Performance comparison of tree-based and neural network models for wear prediction in coated and uncoated Al6061

This study compares the performance of machine learning (ML) and deep learning (DL) models in predicting the dry sliding wear of uncoated Al6061, plasma-sprayed flyash-Al 2 O 3 and flyash-SiC coatings. Ensemble models, including random forest (RF), XGBoost and LightGBM, along with neural network models such as multilayer perceptron (MLP) regressors, backpropagation neural networks (BPNN) and deep neural networks (DNN), were trained on experimental data that varied load, sliding speed and sliding distance. The dataset was scaled and split into training (80%) and testing (20%) subsets. Model performance was evaluated using the coefficient of determination (R 2 ), mean absolute error (MAE) and root mean square error (RMSE). ML regressors accurately predicted the properties of uncoated alloys, with R 2 scores above 0.97, though their performance decreased on coated samples. RF experienced the largest decline in accuracy, particularly flyash-SiC (R 2 = 0.736). Gradient boosting models exhibited improved robustness, with LightGBM achieving R 2 values of 0.977, 0.936 and 0.794 for uncoated, flyash-Al 2 O 3 and flyash-SiC samples, respectively. Neural networks outperformed tree-based methods for coated systems, with MLP and DNN attaining R 2 values up to 0.992, alongside lower MAE and RMSE. SEM analysis corroborated the predictions, showing severe wear in uncoated alloys, minimal surface damage in flyash-Al 2 O 3 coatings and cracking and delamination in flyash-SiC coatings. • AI-based models predict wear behavior of coated and uncoated Al6061 alloys. • Deep learning shows superior accuracy over ensemble tree-based regressors. • Flyash–Al 2 O 3 plasma coatings provide best resistance to dry sliding wear. • SEM validates reduced damage in Al 2 O 3 coatings versus severe SiC delamination. • Study integrates tribology, AI modeling, and sustainable flyash utilization.