Parametric Analysis of ANFIS, ANFIS-PSO, and ANFIS-GA Models for the Prediction of Aluminum Surface Roughness in End-Milling Operation

Abstract

Milling is one of the old and common cutting processes that utilize rotating tools to take materials off the main component with a combination of tools and workpiece movements. The texture of a machined surface is a key factor in defining how an essential component interacts with its environment. Trial-and-error machining to produce high-quality surfaces has been a time-consuming method that yields lower production and poor revenue. In this paper, the performances of an Adaptive Network-based Fuzzy Inference System (ANFIS) model has been employed for the prediction of the surface roughness (SR) of a block of Aluminum alloy AI6061 machined on an end-mill CNC machine by varying four input settings namely: The spindle speed of rotation, the tool cutting rate, the radial depth, and the axial depth. The approach consisted of a parametric analysis carried out within each system to obtain the finest models for the prediction. The hybrids ANFIS-PSO and ANFIS-GA have been employed to find out which one, either PSO or GA, optimizes better ANFIS for the prediction of Al6061 SR. Their performances produced better results than the stand-alone ANFIS, with ANFIS-GA yielding the best results of the most negligible RMSE value of 0.01097 and the regression values of 0.9939 for training and 0.8102 for testing.