LAUSR

This paper proposes a multicriteria design optimization methodology for permanent magnet (PM) motors used in electric vehicle (EV) applications. In the process, an adaptive-network-based fuzzy inference system (ANFIS) is utilized, coupled with a multiobjective optimization algorithm, as a surrogate model of the electric motor. This allows for the consideration of the full drive cycle and respective efficiency map for every motor design. The prediction error of the ANFIS is minimized by employing appropriate membership functions, initial training data, and an adaptive learning scheme via iterative training. The efficiency map is then implemented in a vehicle dynamic model to compute the total consumed energy over the driving cycle. The optimization profile accounts for total energy efficiency, torque density, and additionally considers complementary design criteria via an a posteriori selection procedure on the resulting Pareto set. The methodology developed is applied to optimize a surface PM motor with concentrated fractional slot winding, mounted on a light EV that competes in fuel economy races. The selected motor design has been validated through measurements on a prototype.