Generalizable DNN based multi-material Hysteresis Modelling

This The effective representation of material properties is fundamental to the simulation of electromagnetic devices such as electrical machines, actuators, sensors, transformers, etc. However, the actual operating point of a material is dependent both on position within the device and the excitation. Every point in an electrical machine can be operating on a different part of the magnetization curve. To determine performance parameters such as the efficiency of the machine, the hysteretic behavior of the material is crucial, and the representation used can impact the performance of a simulation code. In this paper, the use of deep learning methods is proposed to reduce the computational effort needed to implement hysteresis in a finite element based simulation system.