A Taguchi-LHS-RSM Double-Staged Approach for Design Optimization of Self-Ventilated Cooling Systems Utilized in PMSMs

Thermal management is critical for high-power-density permanent magnet synchronous machines (PMSMs). To realize the efficient and effective design optimization of self-ventilated air cooling systems for PMSMs, in this paper, a Taguchi-preconditioned Latin hypercube sampling (LHS) and response surface method (RSM) combined approach is proposed. First, as the optimization variable numbers are mostly large when considering simultaneously the heat-sink and air-fan influences, the Taguchi method is employed to decouple the variables based on comprehensive fluidic-thermal calculations. The values of the less interconnected variables are optimized directly in this stage while the two most interconnected ones are selected and optimized in the next stage. Secondly, LHS is used to generate the sample points of the two variables with the values centered at the initially optimized results in the first stage. The RSM surrogate model is established with the LHS points, and the finally optimized values are then decided based on analytical analysis of the surrogate model. The proposed double-staged method is utilized in optimizing the self-ventilation system of a 2000-rpm outer rotor PMSM, and the optimization effectiveness is validated by comparing the cooling efficiency and effectiveness of the results at each optimization stage.