Towards cold spray coatings optimization via artificial intelligence

The main goal when studying a novel powder material / substrate coating system is to optimize its application in terms of the resulting properties and the overall process efficiency. Several approaches for the cold spray process optimization have been used in the past based on DOE - Design of Experiments - methods offering a systematic exploration of the spraying parameters combinations domain. In this work an optimization approach based on a model based reinforcement learning method is presented. For the cold spray technology reinforcement learning is used to guide researchers in the selection of the proper spraying parameters combination that will maximize a single or several coating properties and the overall process efficiency. Two case studies are presented showing the optimization of Copper and Aluminum coatings where the optimization task converges after few spraying operations. As a result, the reinforcement learning methodology offers a map that describes the powder material response to the most possible spraying parameters combinations known as the "Material Fingerprint". The implementation of the reinforcement learning methodology for cold spray coatings optimization can potentially influence the successful development of new powder material / substrate coating systems and support the standardization of the cold spray technology.