Modelling and optimization of weld bead geometry in robotic gas metal arc-based additive manufacturing using machine learning, finite-element modelling and graph theory and matrix approach

The objective of this study is to investigate effects of the welding speed, wire feed speed, and torch angle on the weld geometry, including height, width, and depth of metal deposition, in additive manufacturing of mild steel. In the present study, artificial neural network was developed to predict weld bead geometry and validate the optimization of process parameters to improve quality of weld bead geometry. Experimental results for the width, depth, and height of the weld bead geometry were collected, and the interaction effect of the process parameters on the weld bead geometry was investigated. Three-dimensional finite-element modelling was performed for the AM, and the width, depth, and height of the weld geometry were predicted. The Taguchi method-based graph theory and matrix approach and the utility concept were used to optimise the process parameters for achieving the dimensional accuracy in AM. The optimal working condition was as follows: a torch angle of 60°, a wire feed speed of 6 m/min, and a welding speed of 0.4 m/min. Under the optimal working conditions, the height, width, and depth of the weld bead were 3.910, 7.615, and 2.000 mm, respectively. The optimization was also validated with ANN and a comparison among the ANN, simulation and experimental results revealed good agreement.