Comparative Analysis for Prediction of Box–Behnken Design Based Surface Roughness for MAFM Finished Aluminium-6063/Silicon Carbide/Boron Carbide (Al-6063/SiC/B4C) Composites Using Nature Inspired Optimization Techniques

Abstract

Abstract Advanced manufacturing materials like silicon carbide, silicon nitride, and boron carbide are challenging to finish using traditional polishing and grinding procedures with accuracy, super finish and few surface flaws like micro-cracks. In recent years, magnetic field assisted manufacturing technologies have emerged as efficient methods for cleaning, deburring and polishing parts that are made of metal and high-tech engineering materials. The present research work considers the MAFM setup for experimental readings by finishing the hybrid Al/SiC/B4C-MMCs. By making use of the magnetic field to envelope the work-piece in abrasive flow machining, it can result in improving or increasing the materials removal rate along with the surface finish. At times, it becomes necessary to predict the generated surface with a high degree of precision and accuracy. This aptly requires the use of regression models for obtaining such exactness. In the present work, the MAFM process is investigated by the use of hybrid ANN approach. Here ANN is employed to model the input–output relations between the various parameters. For this, purpose, a generalized framework is designed using six inputs and two outputs. The input parameters are extrusion pressure (Ep), mesh number (M), concentration of abrasives (C), work-piece material (Wp), number of cycles (N) and magnetic flux density (Mf). In contrast, the output parameters are material removal rate (MRR) and change in surface roughness (ΔRa). The optimizations implemented through the hybrid PSO-GA-SA-PS algorithms have been used to optimize the MAFM process. Material removal has been accomplished using loosely bonded magnetic abrasive medium composed of silicone carbide (SiC) and the experiments have been carried out on a MAFM experimental setup. The main experimentation is designed by Design Expert® (12.0 version) software and a total of 54 runs are planned and implemented based upon a design matrix suggested by the Box-Behnken design (BBD) of response surface methodology (RSM). Incidentally, the comparison shows that trained artificial neural network models have superior prediction abilities as compared to results obtained from BBD model. It is therefore concluded that the composition of the loosely bound SiC-based magnetic abrasive medium plays a significant role in how well Al/SiC/B4C finishes.