Prediction of the interpass temperature of a wire arc additive manufactured wall: FEM simulations and artificial neural network

Abstract

Finite element method (FEM) simulations are a powerful tool for understanding the thermal–metallurgical–mechanical effects of wire arc additive manufacturing (WAAM). Nonetheless, owing to the multiphysical nonlinear nature of welding coupled with the longer deposition time of WAAM, FEM simulations can be laborious and time-consuming, which makes it difficult to simulate the numerous procedural parameters of WAAM. Therefore, the present work aimed to employ an FEM mode to analyze the influence of idle time (30–240 s) on the interpass temperature (IT) of 20-layer single-bead walls produced via WAAM and use the FEM results to develop a predictive model for the IT based on an artificial neural network (ANN). The FEM simulations were performed using a heat source and a 20-layer single-bead wall model that was experimentally calibrated and validated. The first layers exhibited similar energy accumulation; however, as the wall height increased, the IT rapidly increased under to low idle times (≤120 s). The ANN was trained using the FEM simulations results, validated with FEM results (not included in the training database), and used to establish a process map (including the idle time, number of layers, and IT). This can help the manufacturers to obtain a suitable balance between productivity (lower idle times) and part behavior (e.g., microstructure and mechanical properties). • A FEM 20 single-bead layer wall WAAMed model was accurately calibrated and experimentally validated; • The influence of idle time (30–240 s) on interpass temperature (IT) was analyzed trough the FEM model; • FEM results were used as database to train an artificial neural network (ANN); • ANN was used to predict the process map considering the idle time, layer number, and IT.