Present day PMSM design methodologies often use optimization tools. Optimizing the geometry alone, may lead to designs that have high efficiency and torque density but may have high torque ripple. By injecting high frequency currents, some optimal geometries with too high torque ripple, become acceptable and optimal, even with an increase in machine and inverter losses, and a decrease in efficiency. An optimization methodology is proposed that uses an analysis of FEM sample designs and a regression model to optimize a PMSM, using a multi-objective genetic algorithm (NSGA-II) and including high frequency injection for torque ripple reduction.