A multi-criteria decision-making and machine learning approach to predict the tribological behavior of microwave-sintered Ti-0.8Ni–0.3Mo/X wt.% TiB composites

This paper investigates the tribological behavior of microwave-sintered Ti-0.8Ni–0.3Mo/X wt.% TiB composites using an integrated Multi-Criteria Decision-Making (MCDM) and Machine Learning (ML) approach. It explores the relationship between composition and tribological properties. Criteria Importance Through Intercriteria Correlation (CRITIC) and entropy methods were used to determine criteria weights, revealing wear rate as more influential than friction. Weighted Aggregated Sum Product Assessment (WASPAS) ranked alternatives based on input factors, and various ML algorithms—Decision Tree, Random Forest, Linear Regression, and AdaBoost—were trained to predict the MCDM-derived closeness coefficient. Among these, AdaBoost showed superior performance. The study confirms the accuracy of ML models in predicting wear and friction under varying conditions, highlighting their reliability in optimizing composite performance. This approach effectively supports tribological evaluation and materials design for engineering applications.