Application of Machine Learning for Optimization of HVOF Process Parameters

Abstract

Abstract A variety of process parameters affect the properties of the deposited coatings in the High Velocity Oxygen Fuel (HVOF) spraying process. In fact, the quality of coatings can be improved without changing feedstock or deposition technology by the application of optimized spraying process parameters. In this study, a large set of data “Big Data” is used to create a variety of machine learning models for prediction of porosity content and hardness values of HVOF deposited coatings. A set of process parameters was selected as validation run and actual HVOF coating was deposited using those parameters. The porosity level and hardness were measured and compared to those predicted by models. The models differ based on the number of neurons utilized in each layer for the calculations. A model with six neurons could predict closest porosity level and the one with three was the best in prediction of hardness. The final model could be obtained by running data through both models. Through this study, a robust machine learning model for the optimization of HVOF process parameters will be developed that could be used for other coatings and thermal spraying techniques.