Abstract:

An accurate magnetic model is essential for highperformance control of electric motors because it characterizes magnetic flux as a nonlinear function of the stator current. However, both acquiring and using magnetic flux maps can be time-consuming, making them unsuitable for low-latency control applications. To address this limitation, we propose a compressedsensing-based approach that reconstructs magnetic flux maps from a limited set of randomly sampled data points. Experiments are conducted on a permanent-magnet machine to validate the proposed method. Extensive simulation and experimental results demonstrate that, by exploiting the sparsity of flux maps, the proposed approach can accurately reconstruct the maps using only 25% of the full experimental samples, thereby significantly reducing the acquisition time.