Process Parameter Sensitivity in Magnetic Pulse Welding: An Artificial Neural Network approach

Abstract

Magnetic pulse welding (MPW), a solid-state impact welding technique provides the ability to join a wide array of material combinations, whilst introducing little to no heat to the system and preserving the base metal microstructure. Impact velocity is one of the key criteria which determines the weldability of the joint during MPW. Experimental measurement of impact velocity in MPW across wide-ranging parameters is expensive and time-consuming. Therefore, guidelines for process selection and knowledge of relative influence of parameters on impact velocity is limited. This study presents the applicability of coupling finite element method (FEM) and artificial neural network (ANN) modelling to perform sensitivity analysis of MPW. The welding process was simulated using FEM, and multilayer modular feedforward networks based on the results from finite element simulations were developed. The results of the present study revealed that the coil cross sectional area and turns primarily governed the process, followed by the voltage. The relative sensitivity of the parameters remained independent of the material combination. Inclusion of shop floor applicable process parameters suggests that the developed ANN models can substantially narrow down experimental runs and simultaneously act as a decision support tool for end users.