Forecasting process parameters for GMAW-based rapid manufacturing using closed-loop iteration based on neural network

During the layered deposition of forming metallic parts with robotic gas metal arc welding, the geometry of a single weld bead plays an important role in determining the layer thickness and dimensional precision of the deposited layer. This paper addresses prediction of welding process parameters for the expected bead geometry with accordance to the adaptive slicing process in rapid manufacturing. The correlation between the process parameters and the bead geometry was established by applying an artificial neural network. A central composite design was employed to conduct experiments for collecting input-output data. A closed-loop iteration system, consisting of a forward model for predicting bead size and a reverse model for forecasting process parameters, was proposed to predict the optimal welding parameters for the desired bead geometry. The results show that the prediction of process parameters obtained from the designed closed-loop iteration system confirms the feasibility of this system in terms of applicability and automation in the additive manufacturing process.