The use of a genetic algorithm in the design optimisation of a brushless DC permanent magnet machine rotor

The paper describes the use of a genetic algorithm in the design optimisation of a surface magnet brushless DC machine. The method combines a two-dimensional magnetostatic finite element model (FEM) with a genetic algorithm (GA) to investigate the influence of the shape of the permanent magnets and their direction of magnetisation. The objective function used in the optimisation procedure is based on maximising torque for a given volume of permanent magnet, assuming the electrical loading of the machine remains unchanged. Compared to a standard radial magnetisation, the optimised rotor designs yield a significant improvement in output torque. To confirm the GA technique a gradient descent optimisation is also applied to the problem.