An agent-based collaborative FEA system for the optimal design of electromagnets

Abstract

To achieve the optimum design of an electromagnet, it is usually necessary to predict its static and dynamic characteristics. The finite element method (FEM) rather than lumped parameters is frequently adopted to calculate static characteristics. Generally, dynamic characteristics are calculated by iterations of many static characteristics derived from diverse excitation currents and strokes with the Runge-Kutta method, which costs much time. To use finite element analysis (FEA) codes to compute dynamic characteristics more efficiently, this paper proposes a collaborative FEA system (C-FEAS) for the optimal design of electromagnets based on multi-agent system technology. The C-FEAS can perform lots of static characteristics in several FEA hosts concurrently. A Contract Net protocol is used for multi-task allocation among multi-agents, each of which stands for one computational FEA host. A prototype is implemented using the Java Agent Development Framework (JADE) and ANSYS software.