The quality assessment and prediction becomes one of the most critical requirements\nfor improving reliability, efficiency and safety of laser welding.\nAccurate and efficient model to perform non-destructive quality estimation is\nan essential part of this assessment. This paper presents a structured and comprehensive\napproach developed to design an effective artificial neural network\nbased model for weld bead geometry prediction and control in laser welding of\ngalvanized steel in butt joint configurations. The proposed approach examines\nlaser welding parameters and conditions known to have an influence on geometric\ncharacteristics of the welds and builds a weld quality prediction model\nstep by step. The modelling procedure begins by examining, through structured\nexperimental investigations and exhaustive 3D modelling and simulation efforts,\nthe direct and the interaction effects of laser welding parameters such as\nlaser power, welding speed, fibre diameter and gap, on the weld bead geometry\n(i.e. depth of penetration and bead width). Using these results and various statistical\ntools, various neural network based prediction models are developed and\nevaluated. The results demonstrate that the proposed approach can effectively\nlead to a consistent model able to accurately and reliably provide an appropriate\nprediction of weld bead geometry under variable welding conditions.
Loading....