A experiment study for welding optimization of fillet welded structure

Purpose: This study aims to examine the interaction between process parameters and bead geometry, to perform the research to predict optimal bead geometry (bead width, reinforcement height, left leg length, right leg length) through the analysis of experimental data. For this, not only linear and the curvilinear equations were developed to predict bead geometry, but also interactions between process parameters and bead geometry were analysed through sensitivity analysis. Design/methodology/approach: A Taguchi method was applied for the optimization of process parameters, as well as bead geometry was predicted using a Neural Networks (LM) learning algorithm. Findings: The data generated through experimental studies conducted in this study has employed to validate its effectiveness for the optimization of bead geometry on process parameters (welding current, welding voltage, welding speed), and to present the criteria to control the process parameters to achieve a good bead geometry. Research limitations/implications: By applying Taguchi method, process parameters (welding current, welding voltage, welding speed) and bead geometry (bead width, reinforcement height, leg length) were analysed. Originality/value: This study has developed mathematical models and algorithms to predict or control the bead geometry in GMA fillet welding process, and analysed the S/N ratio to which Taguchi theory was applied for sensibly to the process parameters.