Machine Learning-Based Erosion Behavior of Silicon Carbide Reinforced Polymer Composites

Abstract

The machine learning methodology is gaining immense exposure as a potential methodology for solving and modelling the machining behaviour of advanced materials. The present paper deals with the application of machining learning approach in analyzing and predicting the effect of reinforced silicon carbide (SiC) particle size on the erosion behaviour of silicon carbide reinforced polymer composites. L₂₇ orthogonal array was designed based on Taguchi’s methodology to execute the experiments. Support vector machine (SVM) and multi-linear regression (MLR) approach were coupled with Taguchi’s methodology to validate obtained optimized response characteristics. These machine learning-based SVM and MLR models are adopted to analyze the absurdity among obtained experimental results and predicted response. Out of 27 experimental runs based on experimental design, 19 experimental runs were selected for training models whereas 08 models were selected for the testing phase. Impingement angle, workpiece reinforcement, standoff distance and slurry pressure were used as input process parameters, whereas material loss was observed as response characteristics. The kernel functions, i.e. Pearson VII based universal kernel (PUK) and radial based function (RBF) kernel were used with machine learning models to obtain the best performing machine learning approach in predicting erosion behaviour of polymer composites.