Tube-cap welding is a method commonly used in the production of products by sealing various materials into tubes and sealing them with caps, which are used in nuclear power to make fuel rods. Welding methods using laser and arc welding or butt welding using resistance heat are mainly used for manufacturing fuel rods by putting uranium raw materials used in nuclear power generation into tubes and welding both ends to caps. In this study, we analyzed the mechanism of welding phenomena by varying welding current, overlap length, and force, which are the main variables in the butt welding of the tube-cap used in nuclear fuel rods manufacturing thought simulation. In order to predict the melted volume of the weld, a regression model and a neural network were used to predict the amount of melt with the main process variables as input variables. The effect of each welding variable on melted volume was quantitatively determined and the correlation between two factors was analyzed.