Development and validation of condition monitoring techniques for electrical machines

Abstract

This research focuses on rotor retaining band faults in Permanent Magnet Synchronous Generator, specifically in aeronautical applications. Inconel718 are commonly used in the rotor retaining sleeves as it can withstand high mechanical force and high conductivity for heat exchange. The objective of the project is to investigate the effects of faults retaining band and analysis the effects based on simulation and data analysis method. This project involved conduction Finite Element Modelling to simulate produce the electromagnetic signature of a Permanent Magnet Synchronous Generator. Simulation was conducted on Ansys Maxwell. FEM simulations produce accurate results without the need of using physical equipment for testing. Results from FEM simulation are retrieved and imported to MATLAB for data processing. As results from FEM are in time-domain, often the effects of fault cannot be identified, especially for signals high harmonics order. Time-frequency transformations methods enable easy identification of fault harmonics as the resultant waveform can be easily compared to the baseline or healthy models. There are many methods used by other researchers when analysing electrical machines faults such as Fourier Transform based methods like Fast Fourier Transform, Short Time Fourier Transform, or other methods like Wavelet Transform and Hilbert Transform. Different methods utilised different function and hence the results will vary based on the method selected. In this project, Fast Fourier Transform and Wavelet Transform were used to analyse the results. Using Fourier Transform methods like Fast Fourier Transform and wavelet transform, time-domain signals and transformed int frequency-domain signal. Amplitude from the transformation was analysed in this project. Wavelet transform is another method used to analyse results from FEM simulations. From the simulation done in this project, we can see torque and current variations, especially in higher order harmonics, where a 2% variation can be seen in the fault models. And from wavelet transforms, we can see signal variations caused by the faults.