Machine learning predictive modeling of plasma-sprayed lanthanum zirconate coatings

Abstract

Thermal barrier coatings play an indispensable role in the protection of nickel-based superalloy components used in gas turbines. Lanthanum zirconate has emerged as a prosperous candidate material to substitute for the present yttria-stabilized zirconia topcoat material. This study examines the influence of plasma spray parameters on the porosity of the lanthanum zirconate coating and its Young's modulus. Response Surface Methodology has been utilized to evaluate the combined effect of plasma spray parameters on coatings microstructure and mechanical properties. Further, mathematical models have been developed and statistically validated through an analysis of variance technique. Two-dimensional contour and three-dimensional surface plots demonstrated the importance of interactions between parameters. Furthermore, machine learning techniques, such as support vector machines with radial basis function and random forest regression, have been utilized for the prediction of Young's modulus based on porosity. SVM-RBF exhibited higher predictive accuracy (R² = 0.9615), whereas RFR offered interpretability and robustness for real-time use. The integration of RSM, ML and visualization tools proves effective in developing high-performance TBCs for aerospace and gas turbine applications.