Analysis and optimization of advance gas metal arc welding process parameters for controlled dilution using computational models

The work presented here is based upon an experimental study which examined the influences of Gas Metal Arc Welding (GMAW) and Advanced Gas Metal Arc Welding (A-GMAW) process parameters in the dilution of single layer weldments formed using AISI 1023 steel. Five factors at five levels were used in conjunction with a Central Composite Design (CCD) to produce a predictive model of dilution. The individual and interaction effect of all five process parameters: welding current; welding voltage; welding speed; nozzle-to-plate distance; and gas flow rate/powder feed rate (gas flow for GMAW and powder for A-GMAW) in addition to a categorical variable of preheat current has been measured. The results of this comparative study clearly show a significant reduction (maximum up to 32.7%) in dilution when utilizing the A-GMAW method compared to the GMAW method. These results are due to two primary reasons: first, the use of preheated filler wire provides additional arc energy through resistive heating in A-GMAW. Second, the powder feeding mechanism in the A-GMAW process utilizes excess energy from the arc and therefore reduces the amount of dilution. The mathematical models developed based on the experimental results show that dilution exhibits a direct linear relationship with welding current. The results show that, in the A-GMAW process, welding current is the most significant parameter affecting dilution, followed by powder feed rate and welding speed. However, an inverse relationship exhibits between the dilution and both the nozzle-to-plate distance and welding voltage. An optimization of the input parameters was also performed using a single objective methodology incorporating the use of the Artificial Neural Network (ANN) and Genetic Algorithms (GA). The results of the optimization procedures showed that the ANN algorithm had better results than the Response Surface Methodology (RSM), whereas the GA had the least favored results concerning optimization.