Hysteresis Modeling in Iron-Dominated Magnets Based on a Multi-Layered NARX Neural Network Approach

A full-fledged neural network modeling, based on a Multi-layered Nonlinear Autoregressive Exogenous Neural Network (NARX) architecture, is proposed for quasi-static and dynamic hysteresis loops, one of the most challenging topics computational magnetism. This modeling approach overcomes drawbacks in attaining better than percent-level accuracy classical recent approaches accelerator magnets, that combine hybridization standard hysteretic models architectures. By means an incremental procedure, different Deep Architectures are selected, fine-tuned tested order to predict magnetic context electromagnets. Tests results show NARX architecture best fits measured field behavior reference quadrupole at CERN. In particular, framework leads percent error below 0.02% prediction, thus outperforming state art paving very promising way future real time applications.