Predicting time-dependent demagnetization behavior of permanent magnetic materials by a physics-informed neural network-based strategy

Based on physics-informed neural networks (PINNs), this work proposed a prediction strategy for the time-dependent demagnetization of SmCo magnets to address challenges in evaluating their stability and lifespan in complex and extreme environments. The PINN model was trained to predict the demagnetization rate and was integrated with numerical simulations to forecast time-dependent demagnetization behavior. Experimental validation demonstrated that this strategy effectively predicted both short-term and long-term demagnetization behavior of permanent magnets in high-temperature environments. Furthermore, it enabled predictions of demagnetization during temperature increasing and cycling in real time scale, offering greater accuracy compared to traditional M-H loop-based methods. This work is promising for various types of permanent magnets and is expected to significantly reduce the verification cost for permanent magnet devices, providing new insights and approaches for evaluating the lifespan of permanent magnet materials.