Optimization of MIG Welding Productivity with the Aid of Artificial Neural Network

Abstract

Automated MIG welding has huge industrial application especially In ship building where long lengths of radiographic quality welding is necessary. This constant demand for high quality welding on varying base metal dimensions without sacrificing productivity is something that have posed a problem for these industries since the advent of automated welding machines. Welding parameter optimization for a given application is mostly done using a series of trial runs using a variety of power settings and the most acceptable setting is taken after adequate metallurgical examination of the welds. However this is a time-consuming and tedious process especially in industrial establishments. Deterministic relations between the several welding parameters are hard to come by and they are not accurate in the wide range of welding demands that the industries call for. Artificial Neural Network can provide a possible solution to this problem by predicting optimized variables.

A large number of trial runs have been made using the MAILAM ® MGA 40 flux cored electrode on a large number of mild steel plates of varying thickness at the Welding Technology Centre, Jadavpur University. Voltage, torch travel speed, wire feed rate and current were some of the variables that were taken into account. The developed network optimized the values of Wire feed rate and current drawn from supplied data which include torch travel speed, voltage and plate thickness. Apart from these welding parameters another aspect of industrial large scale welding that has been taken into account is operator productivity. Most welding productivity optimization studies so far has not taken into account the actual time study of welding. Since operator activities like fettling, interpass cooling etc. has a great bearing on welding productivity, the present study aims at giving a more complete picture of welding productivity optimization.