Detection of axial flux coreless PMSG demagnetization fault

Axial flux permanent magnet synchronous generator (PMSG) with high power density is used in a wide range of areas such as power generation plants and wind turbines. It is extremely important to detect magnetization faults that occur in the axial flux coreless PMSG in advance. Fault in these machines will adversely affect their operation. These faults can lead to both costly and irreparable situations for the business. In this study, a new method has been proposed to detect demagnetization faults in axial flux coreless PMSG generators. This proposed method differs from the traditional methods used in the literature. A novel method proposed in this study is a new texture analysis-based texture extraction. It has been revealed that it is a method to be used instead of frequency spectrum analysis, time-frequency analysis, and motor current signature analysis (MCSA) methods used in traditional methods. Both healthy and defective current and voltage signals of the axial flux coreless SMSG were obtained using the finite element method. In addition, defective signals were created to 3% and 6% demagnetization fault, and current and voltage signals were obtained. In this thesis, a new approach for fault detection has been proposed. The obtained signals were transformed into images of certain sizes. Then, the local binary pattern (LBP), which is a textural analysis method, was applied to these images. Histograms of new images formed after the LBP was applied to images were given to k-NN from machine learning techniques as an attribute vector. Trials were conducted with signals generated in different scenarios (different fault rates for different load and speed situations) to test the accuracy of the proposed approach. With the approach suggested in this thesis study, it has been revealed in the literature that demagnetization failure can be made in the axial flux coreless PMSG.