Tribo-informative Analysis of Spark Plasma Sintered Al-SiC-HAp Hybrid Composite Through Machine Learning Techniques

This research epitomizes the tribological analysis of the SiC and hydroxyapatite (HAp) reinforced aluminium composite. Hydroxyapatite was extracted from the mrigal fish and used as a secondary reinforcement to enhance the tribological characteristics of the aluminium matrix. A DUCOM TR20LE pin-on-disc tribometer and ECONEMET VH1MD hardness were used to assess the wear and hardness of the synthesized spark plasma sintered Al- 10%SiC- x%HAp (x = 0, 2.5, 5, 7.5, 10 and 12.5 wt%) hybrid composite respectively. X-ray diffraction analysis confirmed the presence of HAp and beta-tricalcium phosphate in the mrigal fish bone extracted HAp powder. The OH-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${OH}<^>{-}$$\end{document}, CO32-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${CO}_{3}<^>{2-}$$\end{document} and PO43-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${PO}_{4}<^>{3-}$$\end{document} bands were identified by conducting Fourier transform infrared spectroscopy which confirms the HAp in the calcinated powder. Hardness results reveal that the composite corresponds to the 7.5 wt% HAp achieved the maximum hardness of 129 VHN. The composite with 2.5 wt% reinforced HAp in the Al- 10wt%SiC- HAp composite exhibits the minimum wear and COF of 0.37 mg and 0.3 respectively for 5 N applied load. To investigate the wear behaviour of the synthesized composite, different machine learning techniques such as KNN, SVM, and ANN have been utilized where the SVM machine learning model significantly predicts the output results for wear (R2 = 0.9354, MAE = 0.0832, MSE = 0.0063 and RSME = 0.0824) and COF (MAE = 0.0192, MSE = 0.0009, RSME = 0.0249 and R2 = 0.9423) with minimum error.