Multi‐source data‐driven model to predict the deposition rate of CVD‐SiC coatings with improved accuracy and robustness

Abstract Chemical vapor deposition (CVD) is widely used for the preparation of silicon carbide (SiC) coatings. The growth of controllable SiC coatings at desired efficiencies is of critical importance in industrial production. In this study, we developed a multi‐source data‐driven model to predict the deposition rate of CVD‐SiC coatings. To address the challenge of missing values in the training data, we introduced a simulation‐based filling (SBF) method. The correlation between parameters and the deposition rate was analyzed using machine learning (ML) techniques. Our results demonstrate that leveraging a multi‐source dataset reduces the prediction error from 12% to 3% and improves the generalizability beyond boundary conditions. Comprehensive Sobol sensitivity analysis quantitatively confirmed temperature as the dominant parameter, accounting for 63% of output variance directly and 78% when including interactions. The incorporation of multi‐source dataset improves the accuracy and robustness of the ML model, which also is beneficial in optimization efficiency. The framework presented here highlights the potential of data‐driven approaches for guiding the CVD‐SiC process, offering a novel perspective for future research.