Torque-Speed Characteristic Estimation based on Gaussian Processes and Adaptive Sampling Strategy for Permanent Magnet Synchronous Machines

Authors

• Marcelo D. Silva, Uppsala UniversityFollow
• Pedram Asef, University College LondonFollow
• Oluwaseun A. Badewa, University of KentuckyFollow
• Rosemary E. Alden, University of KentuckyFollow
• Dan M. Ionel, University of KentuckyFollow

Abstract

Internal Permanent Magnet Synchronous Machines (IPMs) are widely used and typically optimized to meet specific performance requirements. Parameters such as base speed, maximum torque, and maximum speed commonly define the torque- speed characteristic of a given design. This study introduces a novel machine learning approach for statistically estimating the torque-speed characteristics of IPMs using Gaussian Process Regression (GPR), which models predictions as random variables. By leveraging uncertainty quantification, the study explores sampling strategies that enable the construction of a high-precision meta-model with minimal error and uncertainty. The proposed adaptive sampling strategy, combined with GPR, accurately estimates torque-speed characteristics and associated loss components across the design space using only a limited number of Finite Element Method (FEM)-based simulations. The results demonstrate good agreement with full FEM evaluations and experimental measurements, validating the effectiveness of the proposed method.

Document Type

Conference Proceeding

Publication Date

10-2025

Digital Object Identifier (DOI)

doi: 10.1109/ECCE58356.2025.11260187

Repository Citation

Silva, Marcelo D.; Asef, Pedram; Badewa, Oluwaseun A.; Alden, Rosemary E.; and Ionel, Dan M., "Torque-Speed Characteristic Estimation based on Gaussian Processes and Adaptive Sampling Strategy for Permanent Magnet Synchronous Machines" (2025). Electrical and Computer Engineering Graduate Research. 22.
https://uknowledge.uky.edu/ece_gradpub/22