Development of automatic orbital pipe MIG welding system and process parameters' optimization of AISI 1020 mild steel pipe using hybrid artificial neural network and genetic algorithm

Currently, in pipe welding, it is nearly difficult for a human welder to weld the whole circumference of a pipe in a single uninterrupted pass using MIG welding causing inconsistencies in weld quality around the welded pipe. The aim of this study was to develop an automated orbital pipe MIG welding system and to optimize welding parameters for enhancing ultimate tensile strength and Rockwell hardness of mild steel 1020 grade pipe. Three levels of variation were applied to the four input parameters that were chosen. Nine experiments were carried out using orthogonal array of L9. In this experimental investigation, the highest ultimate tensile strength (UTS) of 411.2 MPa and Rockwell hardness (RH) of 95 HRB were achieved at 110 A of current, 24 V of voltage, welding gun travel speed of 30 cm/min, and 3 mm of arc length. For modeling the orbital pipe MIG welding process experimental input parameters and response results, a hybrid artificial neural network and genetic algorithm (ANN-GA) model was constructed. This model was used to forecast and optimize UTS and RH, as well as the process factors that go with it. The results indicated that the ANN-GA model could predict the output responses with a mean square error of 5.06e-05. During optimization, a 4-9-2 network trained with neural network of back propagation by Bayesian regularization approach was determined to have the greatest prediction capability, with maximum UTS and RH of 417.857 MPa and 96.5364 HRB, respectively. From the confirmation tests, the average results of 412.7 MPa of UTS and 95 HRB of HR were obtained. The percentage of errors between the ANN-GA predicted optimal responses' results and the confirmatory experimental results were found 1.23% and 1.59% for UTS and RH, respectively.

Automated summary

Currently, it is difficult for a human welder to weld the entire circumference of a pipe in a single pass using MIG welding, resulting in inconsistencies in weld quality. This study developed an automated orbital pipe MIG welding system and optimized welding parameters to enhance the strength and hardness of mild steel pipes. The results showed that the developed hybrid artificial neural network and genetic algorithm model successfully predicted and optimized the ultimate tensile strength and Rockwell hardness. The model achieved a mean square error of 5.06e-05 and accurately forecasted the optimal responses.