Prediction of friction coefficient of su-8 and its composite coatings using machine learning techniques

Machine learning (ML) techniques are used to predict the coefficient of friction of an epoxy polymer resin (SU-8) and its composite coatings deposited on a silicon wafer. Filler type and the number of cycles are taken as the input parameters. The filler types included, two solid fillers namely, graphite and talc, and a liquid filler such as Perfluoropolyether (PFPE). Six variations of the SU8 coatings were developed based on the different combinations of filers used and tested. The experimental data generated for these different coatings for varying number of cycles (0 to 499) was used to train the different ML algorithms like ANN, SVM, CART, and RF to predict the coefficient of friction. The performance of these ML techniques was compared by calculating mean absolute error (MAE), root means square error (RMSE), and square of the correlation coefficient (R²). The ANN algorithm was observed to have the best (R²) metrics while the other ML techniques SVM, CART, and RF had a satisfactory performance with some inaccuracies seen for the CART algorithm for the data set under consideration.