Turbulent flow phenomena and boundary layer transition at the circular arc leading edge of an axial compressor stator blade

Abstract

As fuel prices rise and environmental awareness becomes an increasingly important topic, the efficiency of engines used for power production and transport must be increased whilst decreasing exhaust gas emissions and noise levels. From results obtained during this research project, in combination with work being produced at other research facilities, it is hoped that a greater understanding of how the leading edge region of compressor blades react to changes in engine operating points in a steady and unsteady environment is gained. This thesis investigates the boundary layer development at the leading edge of a controlled diffusion stator blade with a circular arc leading edge profile. Steady flow measurements were made inside a large scale 2D compressor cascade at Reynolds numbers of 260, 000 and 400, 000 over a range of inlet flow angles corresponding to both positive and negative incidence at a level of freestream turbulence similar to that seen in an embedded stage of industrial axial flow compressor. The instrumented blade of a large scale 2D cascade contained a series of very high resolution static pressure tappings and an array of hot-film sensors in the first 10% of surface length from the leading edge. Detailed static pressure measurements in the leading edge region show the time-mean boundary layer development through the velocity over-speed and following region of accelerating flow on the suction surface. The formation of separation bubbles at the leading edge of the pressure and suction surfaces trigger the boundary layer to undergo an initial and rapid transition to turbulence. On the pressure surface the bubble forms at all values of incidence tested, whereas on the suction surface a bubble only forms for incidence greater than design. In all cases the bubble length was reduced significantly as Reynolds number was increased. These trends are supported by the qualitative analysis of surface flow visualisation images. Quasi-wall shear stress measurements from hot-film sensors were interpreted using a hybrid threshold peak-valley-counting algorithm to yield time-averaged turbulent intermittency on the blade’s suction surface. These results, in combination with raw quasi-wall shear stress traces show evidence of boundary layer relaminarisation on the suction surface downstream of the leading edge velocity over-speed in the favorable pressure gradient leading to peak suction. The relaminarisation process is observed to become less effective as Reynolds number and inlet flow angle are increased. The boundary layer development is shown to have a large influence on the blade total pressure loss. Initial observations were made without unsteady wakes and at negative incidence loss was seen to increase as the Reynolds number was decreased and, in contrast, at positive incidence the opposite trend was displayed. The cascade’s rotating bar mechanism was used for unsteady tests where the influence of changing reduced frequency was investigated and compared to the performance of the cascade in steady operation. Results showed that increasing the stator reduced frequency brought about an increase in total blade pressure loss. The proportion of total loss generated by the suction surface increased linearly as the reduced frequency was increased from 0.47 􀀀 1.88. The opposite trend was seen on the pressure surface. These differences were attributed to the influence of wake passing events on the boundary layer in the leading edge region of both surfaces. Results from both steady and unsteady tests were compared against predictions made using the MISES solver. MISES is not capable of predicting boundary layer relaminarisation, which was shown to be clearly present on the suction surface leading to peak suction, particularly at low to moderate values of inlet flow angle and low Reynolds numbers. As a result MISES over predicts the turbulent wetted area on the blades surface when transition occurs in the leading edge region. MISES does however predict the same Reynolds number dependency of separated flow regions. The turbulent freestream flow field in the leading edge region of the blades was investigated using 2D Particle Image Velocimetry (PIV) and a commercial Reynolds Averaged Navier Stokes (RANS) based solver. The camera focused on the leading edge region of the suction surface and computational simulations were post-processed to display the equivalent field of view. Simulations were performed using the Menter Shear Stress Transport (SST) model and the Speziale, Sarkar and Gatski (SSG) Reynolds Stress Model (RSM) implemented inside ANSYS CFX. Of particular interest was how the free-stream turbulence intensity is modeled in areas of high shear strain rate, how turbulence intensity varies as the blades leading edge is approached and the production of anisotropic turbulent fluctuations close to the edge of the boundary layer. The freestream flow field was heavily influenced by the blade row displaying an increase in turbulent intermittency and significant levels of anisotropy close to the blade’s leading edge. Although the CFD did predict rises in intermittency and some anisotropy, it oversimplified the flow field and missed some of the more subtle complexities seen in the experimental results. The importance of understanding details about transition phenomena and freestream turbulence are highlighted and new insights into how they react to changes in Reynolds number and inlet flow angle are provided which can be used by code developers and users as validation. This body of work also highlights areas that still need more investigation.