Wire Arc Additive Manufacturing (WAAM) is an advanced digital technology for three-dimensional solid manufacturing. In contrast to traditional subtractive and equal-material manufacturing methods, WAAM is founded on the “bottom-up” layered discrete free form of digital 3D models, offering innovative design and manufacturing flexibility. This technology is not limited by part shapes and can efficiently create complex structural components that are difficult to produce with traditional methods. To tackle challenges in process parameter selection and predicting forming outcomes in practical arc additive manufacturing, a precise and efficient mathematical method has been devised to forecast the morphology of single-pass forming. This method enables rapid and convenient process parameter selection for Wire Arc Additive Manufacturing and assists in quality control during forming processes. This article introduces a pure convolutional neural network (CNN) model along with a loss function. Based on a single-pass single-layer wire arc additive manufacturing experiment, the CNN algorithm uses process parameters as input to predict the model and develop a comprehensive prediction model for the cross-sectional profile of the weld bead. Furthermore, employing this pure CNN network model and loss function greatly aids in predicting the single-layer morphology of arc additive manufacturing. The results show that the method proposed in this article can accurately forecast the cross-sectional profiles of single-layer single-pass and multi-layer single-pass welding beads in arc additive manufacturing.