Rapid optimization of interior permanent magnet (IPM) machines using the response surface method and dimensionless parameters
The primary objective for this work is to create a method to rapidly optimize interior permanent magnet (IPM) machines. For this research, a method was created called dimensionless optimization (DO). Eight dimensionless parameters determine nineteen dimensions of the machine based on design rules, stresses in the rotor and back EMF. Given a maximum torque specification and a local design space, a hypercube design of 64 machines is created using an 8 factor I-optimal design. Analysis is performed using a commercial program which incorporates electromagnetic FEA, and provides the current and phase angle requirements, and efficiency of the machines at the defined test points. Response surfaces are created for efficiency and current at each test point. The method determines the variance of the response surfaces, the optimum predicted design in the local design space, and the location of design spaces which may yield better designs. The method was validated by comparing 20 kW and 50 kW machines designed by an electric machines manufacturer, to machines designed using dimensionless optimization. The efficiencies of both machines were improved using dimensionless optimization.