Optimizing parameter selection for machine learning models to predict coating characteristics in cold spray processes on polymeric materials

Cold spray (CS) is an advanced manufacturing technique used to deposit metallic layers onto various materials. It works by accelerating metallic particles to supersonic speeds with pressurized gas, enabling adhesion through mechanical interlocking upon impact with the substrate. The integration of Artificial Intelligence (AI) can significantly improve process understanding and enhance the quality of this additive manufacturing method. This work employs Finite Element Method simulations combined with an additional parameter, density, to improve the predictive accuracy of the cold spray process. To ensure consistency and comparability, Z-normalization was applied to standardize the data. Among the tested approaches, the best results for flattening and penetration depth output parameters (RMSE of 0.26 and 0.27, respectively) are achieved by Gaussian Process Regression model with a Matern 5/2 kernel applying the Z-normalization