Abstract:

An accurate magnetic model for electric motors is essential for high performance control strategies. The magnetic model is typically acquired by massive experiments of measuring magnetic flux data throughout the operating current range, and then applied in the control process via a look-up table of measurements. Both the acquisition and the application processes are time-consuming and not suitable for low-latency controls. To address this issue, we propose a novel compressed sensing-based method to recover a high-fidelity flux map from limited randomly sampled data points, and further infer an analytical magnetic model of the recovered flux map. This analytical model can then be used to efficiently compute the magnetic flux instead of looking up measurement data, given the stator current in the control loop. The proposed approach is validated on data simulated by finite element analysis (FEA).