A Semi-Analytic Model for Temperature Distribution Prediction and Hotspot Tracking in SPMSM and Its Application for Improved Multiphysics Multiobjective Optimization Design

Along with the high-quality development of permanent magnet synchronous motors (PMSMs), the problem of temperature rise has become increasingly prominent. To this end, a semi-analytic model (SAM) combining with the analytical method (AM), the lumped parameter thermal network method (LPTNM), and the finite-element method (FEM) is proposed with the commonly eccentric surface-mounted PMSM (SPMSM) as the main object of study. The complex structure and the varying heat dissipation characteristics of the eccentric SPMSM are reasonably accounted for. Global temperature distribution prediction and hotspot tracking (especially for critical parts, such as eccentric permanent magnets (PMs) and windings) are realized. Meanwhile, the proposed SAM combines high analytical accuracy and fast analytical speed without consuming too much computational resources. Furthermore, the extensibility and the flexible application of the proposed SAM for the multiphysics multiobjective optimization design of the eccentric SPMSM are also investigated. An optimization model is constructed by combining the proposed SAM, machine learning technique, and nondominated sorting genetic algorithm III (NSGA-III). The multiphysics characteristics of the eccentric SPMSM are well improved. Electromagnetic performance is effectively improved, and temperature rise is significantly suppressed. Finally, this study is fully and robustly validated by simulation analysis and prototype experiment. Overall, the proposed SAM has an extremely positive performance. The global temperature distribution predicted by the SAM agrees well with the simulation results. Moreover, the prediction deviations of the hotspot of critical parts are very small, basically within 5.00%, and the ones of the PMs and windings are even lower than 3.50%. Furthermore, the optimization model developed based on the proposed SAM also has excellent optimization performance. The average torque (0.27%) is enhanced, while the torque ripple (74.07%), total loss (9.42%), hotspot of PMs (12.67%), and hotspot of windings (9.92%) are significantly suppressed.