, ensuring precision in parameter selection. Ultimately, optimized parameters were used to deposit IN625 for the successful repair of cracked specimens, demonstrating the practical applicability of the developed methodology.