Numerical simulation and designing artificial neural network for estimating melt pool geometry and temperature distribution in laser welding of Ti6Al4V alloy

Abstract

Weld geometry is a critical factor for determining the quality of Ti6Al4V welded joints. The size of the weld cross section profile has been quantitatively investigated through experimental and numerical analysis. Due to the difficulties in temperature measuring of the molten pool region, the temperature distribution through numerical simulation was exerted as an indirect approach for estimating the size of the melt pool profile and HAZ region. Moreover, the numerical model was used for prediction of cooling rate in the melt pool and thereby characterization of fusion zone microstructure. To achieve an accurate prediction of the weld geometry at low time and cost, the process was simulated based on artificial neural network. Different ANNs were developed for progressive prediction of the weld pool temperature distribution and weld geometry. Two feed-forward back propagation neural network models with 11 and 14 neurons were developed to predict optimum process parameters. The proposed artificial neural network models perfectly predicted the process with mean square errors of 0.079 and 0.063. The results indicated that ANN outputs were in good agreement with the experimental and numerical data.