Integrating Machine Learning Approach into Plasma-Sprayed Lanthanum Zirconate Coatings for Thermal Barrier Applications
摘要
Thermal barrier coatings are protective systems used to increase the efficiency and lifespan of gas turbines. Lanthanum zirconate is a promising TBC material owing to its exceptional thermal stability, low thermal conductivity, and high melting point. This study presents an integrated experimental and machine learning methodology aimed to enhance the performance of plasma-sprayed lanthanum zirconate coatings in advanced thermal barrier applications. The primary objective of this study was to establish quantitative correlations between powder processing parameters, coating porosity and deposition efficiency. Crushed and milled thermal spray grade lanthanum zirconate powder was prepared using solid-state sintering process at 1600 °C for 12 h, which produced a particle size distribution range of D10 = 12 µm, D50 = 34 µm and D90 = 68 µm. Plasma spraying conducted at 25–35 kW demonstrated that coatings developed from the 12-hour sintered, 4-hour milled powder at 30 kW plasma power achieved a deposition efficiency of 63.9% and a porosity of 4.1 ± 0.3%. The ML models, particularly the random forest and XGBoost regressors, demonstrated high predictive accuracy of R2 = 0.98 and RMSE = 2.1% in correlating the plasma spray process parameters with the coating microstructure. Therefore, it was concluded that the integrated ML strategy provides a robust data-driven method for the predictive optimization and quality control of LZ-based plasma-sprayed thermal barrier coatings.