TL;DR
This paper presents a novel method to accurately model permanent magnet synchronous machines (PMSMs) by first identifying their real magnetic material properties through an inverse problem using torque measurements. This accurate model is then used to optimize the machine's geometry for maximum torque-to-loss ratio, leading to more efficient designs.
What Problem Does This Paper Address?
Designing highly efficient and compact permanent magnet synchronous machines (PMSMs) is crucial for applications like electric vehicles where space and weight are critical. A key challenge in accurately modeling these machines is the non-linear behavior of magnetic materials, especially under saturation. Standard modeling approaches often rely on manufacturer-provided material data, which can be inaccurate due to factors like mechanical stress and material degradation that occur during the manufacturing process. This leads to discrepancies between simulated and actual machine performance. The paper addresses this by proposing a method to derive the actual magnetic material properties of a machine from experimental measurements, and then using this refined model for geometric optimization.
Methods
The core of the paper's methodology involves two main stages:
- Material Characterization via Inverse Problem:
- A 2D finite element model (FEM) of the PMSM is developed.
- The magnetic material's non-linear characteristic is represented by a three-parameter model (specifically, a permeability model $\mu(B) = \frac{B_0}{\mu_0} \left( \frac{1}{1 + (\frac{B}{B_0})^\nu} \right)$).
- An "inverse problem" is formulated: torque measurements at standstill, across various rotor angles and current amplitudes, are used as input.
- A minimization algorithm (interior-reflective Newton method) is employed to find the optimal parameters of the material model that best match the measured torque data. This approach accounts for the fact that material properties can change within the machine.
- To mitigate the impact of measurement noise and torque ripple (potentially caused by rotor eccentricity), an alternative cost function is introduced that fits a smooth analytical curve to the measured torque data before comparing it to the FEM-simulated torque.
- Geometric Optimization:
- Once the material properties are accurately identified, the validated FEM model is used for geometric optimization.
- The goal is to maximize the ratio of torque to total losses (iron losses + copper losses).
- Key geometrical parameters optimized include the thickness of the permanent magnets, the width of the stator teeth, and the angle of the stator current components (d- and q-axis).
- The optimization is performed for both low-load (small currents) and high-load (large currents) conditions, as the optimal geometry differs between these scenarios.
Datasets
- Torque Measurement Data: The paper utilizes experimental torque measurements taken at standstill for various rotor angles and stator current amplitudes. These measurements are crucial for the inverse problem to identify material parameters.
- Local BH-Measurements: Local magnetic flux density (B) and magnetic field (H) measurements were performed on the stator yoke to validate the material characteristics derived from the inverse problem.
- Electromotive Force (EMF) Measurements: No-load EMF measurements were used to validate the overall FEM model's accuracy in predicting machine behavior.
Key Insights and Ideas
- Material Properties Matter (and Change): The paper strongly emphasizes that relying on manufacturer data for magnetic materials in PMSMs can lead to inaccurate models. Real-world conditions (mechanical stress, degradation) alter these properties, and an inverse problem approach using actual machine measurements is necessary for accurate characterization.
- Torque as a Sensitive Indicator: Torque measurements at standstill are shown to be highly sensitive to the magnetic material properties, making them an effective input for the inverse problem.
- Addressing Measurement Noise: The introduction of an alternative cost function that smooths measured torque data before optimization is a clever way to handle measurement inaccuracies and torque ripple, leading to more robust material parameter identification.
- Load-Dependent Optimization: The optimal geometry for a PMSM is not universal; it depends significantly on the operating load. Optimizing for low load might prioritize reducing iron losses (wider teeth), while optimizing for high load might focus on reducing copper losses (larger copper sections, thicker magnets).
- Air Gap Sensitivity: The air gap length, even with small variations, has a significant impact on the saturation induction and thus on the overall machine performance. This highlights the importance of precise manufacturing for optimal results.
Key Results
- The developed inverse problem method successfully identified the magnetic material parameters of the PMSM, leading to a numerical model that accurately predicts measured torque and EMF.
- The optimized geometries showed improvements in torque-to-loss ratio. For instance, at high currents (16A), the optimal design featured thicker magnets and narrower stator teeth compared to the low-current (2A) optimization.
- The paper demonstrates that the derived material characteristics, when used in the FEM, provide a better match to experimental torque profiles than using standard material data.
- Sensitivity analysis showed that the air gap length significantly influences the optimal magnetic parameters and torque profiles.
Limitations and Future Work
- 2D Model Assumption: The study uses a 2D FEM, which neglects 3D effects that can influence material parameter identification and machine performance.
- Air Gap Tolerance: The accuracy of material parameter identification is shown to be influenced by the tolerance on the air gap.
- Future Work: The authors suggest that extending the analysis to 3D models would provide even more accurate results. Further investigation into the impact of manufacturing tolerances, particularly on the air gap, is also warranted. The method's generalizability to other electromagnetic devices (transformers, inductors) is also implied.
Summary updated