High-frequency analysis methods of inter-turn stresses in the windings of electrical machines

Mrs Kaoutar Hazim will publicly defend her thesis entitled "High-frequency analysis methods of inter-turn stresses in the windings of electrical machines".

Summary

This Ph.D. thesis focuses on modelling the electrical constraints imposed by the new generation of semiconductor devices on electrical machine windings. We make use of a previously developed circuit model whose parameters were computed at a single frequency and improve it by properly configuring the different parameter identification techniques in terms of accuracy, computational complexity, and frequency-dependency. Through experimental investigation, we explore the frequency-dependent behaviour of capacitive coupling and provide justification for its neglect in the model. The frequency-dependent elements are initially incorporated in the frequency-domain. Our proposed method is automatic and does not require the use of any fitting technique. The accuracy of the frequency-domain model is validated against various test cases by comparing the computed impedances with their respective measured values. Subsequently, the frequency model is employed to compute the nodal admittance matrix, focusing only on the nodes of interest for a straightforward computation of the nodal voltages. These voltages are then transformed into the time-domain using an inverse fast Fourier transform. To evaluate the accuracy and computational complexity, we compare our results with an equivalent vector fitting implementation. The inverse fast Fourier transform implementation is automatic, twice as fast as a vector fitting implementation and accurately predicts time-domain over-voltages. However, it exhibits spurious oscillations, which are substantially reduced by the equivalent fitting implementation. The computed time-domain nodal voltages are used to deduce the inter-turn voltages, demonstrating good agreement with time-domain measurements. The proposed method can be used as input to the various numerical methods employed in predicting premature failure of electrical machine windings.

Practical information

Defence will take place on September 20th 2023 at 10:30, to all at Institut de Montefiore (salle R7, Bât. B28 - Sart Tilman).