Predicting the Effects of Process Parameters on Weldment Characteristics in MIG Welding using Artificial Neural Networks

Abstract

In this project work an attempt has been taken to predicting the effects of process parameters on weldment characteristics in MIG welding with the help of artificial neural network technique. Electrode wire diameter, Electrode wire feed rate, Welding speed, Welding current and Arc length have been chosen as influential process parameters. More or less, these are the influential factors in deciding the weldment characteristics. Weldment characteristics like Bead Geometry, Depth of Penetration, Depth of Heat Affected Zone (HAZ) and Hardness of weld metal are important characteristics on the basis of structure and these have been considered in this project work. Metal Inert Gas (MIG) Welding process with automatic or robotic system in various industries is a demanding welding process now-a-days and this process is being used with increasing rate of applications. Due to these reasons, MIG welding process has been chosen for this project work and a semi-automatic MIG welding machine have been used. Single straight beads have been welded on the surface of the specimens of the medium carbon steel plate. Artificial Neural Network (ANN) refers to computing systems whose central theme is borrowed from the analogy of biological neural networks and the basic unit of such networks is a simple mathematical model. In this research work the Real-Time Recurrent Learning algorithm of ANN have been used with actual inputs and outputs of experimental values as inputs to the algorithm to complete the computational tasks. It has been observed that the computational values of weldment characteristics obtained by ANN are very close to the experimental values of those. So the ANN based approach can be used effectively for predicting the weldment characteristics in MIG welding.