A Novel High-Precision Network Auto-Generation Approach for PMSM Thermal Analysis via Pixel Information

High power density is a key design specification for modern permanent magnet synchronous motors (PMSMs), which imposes strict requirements on effective high-precision temperature prediction methods. This paper proposes a pixel information-based method for generating multi-node thermal networks to enable efficient thermal analyses of PMSMs. This method takes cross-sectional motor images as input and automatically generates a high-node-density thermal network by structured grid discretization and pixel-level material information extraction, enabling fully automated thermal network construction. Based on the multi-slice three-dimensional network, the ATNG method achieves high computational efficiency while maintaining accurate temperature prediction, allowing effective representation of internal temperature distributions and transient thermal behavior. Experimental validation on stator modules and a PMSM prototype under various operating conditions shows that the ATNG method can reliably predict temperature trends in both steady-state and transient regimes, with a maximum prediction error within 5%, while offering clear advantages in modeling and simulation efficiency.