Experimental Investigation and Machine Learning Modeling of Wear Characteristics of AZ91 Composites

Abstract

This study's primary goal is to examine the effects of wear parameters on the wear-rate (WR) of magnesium (AZ91) composites. The composites are made up of using a stir casting process with aluminum oxide (Al2O3) and graphene as reinforcements. In the present work, one material factor (material type (MT)) and three tribological factors (load(L), velocity (V), and sliding distance (D)) were chosen to study their influence on the wear-rate. Taguchi technique is employed for the design of experiments, and it was observed that load (L) is the most influencing parameter on WR, followed by MT, D, and V. The optimal values of influencing parameters for WR are as follows: MT = T2, L = 10 N, V = 2 m/s, and D = 500 m. The wear mechanisms at the highest and lowest WR conditions were also studied by observing their scanning electron micrographs (SEM) on wear pin’s surface and its debris. From the SEM analysis, it was observed that abrasion, delamination, adhesion, and oxidation mechanisms were exhibited on the wear surface. Machine learning (ML) models such as artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), and decision tree (DT) were used to develop an effective prediction model to predict the output responses at the corresponding input variables. Confirmation tests were conducted under optimal conditions, and the same were examined with the results of ANN, ANFIS and DT. It was noticed that the DT model exhibited higher accuracy when compared to other models considered in this study.