Determination of laser welding process parameters through a combined neural network and computational fluid dynamics approach

Calculations of fluid flow and heat transfer in the weld pool are strongly influenced by process parameters like laser spot diameter and laser power reaching the surface of the weld pool. Additionally, the chemical properties of the material give raise to large variations in the material properties. The present work develops a neural network model to estimate the uncertain parameters necessary for weld pool modeling following an inverse modeling approach. A set of weld pool shapes of type 304 stainless steel obtained through computational fluid dynamics simulations is used to train the artificial neural networks. The neural network then predicts the input parameters to the simulations like laser power and diameter. These parameters are then used to obtain an accurate, high resolution solution of the fluid flow and heat transfer in the weld pool.