Thermal spraying is a mature surface treatment technology with broad applications in aerospace, energy equipment, and related fields. The thermal spray process involves complex multiphysics coupling, where traditional methods face challenges in parameter optimization, performance prediction, and quality control. In recent years, machine learning has advanced rapidly, offering new solutions for digital and intelligent research in thermal spraying. This paper reviews the applications and research progress of machine learning in thermal spraying. First, the basic principles of supervised learning, unsupervised learning, and reinforcement learning are introduced, along with their applicability to thermal spraying. Second, the current status of machine learning applications is examined in key areas, including composition design and optimization of powder materials, process parameter optimization and control, coating structure and performance prediction, defect detection and quality control, as well as digital twins and robotic autonomous spraying. Additionally, the advantages and limitations of different algorithms are analyzed. Finally, the challenges in current research are summarized, including inadequate data quality and limited model generalization capability. Future development directions are proposed, encompassing multimodal data fusion, interpretable machine learning, and cross-scale modeling.
Gao et al. (Thu,) studied this question.