Abstract: In the laser welding process of 316L stainless steel, the welding process parameters directly affect the appearance of the weld. Establishing the relationship between welding process parameters and weld geometry is very important for optimizing welding process parameters and reducing welding costs. According to the characteristics of the two-dimensional rough contours of 316L stainless steel welds, the idea of piecewise function is used to describe the contours of the welds using Hermite interpolation and least square fitting respectively, and a generalized regression neural network two-dimensional weld topography prediction model is constructed.The experimental results show that the Hermite interpolation method can obtain a more accurate weld profile, with an average relative error of −3.49%. The proposed model provides an effective method for welding seam prediction of 316L stainless steel welding process.