Taguchi approach and decision tree algorithm for prediction of wear rate in zinc oxide-filled AA7075 matrix composites

Abstract

This work aims to apply data-driven approach [decision tree (DT) algorithm] to analyse the wear rate (WR) of ZnO-filled AA7075 composites. The results of model-based analysis was compared with Taguchi analysis. Stir casting was used to produce the composite samples. Characterization studies were conducted to analyse the composition and morphology. The scanning electron microscopy results indicated the even dispersion of ZnO in the AA7075. The energy-dispersive x-ray spectroscopy pattern ensures the presence of matrix elements and the inclusion of reinforcement particles into the proposed composites. To minimize the number of experimentation, L27 Orthogonal array is used for finding WR. The ‘DuCom’ Pin-on-Disc apparatus were used to prepare WR data for the set of the proposed composites. Taguchi technique reveals the optimum level factors for obtaining the minimum ‘WR’ is reinforcement content of 10 wt.%, applied load (P) at 10 N, sliding velocity (V) at 1 m s−1 and sliding distance (D) of 1000 m. The experiments results from DT algorithm, and analysis of variance and signal-to-noise ratio analysis from Taguchi-based approach confirmed that reinforcement is the primary element for affecting wear of the composites. The reason for applying DT algorithm is that, the low-level knowledge could be converted into high-level knowledge (If-then-else rules), which can be effortlessly explicable by semiskilled personnel.