High-performance functional coatings, owing to their excellent wear resistance, corrosion resistance, high-temperature tolerance, and catalytic properties, hold significant application value in strategic fields such as aerospace, marine engineering equipment, and green energy. Thermal spraying technology enables the efficient deposition of ceramic, metallic, and composite coatings; however, the deposition process involves multiphysics coupling and complex physicochemical transformations, resulting in a nonlinear, coupled interdependence between process parameters and coating properties that conventional trial-and-error approaches struggle to control precisely. To address these challenges, machine learning methods were introduced into the thermal spraying process, yielding the following main results: 1) the contribution weights of thermal spray process parameters to in-flight particle characteristics and coating properties were quantitatively resolved; 2) a "map-guided" forward prediction model linking process parameters to coating properties was established; and 3) a "target-oriented" inverse design of process parameters based on desired coating properties was realized for the first time. This study not only deepens the understanding of thermal spraying mechanisms but also provides a viable pathway toward the controllable deposition of coatings, thereby promoting the engineering application of functional coatings in key fields.

thermal spraying,machine learning,functional coatings,prediction of coatings,optimization of coatings