Enhanced Temperature Prediction in Wire Arc-Directed Energy Deposition Using Physics-Informed Neural Networks

Abstract Wire Arc-Directed Energy Deposition (WA-DED) is an advanced metal 3D printing process that enables the production of near-net-shape metal parts by depositing material layer-upon-layer at a high deposition rate. Accurate temperature control in WA-DED is crucial for controlling the microstructure of the fabricated metals. In the past decade, machine learning and computational capability have grown rapidly. Various industries are eager to rely on data-driven models. Data-driven models can deliver accurate predictions when large and quality datasets are available. However, their accuracy tends to diminish when small but high-quality data is provided, which makes precise predictions more challenging. In high-deposition rate WA-DED printing processes, relying solely on data-driven models is not an ideal choice for ensuring accurate predictions. An effective method for monitoring temperature is Physics-Informed Neural Networks (PINNs). PINNs provide a promising approach for real-time temperature monitoring by integrating physics equations into conventional neural networks. This allows PINNs not to rely on data but to make predictions from physics equations. In this study, we used PINNs to predict the temperature evolution in a metal substrate subjected to a moving heat source, simulating the heat transfer process in WA-DED. The results show that our fully trained PINN models with different architectures achieved accurate predictions of approximately 2% average percent error over the Finite Element (FE) model. Therefore, this model holds significant potential for being utilized as a real-time monitoring tool in the WA-DED printing process.