Multiphysics Multi-Objective Optimization for Electric Motor NVH

Prediction and reduction of noise/vibration at the early design stage is important for motor design. Rapid design iterations require a platform where electromagnetic, structural and acoustic solvers can communicate with each other without user scripting or interventions. Based on the platform, multiple designs in a given design space need to be analyzed by distributed high performance computers automatically. To demonstrate such a multiphysics multi-objective optimization workflow, four geometrical variables for an interior permanent magnet motor are selected for optimizing the electric and acoustic performance (Figure 1). Average torque and equivalent radiated power level (ERP) are calculated for multiple design points and response surfaces are then created for the sensitivity study and optimization. Changes on the stator and rotor are adopted to directly address the source of the noise, and a much larger reduction is achieved compared to design changes on the housing only (7.3 dB vs. 0.4 dB).