Modeling of additive manufacturing processes with time‐dependent material properties using physics‐informed neural networks

Abstract

Recently, physics‐informed neural networks (PINNs) have been effectively utilized in a wide range of problems within the domains of applied mathematics and engineering. In PINNs, the governing physical equations are directly incorporated into the loss function of the network and a conventional labeled dataset is not required for its training. In order to successfully simulate the additive manufacturing processes with concrete, a novel process‐based FE‐simulation has been developed where the Drucker–Prager plasticity model is used as the material model. In this work, we will examine the deployment of a PINN to substitute the Newton–Raphson iterations that occur in the return‐mapping algorithm of the Drucker–Prager plasticity model.