Surface Aware Triboinformatics Framework for Wear Prediction of MWCNT Reinforced Epoxy Composites Using Run-Wise AFM Descriptors and Machine Learning

Accurate prediction of wear behavior in polymer nanocomposites remains challenging due to the coupled influence of operating conditions and evolving surface morphology. In this study, a surface-aware triboinformatics framework is proposed to predict the dry sliding wear behavior of multi-walled carbon nanotube (MWCNT) reinforced epoxy composites by integrating operating parameters with run-wise atomic force microscopy (AFM) surface descriptors. Wear experiments were conducted using a Taguchi L16 design by varying CNT content (0–0.75 wt.%), applied load (10–40 N), sliding speed (183–458 rpm), and sliding distance (500–1250 m). AFM-derived parameters, including Ra, Rq, Z-range, and surface area difference, were extracted from the worn surface corresponding to each experimental run. Multiple regression-based machine learning models were evaluated using leave-one-out cross-validation, with ensemble-based models providing the best predictive performance (R2 > 0.85 with low RMSE and MAE). Feature importance and partial dependence analyses identified CNT content as the dominant factor controlling wear reduction, followed by Z-range and Ra, highlighting the critical role of surface damage severity. Neat epoxy exhibited a maximum wear loss of 0.444 mg, whereas the 0.75 wt.% CNT composite showed values as low as 0.003 mg under comparable conditions, corresponding to a reduction of approximately 99%. The proposed framework enables mechanistically interpretable wear prediction and supports the design of durable polymer composites, contributing to SDG 9 (Industry, Innovation and Infrastructure) and SDG 12 (Responsible Consumption and Production).