Three-dimensional separations in compressors
A lack of understanding of the nature and characteristics of three-dimensional (3D) separation, which is inherent in compressor blade passages and tip clearance dominated 3D flow in the rotor tip region, has limited the success of 3D blade design. This dissertation is therefore aimed at contributing to compressor design process by exploring the formation of 3D separations. Detailed measurements in compressor cascades, tests in a low-speed single stage Deverson compressor rig and considerable use of Denton multi-stage Reynolds averaged Navier-Stokes (RANS) solver MULTIP have provided better understanding and improved predictive capability of this important phenomenon. An attempt has also been made to clarify the role of different flow mechanisms.
Tests carried out to determine the effect of surface roughness in the single-stage axial compressor show that stage performance, even at design point is extremely sensitive to roughness around the leading edge and peak suction regions because of the effect on 3D separations. A numerical model to simulate the effect of roughness was formulated and incorporated into MULTIP and this showed good agreement with measurements.
A method of estimating the thickness of the 3D separated layer has been developed, which was used in a parametric study to assess the sensitivity of 3D separations to key flow and design parameters in axial compressors. Investigation into endwall flow control methods was undertaken, based on the understanding of the formation of 3D separation. These include clearance flows, endwall fences and endwall dividing streamline suction. The latter was demonstrated to be the most effective method of control.
The knowledge gained in predicting 3D separation was then used to explore the 3D nature of the flow around rotor tip sections and how this is influenced by key 3D design techniques. The performance of the existing 3D rotor and the modified version with tip chord extension was analysed using detailed CFD, with carefully refined mesh especially in the blade tip/clearance region. This approach enabled key problems with the oliginal 3D design to be diagnosed. The lesson learned from the performance of the 3D rotor therefore guided a re-design of the rotor tip. It was found that a more modest amount of blade lean is enough to achieve improved performance improvement in the tip region.
Suggested future work that will aid further understanding of 3D separations in terms of flow mechanism, unsteadiness and control are also presented.