Data Augmentation–Based Measurement Framework for Predicting the Performance of Electrodeposited Ni-W/SiC Nanocoatings

The limited availability of reference data for electrodeposited Ni-W/SiC nanocoatings poses a major challenge to the indirect estimation of key performance measurands from process parameters. To address this problem, this study develops a data augmentation–based measurement framework for the estimation of coefficient of friction (COF) and hardness under small-sample conditions. A literature-derived multi-source dataset was first established and harmonized, after which controlled data augmentation was introduced using a Regression-Conditional Deep Convolutional Generative Adversarial Network with a dual-branch architecture. To ensure the reliability of the generated data, their quality was evaluated through both statistical and physical assessments. Five Machine Learning (ML) algorithms were then compared as candidate indirect estimators. The results show that the generated data achieved the best overall quality at 2000 training epochs, and that an augmentation level of 1,500 synthetic samples yielded the best predictive performance. Using XGBoost as the final estimator, the framework achieved for COF and for hardness under 5-fold cross-validation, and and , respectively, under Leave-One-Literature-Out Cross-Validation (LOLO-CV). Additional comparison with conventional augmentation-based frameworks and independent laboratory validation further confirmed the practical reliability of the proposed framework. These findings demonstrate that controlled data augmentation can effectively improve the learning of measurement functions under limited-data conditions and provide a useful basis for the indirect estimation of Ni-W/SiC nanocoating performance.