Yokeless and Segmented Armature (YASA) Axial Flux Permanent Magnet Synchronous Machines may have asymmetrical demagnetization defects in their two rotors. Condition monitoring therefore becomes more complicated than in conventional single rotor machines. In order to develop a real-time condition monitoring algorithm for demagnetization faults in YASA machines -which requires solving an inverse problem -a fast forward model is necessary. A frequency-based analytical model is used to calculate the three phase terminal voltages based on the three phase current waveforms and the demagnetization defects, including asymmetrical defects in the two rotors. In order to reduce the computational time of the forward model, the paper focuses on the sensitivity of different voltage harmonics on several demagnetization faults. To this end, the Cramér-Rao lower bound technique is used to analyse the harmonics sensitivity to the demagnetization faults. The goal of the sensitivity analysis is to select an appropriate set of harmonics that gives maximal information about the defects. It becomes much faster and useful for real-time condition monitoring. The results of the analysis show that the subharmonics around and including the fundamental are the most sensitive harmonics to the demagnetization defects. The analytical model is validated with experimental data and finite element simulations.
IntroductionAxial flux permanent magnet synchronous machines (AFPMSMs) of the yokeless and segmented armature type (YASA) have been extensively used in different applications thanks to their excellent. Particularly, AFPMSMs are well dedicated for applications that acquire high power density, such as sustainable energy applications. In practice, these types of machines may suffer from different faults, such as rotor eccentricity and permanent magnet (PM) demagnetization, which decrease their reliability [4] [5]. In order to prevent the defect from progressing and creating more damage in the machine, online condition monitoring systems are highly needed to detect the faults in early stages. The demagnetization defects may occur due to an excessive temperature rise in the rotor of the machine, or due to high load currents or short circuit currents in the stator. When magnets are exposed to high temperatures, their magnetization may be lost partially or completely [6][7]. This problem can be mitigated by using magnets of a higher temperature class, but these magnets are more expensive and have a lower remanence. A better solution is to detect the demagnetization via a real-time monitoring system. Demagnetization faults can be detected on-line by several sensor and/or sensorless techniques. In [8] and [9], an overview of these techniques is given. Four types of detection techniques are identified. The first type consists of current-based indexes. Here the current waveforms are analysed in either the time or the frequency domain [10][11][12] [13]. Some techniques require machine standstill while others require steady-state operation. If ...