The comparative evaluation of the wear behavior of epoxy matrix hybrid nano-composites via experiments and machine learning models

• Published in: Tribology international 2025-04, Vol.204, Article 110451
• Main Authors: Aydın, Fatih, Karaoğlan, Kürşat Mustafa, Pektürk, Hatice Yakut, Demir, Bilge, Karakurt, Volkan, Ahlatçı, Hayrettin
• Format: Article
• Language: English
• Subjects: Carbon fiber; Machine learning; MWCNT; Quadriaxial non-crimp fabric; Wear behavior; Wear loss prediction
• ISSN: 0301-679X
• DOI: 10.1016/j.triboint.2024.110451

This study evaluated the wear behavior of multiwall carbon nanotube (MWCNT) doped non-crimp fabric carbon fiber reinforced polymer (NCF-CFRP) composites produced through vacuum infusion. Compared to 0 wt% MWCNT reinforced composite, the wear loss of 1 wt% MWCNT reinforced composite under loads of 10 N and 30 N decreased by 48.1 % and 61.1 %, respectively, for sliding distance of 1000 m. Additionally, the study evaluated various Machine Learning models including Deep Multi-Layer Perceptron (DMLP), Random Forest Regression, Gradient Boosting Regression, Linear Regression (LR), and Polynomial Regression for predicting wear loss. The DMLP model exhibited enhanced predictive capabilities in the testing phase (R²=0.9726) compared to its training performance (R²=0.9531), while the LR model maintained stable performance characteristics between training (R²=0.9712) and testing (R²=0.9454) phases. •MWCNT doped non-crimp fabric carbon fiber reinforced composites were produced.•The 1 wt% MWCNT reinforced composite exhibited the best wear performance.•Matrix delamination and carbon fiber fractures were observed under load of 30 N.•Different machine learning models were used to predict the wear performance.