Neural Simulation of Surface Generated During Magnetic Abrasive Flow Machining of Hybrid Al/SiC/B4C-MMCs

Abstract

Modern manufacturing processes like abrasive flow machining (AFM) are often employed for finishing the conventionally inaccessible surfaces that include the inner surfaces of cylindrical workpieces. This process is used for accurate prediction of the generated surfaces. As such, regression models are useful for such a prediction. In the existing AFM setup, the nylon fixture is replaced by an aluminum fixture, and a modified MAFM setup is fabricated by including the effect of magnetization. The research explores the MAFM process for finishing the hybrid metal matrix composites (MMCs) of SiC/B4C using aluminum-6063 as a base material. The paper employs artificial neural networks to model and simulate the response characteristics during the MAFM process, developed in the MATLAB 2016 b environment. The neural networks are trained for finishing the cylindrical components of Al/SiC/B4C-MMCs. A neural network having a configuration of generalized back-propagation is employed with six inputs, two outputs, and two hidden layers. The experimentally observed data take into account the influence of MAFM process parameters including the number of cycles, extrusion pressure, the concentration of abrasives, magnetic flux density, mesh size, and workpiece material. The process responses are characterized as material removal rate (MRR; μg/s) and change in surface roughness (ΔRa; μm). Moreover, the microstructure analysis of workpiece materials is also done using scanning electron microscope.