Our paper, titled, “Physics-informed machine learning surrogate for scalable simulation of thermal histories during wire-arc directed energy deposition” has recently been published in Additive Manufacturing Letters.
This paper aimed to surrogate the computationally-expensive FEM simulation necessary to determine temperature histories during the additive manufacturing method, wire-arc directed energy deposition, using a purely physics-informed neural network (PINN). Rather than traditional data-driven approaches, physics-informed neural networks train the model based on residual losses as determined by previously established constitutive equations, such as the conservation of energy or mass.
Key Highlights:
- PINN was trained purely with physics without any experimental or external data.
- PINN reduced runtime by up to 98.6% compared to FEM.
- PINN provides resolution-agnostic predictions in both time and space.
Read the full article here: https://doi.org/10.1016/j.addlet.2025.100327
