The student Mikel Mendizabal Eizaguirre an EXCELLENT CUM LAUDE
The student Mikel Mendizabal Eizaguirre an EXCELLENT CUM LAUDE
The student Mikel Mendizabal Eizaguirre an EXCELLENT CUM LAUDE
- Thesis title: Vibration Estimator for the Reduction of Vibrations and Noise of Electric Machines by Control
Court:
- Presidency: David Diaz Reigosa (Universidad de Oviedo)
- Vocal: Frédéric Druesne (Université de technologie de Compiègne)
- Secretary: Izaskun Sarasola Altuna (Mondragon Unibertsitatea)
Abstract:
Due to the increasing demand for high power-density solutions, the use of Permanent Magnet Synchronous Machines (PMSM) has significantly grown in recent years. However, higher power density often leads to increased vibration and noise levels, which not only affect user comfort, but can also compromise the reliability and durability of machine components. Consequently, optimizing the vibration and noise behavior of PMSMs has become a crucial aspect.
Control-based strategies represent a more flexible and cost-effective alternative to design modifications for reducing vibrations and noise. For an effective control, the global vibration behavior of the machine needs to be known in real-time, and the signal from a single physical sensor is usually not a reliable reference. Therefore, the main objective of this thesis is to develop a model that provides a fast and accurate prediction of the vibration behavior of PMSMs under varying working conditions, to be used as a virtual sensor within a vibration reduction control strategy.
Detailed Finite Element (FE) models are employed to characterize both the electromagnetic forces and the structural vibration modes of the electric machine. Simulations are performed to capture the vibration behavior of the machine across a wide range of working conditions, and the simulation results are implemented in a novel multiphysical reduced model using Look-Up Tables. This methodology achieves an accuracy comparable to conventional FE models while reducing the computational time from several days to a few seconds.
Several modeling aspects are thoroughly investigated to define clear guidelines on their influence on the performance of the model, and to optimize the balance between accuracy and computational efficiency. The study highlights the significant impact of manufacturing tolerances; two approaches are proposed to model their effect, which substantially improve the accuracy of the model. A computationally efficient method for modeling rotor skew is introduced, demonstrating its strong influence on vibration harmonics. The contribution of the rotor to the vibration response is also shown to be significant, but tangential forces are found to have a minor effect in the studied machine. The use of concentrated forces is shown to greatly reduce the calculation time without compromising accuracy, and criteria are established for selecting the optimum axial discretization and number of vibration modes.
The proposed model is validated using conventional FE simulations and experimental vibration measurements under various operating conditions. The results confirm its ability to accurately predict both the amplitudes and the load and speed dependence of the main vibration harmonics, with a computational cost close to real-time. Therefore, the potential of the reduced model to be used as a virtual sensor in a vibration reduction control strategy is demonstrated.