Cold Spray is a solid-state deposition technique that accelerates metallic particles to supersonic speed through a converging-diverging nozzle. Particles adhere to the substrate thanks to the high kinetic energy they acquire when reaching a material-specific critical velocity. The optimal choice of process parameters — essential to obtain suitable particle temperatures and velocities — is crucial for deposition efficiency. In the present study, a computational fluid dynamic (CFD) model was developed to simulate the cold spraying process. The model was validated experimentally through a high-speed camera for in-flight particle tracking. The simulations were repeated on a wide range of process parameters and on different substrate geometries, applying Latin Hypercube sampling to ensure a homogeneous distribution within the variable space. The generated data was used to train an artificial intelligence (AI) model of Support Vector Regression (SVR) with the objective of directly predicting the thermo-kinetic properties of the metallic powders. To strengthen the interpretability of the prediction model, explainable AI method of SHapley Additive exPlanations (SHAP) was implemented to identify how each input parameter affects the model predictions for particle temperatures and velocities. The combined CFD-AI approach showed high accuracy and efficiency in predicting the thermo-kinetic conditions of the powder while maintaining the physical interpretability of the related phenomena. This integrated method enables advanced optimization strategies for controlling the Cold Spray process.
Falco et al. (Wed,) studied this question.