Erosion wear performance of titania filled ramie-epoxy composites: A data driven optimization study using supervised machine learning approach

This paper reports on a data driven machine learning (ML) approach to analyze and predict the erosion behavior of titanium oxide (titania) filled ramie-epoxy composites. ML models are extensively used in recent years to mimic human decisions in various industries. After fabrication and well-designed erosion trials following design of experiments, the experimental data is critically analyzed to examine the effect of each input factor (erodent temperature, striking angle, striking velocity and filler content) on the output that is erosion wear rate. It is found that the erosion wear rate increases with increase in striking velocity and striking angle and decreases with increase in filler content. The experimental data is further used to feed five different ML models. The performance and adequacy of ML models are compared using five different performance metrics. It is noticed that although all the machine learning techniques effectively predicted the erosion rate, the Gradient boosting machine (GBM) model exhibited superior performance with an R 2 value of 0.9486. The feature importance plot confirms that the, filler content, striking velocity and the striking angle played a major role in predicting the erosion rate of the hybrid composites.