Cognitive Power Electronics for Detection of Demagnetization in Electric Drives

The increasing number of safety-critical applications for electric drives are chasing the need for methods to monitor the health status of the motor. Especially for safety-critical steering and traction functions for autonomous driving or for different types of aerial vehicles, a high reliability of the drive is crucial. A significant failure mode in permanent magnet synchronous machines (PMSM) is the demagnetization of the rotor magnets. If demagnetization is detected early, system parameters can be adapted to the reduced performance, or to bring the system in a safe state. In this paper a detection method for demagnetization detection during normal operation is described. Both the simulation-based and experimental steps for data acquisition and understanding will be explained, as well as the development and evaluation of the detection pipeline for demagnetization detection. It can be shown that partial demagnetization leads to specific asymmetries in the magnetic flux in the airgap of the machine, which produce additional harmonics in the induced voltage and the phase current. Here, the harmonics at the demagnetization depend on the specific slot-pole-combination of the motor. This is demonstrated both in electromagnetic simulation and in experiments with artificially damaged motors with different levels of demagnetization. For the data-driven recognition of the demagnetization the phase currents of intact and damaged motors are recorded on a test rig. The detection pipeline comprises a spectral analysis and a dimension reduction using the Fast Fourier Transform (FFT), followed by a kernel principial component analysis (kPCA). With this a clear differentiation of the intact motors and investigated fault cases is possible.