Multi-stage Aeromechanical Phenomena and Computation Principles of a Compressor

This dissertation presents an investigation of the complex aeromechanical phenomena in a multi-stage turbomachinery. As a major component of the inter-disciplinary design, aeromechanical study demonstrates the interaction of fluid and structure within turbomachines. The non-uniformity or the unsteadiness of the flow interacts with the bladed disks, causing vibrations of blades. As the integrity of the machine is at stake when vibrations are at presence, it is critical to understand the mechanism of aeromechanics behaviors of a turbomachine.

The research is conducted with computational methods on the Purdue 3.5-stage Compressor test facility. The prediction is compared against experimental aerodynamic data and vibratory response measurement data to uncover the interaction among different aeromechanics phenomena and the key drivers influencing the prediction of forced response. Both frequency-domain and time-domain computational methods are used in this study.

The first part of the study addresses the interaction of forced response and flutter. A quasi-3D stator-rotor configuration is selected for this study. The influence from flutter to forced response is observed: a one-way crosstalk at forced response frequency is observed, presented as the anomaly of unsteady velocity and unsteady pressure near the rear section of rotor blades and in the rotor wake region. The anomaly is speculated as the presence of increasing intensity of shedding vortices induced by the vibration of the blade. To further test the impact of this viscous effect, a numerical experiment is performed with inviscid rotor blades. In contrast to the crosstalk at forced response frequency, no obvious influence on the unsteady behavior is detected at the flutter frequency, and this observation is confirmed at multiple vibration amplitudes. Considering the relationship between unsteady pressure at flutter frequency and aerodynamic damping, we conclude the influence of forced response on the aerodynamic damping is negligible. In addition, a linearity of unsteady pressure at the flutter frequency vs. vibration amplitude is uncovered.

The second part of the study demonstrates the influence of the spurious of wave reflections when using reflecting boundary conditions and including the wake from the non-adjacent stator row. The aim is to provide an accurate prediction of forced response with the least computational effort. Previous research indicates that by reducing the computation domain from 7-row to a 3-row stator-rotor-stator (S1-R2-S2) configuration, the forcing function is over-predicted by 80%. To address this over-prediction, an investigation of boundary conditions and a study with additional rows are conducted. The influence of reflecting boundary conditions on the blade modal force is studied by preventing wave reflection. Additionally, a 5-row simulation is studied to take an extra source of excitation force, the IGV row with the same blade count as the other stators, into consideration. Three conclusions are drawn from this study: 1) boundary reflection has a significant influence on unsteady simulation and the modal force, thus should be avoided by using mesh treatment at both up- and down-stream; 2) the IGV wake mildly contributes to the forcing function but cannot be ignored; 3) the clocking feature of IGV, S1, and S2 renders the excitation energy transferred from 1st harmonic to other higher harmonics.

The third part of the study examines the physical wave reflections from blade rows in the compressor. When waves propagate through the compressor, the blade rows serve as “walls” with partial admission and will reflect waves. Those physical reflecting waves might interact with the original excitations and influence the results. The aim is to understand the significance of the physical wave reflections, and test the necessity of including the rows that are not directly contribute to the generation of excitations. By including the downstream R3 in the 4-row simulations (S1-R2-S2-R3), better prediction of the 1T-44EO forcing function is achieved. The rotor reflection contributes to about 30% of resulting modal force. By modifying the blade counts of S1, the forcing generated from up- and downstream of the imbedded rotor are separated and studied at the associated frequencies. Stator wave reflections are identified, which can contribute to to 30-50% of forcing function. Three conclusions are drawn from this study: 1) For a mistuned forced response, the averaged response should be used for as an indicator of the forcing function. 2) The reflecting waves from the downstream rotor create a destructive interference with the original excitations and increase the forcing, whereas the adjacent up- and downstream stators, both create constructive wave reflections. 3) When simulating a blade row sandwiched by two stators, even if the blade counts are not the same and gives excitations at different frequencies, it is still necessary to include both rows in the simulation, as the physical wave reflections are not simulated in 2-row cases.

This dissertation provides a guidance of forced response and flutter modeling, and the distilled computation principles can be employed for industrial analysis in practice.