Optimization based on artificial intelligence for thermal barrier coating with shadow effects using electron beam physical vapor deposition

Abstract

Thermal barrier coatings for gas turbine engine cooling blades require precise thickness control during electron beam physical vapor deposition (EB-PVD), and achieving uniform coating distribution is critical for aerospace applications. However, shadow effects during EB-PVD create significant thickness variations that compromise coating quality and performance reliability. Therefore, this study presents a comprehensive computational formulation that integrates three key models: thermal behavior of the ingot within the EB-PVD machine, coating deposition modeling of the vapor plume, and substrate-manipulator kinematics. Virtual ingots were employed to specifically account for shadow effects, while specimen tests provided experimental coefficients that were incorporated into the formulation. The approach utilizes coating deposition characteristics related to manipulator input profiles to train a computationally efficient multi-layer perceptron (MLP). Using the trained MLP, optimization was performed to minimize thermal barrier coating thickness variation on cooling blades. The results demonstrate that this integrated formulation successfully addresses shadow-induced thickness variations, contributing to a comprehensive database of coating deposition characteristics for thermal barrier coating applications in gas turbine engines. Consequently, the proposed formulation offers a straightforward and computationally efficient solution for optimizing EB-PVD processes, enabling improved coating uniformity and reliability for aerospace thermal barrier coating applications.