Simulation of non-linear magnetic circuits modelled using artificial neural network

Abstract

Simulation of electromechanical systems and other systems where coupling of different physical effects is modelled, is currently a very active research area. It gives rise to development of hardware description languages where description of such different physical devices is possible and to implementation of simulators that support such languages. Models of devices for system simulation should be simple enough, so that simulation time stays in reasonable limits, but must describe correctly enough all relevant physical dependencies. Modelling with such demands is a very difficult task. An automated approach for solving this problem is usage of neural networks for modelling. In this paper, a neural network is used to model some complex physical dependencies of an electromagnetic circuit, while dynamic mechanical behaviour is modelled analytically. Model is described in an object-oriented hardware description language and simulated by the simulator Alecsis. In this way, a fast and accurate simulation of electromagnetic system is obtained.