Effects of surface roughness on evolutions of loss and deviation in a linear compressor cascade

Transcription

1 Journal of Mechanical Science and Technology 31 (11) (2017) 5329~ DOI /s y Effects of surface roughness on evolutions of loss and deviation in a linear compressor cascade Dongjae Kong *, Heechan Jeong and Seung Jin Song School of Mechanical and Aerospace Engineering, Seoul National University, Seoul , Korea (Manuscript Received February 27, 2017; Revised May 9, 2017; Accepted July 4, 2017) Abstract Experiments have been conducted to investigate surface roughness effects on the evolutions of loss and deviation downstream of a compressor blade row. Two cases - smooth (k + = 1.27) and rough (k + = 38.92) - have been investigated. Roughness has been attached to the blade suction side using a spray type glue gun and emery grain. At x/c = 0.2, 0.3 and 0.5 (downstream of the cascade), a five-hole probe has been used to measure the total pressure, mean velocity, and deviation at the mid-span. Static pressure taps on the blades have been used to measure the blade loading. When roughness is increased, the loss is larger and the loss increases more rapidly in the axial direction, indicating enhanced mixing. The deviation also increases at every measurement plane when roughness is increased. In addition, pitch-wise deviation distribution becomes uniform at x/c = 0.5 due to enhanced mixing. Keywords: Compressor cascade; Deviation evolution; Loss evolution; Surface roughness Introduction Gas turbines are widely used in propulsion and power generation, and, for economic and environmental reasons, the efficiency of the gas turbine remains an important research topic. Denton [1] suggested that aerodynamic loss can be divided into three components - tip leakage, end-wall, and profile loss. Tip leakage and end-wall losses are threedimensional, and two-dimensional profile loss results from the boundary layer growth on the blade surfaces. The profile loss can be further divided into that due to boundary layer development on the blades and that which occurs downstream (mixing loss). Blade surface roughness is one factor which affects loss. Compressor can be roughened by fouling from ingested aerosols like dust or sand and erosion [2]. Blade roughness can significantly change flow features which, in turn, affect inlet conditions for the next blade row. These effects accumulate through multiple stages, altering the operation point and degrading gas turbine s performance. Bons et al. [3] and Taylor [4] measured surface roughness of in-service turbine blades and quantified their characteristics. Bons [2] pointed out that surface roughness problems occur not only in turbines but also in compressors. Leipold et al. [5] reported that as the Reynolds number and incidence angle increase, roughness has a greater * Corresponding author. Tel.: , Fax.: address: superdomy@snu.ac.kr Recommended by Associate Editor Weon Gyu Shin KSME & Springer 2017 influence on the boundary layer parameters (displacement thickness, momentum thickness, shape factor, etc.) and loss. Back et al. [6] showed that loss and deviation increased as the magnitude of surface roughness increased. They also found a critical roughness magnitude where the loss abruptly increases and that surface roughness on the blade suction side has a greater impact than that on the pressure side [7]. Lorenz et al. [8] investigated the combined effects of surface roughness and free stream turbulence intensity on boundary layer behaviors at the trailing edge of a turbine blade. They reported that the momentum thickness at the trailing edge increased with augmented surface roughness, except at low Reynolds numbers. Their investigation also implied that boundary layer behavior is more sensitive to surface roughness than free stream turbulence intensity. Zhang and Ligrani [9] investigated the combined effects of surface roughness and free stream turbulence intensity on the wake turbulence structure of a turbine blade. They also found that surface roughness has a greater influence than freestream turbulence intensity on the wake turbulence structures. Results from previous research generally include measurements of aerodynamic parameters at one downstream location of a blade row. However, understanding the downstream evolutions of wake properties is also important because the axial gap is a key design parameter which sets the aerodynamic boundary conditions for the following blade row [10]. Zhang and Ligrani [11] conducted measurements at x/c = 0.25 and 1.00 for a turbine blade. However, such investigation has not

2 5330 D. Kong et al. / Journal of Mechanical Science and Technology 31 (11) (2017) 5329~5335 Table 1. Geometry of test blade. Fig. 1. Low speed wind tunnel at SNU. Nomenclature Number of blades 6 Chord, c 200 mm Span, H 196 mm Pitch, S mm Solidity, σ (= c/s) 1.19 Stagger angle, γ 50 Blade inlet angle, k Inlet flow angle, β Blade exit angle, k Exit flow angle, β2 - Fig. 2. Schematic of test section. yet been done for a compressor blade row. Therefore, to understand the surface roughness effects on the evolutions of compressor wake properties, this study presents measurements of wake properties at three downstream locations under surface roughness. Here, loss and deviation have been described to represent the wake properties. Thus, the specific research questions to be answered are: 1. How dose roughness affect loss evolution? 2. How does roughness affect deviation evolution? 2. Experimental apparatus and instrumentation 2.1 Test facility An open-type low speed wind tunnel at Seoul National University (SNU) has been used for this study. The overall schematic of the wind tunnel is shown in Fig. 1. Ambient air is drawn in by a turbo blower whose maximum volume flow rate is 620 m 3 /min and passes through a diffuser and settling chamber to reach the test section. Free stream turbulent intensity is below 1 % at the test section. A schematic of the test section is shown in Fig. 2. Six blades are mounted in the test section, and tailboards are installed to set periodic flow conditions in the pitch-wise direction. Compressor blades for the current study represent the mid-span of a compressor rotor blade from a modern power generation gas turbine. The Reynolds number based on the blade chord length is The test blade geometry is summarized in Table Instrumentation Fig. 3 shows the measurement locations. The upstream static pressure and total pressure have been measured using a Pitot-tube (United Sensor ) at 1.0 chord length upstream from the leading edge. Static pressures on the blades have been measured using the mid-span pressure taps on the 3 rd blade suction side (24 taps) and 4 th blade pressure side (18 Fig. 3. Measurement planes. taps). End-wall pressure taps covering two blade passages (y/s = 2.0) at x/c = 0.1 have been used to ensure periodicity. A cobra type five-hole probe (United Sensor ) has been traversed to measure total pressure, mean velocity, and deviation at x/c = 0.2, 0.3 and 0.5. A ± 5 kpa pressure scanner with an accuracy of 0.05 % of full scale (NetScanner, Model 9116) has been used to digitize data from the pressure taps and both pneumatic probes. The five-hole probe has been traversed for one pitch length at the mid-span by a Velmex linear traverse (Velmex Inc., BiSlide Series) between y/s = 0 and 1. The accuracy of the linear traverse is mm for a traverse distance of 250 mm. During traversing, leakage flow through the slots has been prevented by a thin plastic film. 2.3 Roughness selection Since the effects of the surface roughness on the blade suction side are dominant, the current study focuses on the suction side roughness. Equivalent sandgrain roughness (k s ) for the current study has been selected to have a value similar to the averaged roughness Reynolds number (k + ) of a real compressor blade. The procedure for the roughness selection is given in the Appendix, and selected roughness values are listed in Table 2. Clean blades without any roughness treatment have been used as the smooth (reference) case. For the rough case, four blades have been roughened by coating the suction side with

3 D. Kong et al. / Journal of Mechanical Science and Technology 31 (11) (2017) 5329~ Table 2. Selected roughness Reynolds numbers. Case k + Smooth 1.27 Rough emery grain using a spray type glue gun. Emery grain size has been selected to have a roughness Reynolds number of To avoid static pressure fluctuations in static pressure taps on the blade suction side, a 5-mm-wide mid-span strip of the suction side has been kept smooth regardless of the experimental case. 2.4 Data reduction Fig. 4. Pitch-wise end wall pressure distribution at x/c = 0.1 for k+ = The pressure coefficient is defined as P P1 Cp = P P 01 1 (1) where P 01 and P 1 refer to the upstream total and static pressures, respectively, and P is the blade static pressure measured via the static pressure taps. Loss coefficient has been calculated from the total pressure measured by the five-hole probe and defined as (, ) P01 P0 x y Yp = P P 01 1 (2) where P 0 is the total pressure measured by the five-hole probe. Deviation is the difference between the exit flow angle β 2 and the blade exit angle k 2, where the exit flow angle has been obtained by the five-hole probe. ( x y) δ = β, k. (3) 2 2 Finally, the mean downstream velocity has been normalized by the upstream velocity. From the uncertainty analysis, uncertainties (with a 95 % confidence interval) in the pressure coefficient, loss coefficient, deviation, and velocity are calculated to be 0.001, 0.002, 0.7 and 0.21 m/s, respectively. 3. Experimental results and discussion 3.1 Periodicity Fig. 4 shows the static pressure coefficient distribution of the smooth blade plotted versus blade pitch at x/c = 0.1. The pitch-to-pitch variation in C p is within 1.9 %, indicating good periodicity. Similar periodicity has been also confirmed for the roughened blade. 3.2 Mean velocity and loss Fig. 5. Pitch-wise mean velocity distribution of smooth ; rough blades at x/c = 0.2, 0.3 and 0.5. Figs. 5 and show the pitch-wise mean velocity distributions of the smooth (Fig. 5) and rough (Fig. 5) blades at x/c = 0.2, 0.3 and 0.5. For the rough case, the wake widens and the symmetric shape of the wake for the smooth case becomes asymmetric. These changes indicate increased loss. Also, the pitch-wise location of the minimum velocity shifts towards the suction side for the rough case, indicating increased deviation. The mean velocity of the core (non-wake) region remains constant for the smooth case. However, that for the rough case decreases from x/c = 0.2 to x/c = 0.5. Figs. 6 and show the pitch-wise loss distributions at x/c = 0.2, 0.3 and 0.5 for the smooth (Fig. 6) and rough (Fig. 6) cases. Similar to the wake distributions, the symmetric shape of the loss distribution for the smooth case becomes

4 5332 D. Kong et al. / Journal of Mechanical Science and Technology 31 (11) (2017) 5329~5335 Fig. 7. Influence of surface roughness on blade loading. Fig. 6. Pitch-wise loss coefficient distribution of smooth ; rough blades at x/c = 0.2, 0.3 and 0.5. asymmetric for the rough case. The loss distribution widens and increases in magnitude for the rough case. Also, the pitchwise location of the maximum loss coefficient shifts towards the suction side. These changes are due to the earlier transition to turbulent flow, increased skin friction, and separation. Earlier transition to turbulent flow on the roughened compressor blade was implied by Schreiber et al. [12], and similar phenomena have been detected on turbine blades [13, 14]. This earlier transition to turbulent flow, in turn, increases skin friction on the blade [15], leading to increased momentum defect and loss. Roughness also causes turbulent separation, another source of loss. Fig. 7 shows the influence of surface roughness on the chord-wise distribution of the blade surface static pressure coefficient. The upper lines correspond to the pressure side and the lower lines to the suction side. As roughness increases, the blade loading is decreased. The pressure coefficient distribution on the pressure side is decreased with roughness but has a similar shape regardless of the surface roughness condition. On the suction side, however, the pressure coefficient rises continuously for the smooth case, but it plateaus at around 64 % for the rough case, indicating flow separation. Leipold et al. [5] and Back et al. [7] also detected turbulent separation with a roughened compressor blade. The mass averaged loss and (estimated) fully mixed out loss are plotted versus axial distance in Fig. 8. In addition, mass averaged losses from similar previous research [5, 7] have been plotted for comparison. Estimated roughness Reynolds numbers of Refs. [5, 7] are 8.72 and 4.80, respectively, and Fig. 8. Influence of surface roughness on mass averaged loss coefficient evolution. the estimation procedure is described in the Appendix. Roughness Reynolds numbers for the current study have been selected to have a value similar to the averaged roughness Reynolds number of a real compressor blade [16], and the values are 1.27 (smooth) and (rough). Back et al. [6] reported that loss increases as the magnitude of surface roughness increases and found a critical roughness magnitude where the loss abruptly increases. As roughness Reynolds numbers from Refs. [5, 7] are smaller than that of the rough case for the current study, their mass averaged losses are smaller. However, as blade geometry and roughness types from the previous research are different from those of the current research, a direct comparison based on the roughness Reynolds number alone is not sufficient. Measurements of previous research were conducted at a single axial location. However, current study conducted measurements at three axial locations and this presents loss evolution in the axial direction under the surface roughness. The mass averaged loss coefficient for the rough case is higher at every measurement plane than that for the smooth case. For the smooth case, the mass averaged loss coefficient at x/c = 0.2 increases by 3.5 % and 8.2 % at x/c = 0.3 and 0.5, respectively. However, for the roughened blade, the corre-

5 D. Kong et al. / Journal of Mechanical Science and Technology 31 (11) (2017) 5329~ sponding increases are 21.0 % and 31.9 %. This augmented increase in the mass averaged loss coefficient indicates the enhanced role of mixing loss for the rough case. The fully mixed out loss has been estimated as follows. Using the wake profile in Fig. 5, a two-dimensional control volume analysis has been conducted to estimate the total pressure (and loss) at the fully mixed out state [17]. This analysis assumes 1) incompressible, steady flow; 2) uniform static pressure distribution along the pitch-wise direction; and 3) constant flow angle. The second and third assumptions are reasonable based on the velocity data measured at x/c = 0.5 for both smooth and rough cases. The loss at x/c = 0.5 has reached 91.1 % of the fully mixed out value for the smooth case and 95.5 % for the rough case, again indicating enhanced mixing due to surface roughness. The two possible reasons for increased mixing loss are increased turbulence intensity and turbulent separation. First, Shin and Song [18] revealed that roughness increases the Reynolds stresses inside the boundary layer of a flat plate. Furthermore, Zhang and Ligrani [9] reported that the augmented turbulence intensity due to surface roughness was still detectable at x/c = 1.0. These research indicate that the augmented turbulent intensity due to surface roughness would exist between x/c = 0.2 and 0.5 in the current study. Second possible reason for the enhanced mixing is the large eddy generation due to turbulent separation (Fig. 7). Denton [1] stated that the separation region (Fig. 7) generates large vortices, increasing loss. Thus, increased turbulent intensity and turbulent separation due to roughness are thought to enhance turbulent dissipation, increasing mixing loss. Fig. 9. Pitch-wise deviation distribution of smooth ; rough blades at x/c = 0.2, 0.3 and Deviation In addition to loss, deviation is another important cascade performance parameter. Figs. 9 and show the pitch-wise deviation distributions for the smooth (Fig. 9) and rough (Fig. 9) blades at x/c = 0.2, 0.3 and 0.5. As x/c increases, the maximum deviation magnitude decreases and the distribution maintains a peak-valley shape for the smooth case. For the rough case, the deviation increases at every measurement plane, relative to the smooth case. Larger deviation means reduced flow turning, or performance degradation [5, 19]. This performance degradation is mainly due to the increased displacement effect due to surface roughness. This phenomenon has been reported by Shin and Song [18] on a flat plate and Leipold et al. [5] on a compressor blade. Enhanced downstream mixing leads to an almost uniform deviation distribution at x/c = 0.5 for the rough case. Fig. 10 shows the mass averaged deviation plotted versus axial distance, including the estimated value at the fully mixed out state. The mass averaged deviation is higher for the rough case at every measurement plane. For the rough case, the increment of the mass averaged deviation from x/c = 0.2 to 0.3 is larger than that of the smooth case due to increased wake mixing [17]. The mass averaged deviation at x/c = 0.5 reaches Fig. 10. Influence of surface roughness on mass averaged deviation evolution. its fully mixed out value for both smooth and rough cases. Thus, deviation reaches its fully mixed out value more quickly than the loss coefficient. 4. Conclusions The aim of the current study is to investigate the effects of surface roughness on the evolutions of loss and deviation. New conclusions from this study are as follows: (1) For the rough case, the wake widens in the pitch-wise direction, and the symmetric shape of the wake for the smooth

6 5334 D. Kong et al. / Journal of Mechanical Science and Technology 31 (11) (2017) 5329~5335 case becomes asymmetric. (2) Pitch-wise loss distribution widens and increases in magnitude for the rough case due to earlier transition to turbulent flow, increased skin friction, and turbulent separation. (3) The mass averaged loss coefficient increases more rapidly downstream of the blade with surface roughness due to increased turbulence intensity and large eddies induced by turbulent separation. (4) For the rough case, the mass averaged deviation increases at every measurement plane. In addition, the pitchwise location of maximum deviation shifts towards the suction side due to the increased displacement effect of surface roughness. (5) The mass averaged deviation increases more rapidly downstream of the blade with surface roughness due to increased wake mixing. Furthermore, deviation reaches its fully mixed out value more quickly than loss coefficient. Acknowledgements Financial support from the National Research Foundation of Korea (NRF) (Project Number: ), BK21+ Program, and SNU-IAMD (Institute for Advanced Machinery and Design) are gratefully acknowledged by the authors. Nomenclature c : Chord length c f : Skin friction coefficient C p : Static pressure coefficient H : Span k 1 : Blade inlet angle k 2 : Blade exit angle k s : Equivalent sandgrain roughness k + : Roughness Reynolds number LE : Leading edge MC : Mid-chord MS : Mid-span P, P 1 : Static pressure P 0, P 01 : Total pressure R a : Centerline averaged roughness Re c : Reynolds number S : Pitch TE : Trailing edge U inlet : Upstream velocity V : Mean velocity x : Axial coordinate y : Pitch wise coordinate Y p : Total pressure loss coefficient β 1 : Inlet flow angle β 1 : Exit flow angle γ : Stagger angle δ : Deviation angle ρ : Density σ : Solidity (= c/s) References [1] J. D. Denton, The 1993 IGTI scholar lecture: Loss mechanisms in turbomachines, Journal of Turbomachinery, 115 (4) (1993) [2] J. P. Bons, A review of surface roughness effects in gas turbines, Journal of Turbomachinery, 132 (2) (2010) [3] J. P. Bons, R. P. Taylor, S. T. McClain and R. B. Rivir, The many faces of turbine surface roughness, Journal of Turbomachinery, 123 (4) (2001) [4] R. P. Taylor, Surface roughness measurements on gas turbine blades, Journal of Turbomachinery, 112 (2) (1990) [5] R. Leipold, M. Boese and L. Fottner, The influence of technical surface roughness caused by precision forging on the flow around a highly loaded compressor cascade, Journal of Turbomachinery, 122 (3) (2000) [6] S. C. Back, J. H. Sohn and S. J. Song, Impact of surface roughness on compressor cascade performance, Journal of Fluids Engineering, 132 (6) (2010) [7] S. C. Back, G. V. Hobson, S. J. Song and K. T. Millsaps, Effects of Reynolds number and surface roughness magnitude and location on compressor cascade performance, Journal of Turbomachinery, 134 (5) (2012) [8] M. Lorenz, A. Schulz and H. J. Bauer, Experimental study of surface roughness effects on a turbine airfoil in a linear cascade Part 2: Aerodynamic losses, Journal of Turbomachinery, 134 (4) (2012) [9] Q. Zhang and P. M. Ligrani, Wake turbulence structure downstream of a cambered airfoil in transonic flow: Effects of surface roughness and freestream turbulence intensity, International Journal of Rotating Machinery (2006). [10] F. E. Ames and M. W. Plesniak, The influence of largescale, high-intensity turbulence on vane aerodynamic losses, wake growth, and the exit turbulence parameters, Journal of Turbomachinery, 119 (2) (1997) [11] Q. Zhang and P. M. Ligrani, Aerodynamic losses of a cambered turbine vane: Influences of surface roughness and freestream turbulence intensity, Journal of Turbomachinery, 128 (3) (2006) [12] H. A. Schreiber, W. Steinert and B. Küsters, Effects of Reynolds number and freestream turbulence on boundary layer transition in a compressor cascade, Journal of Turbomachinery, 124 (1) (2002) 1-9. [13] M. Lorenz, A. Schulz and H. J. Bauer, Experimental study of surface roughness effects on a turbine airfoil in a linear cascade Part 1: External heat transfer, Journal of Turbomachinery, 134 (4) (2012) [14] N. Abuaf, R. S. Bunker and C. P. Lee, Effects of surface roughness on heat transfer and aerodynamic performance of turbine airfoils, Journal of Turbomachinery, 120 (3) (1997) [15] J. E. Dees and D. G. Bogard, Effects of regular and random roughness on the heat transfer and skin friction coefficient on the suction side of a gas turbine vane, Journal of Turbomachinery, 130 (4) (2008) [16] J. H. Sohn, Influence of blade surface roughness on flow

7 D. Kong et al. / Journal of Mechanical Science and Technology 31 (11) (2017) 5329~ characteristics in a linear compressor cascade, Department of Mechanical and Aerospace Engineering, Seoul National University, Seoul (2007). [17] E. M. Greitzer, C. S. Tan and M. B. Graf, Internal flow: Concepts and applications, Cambridge University Press, Cambridge, England, 3 (2007). [18] J. H. Shin and S. J. Song, Pressure gradient effects on smooth and rough surface turbulent boundary layers Part 2: Adverse pressure gradient, Journal of Fluids Engineering, 137 (1) (2015) [19] J. W. Dreon Jr., Controlled diffusion compressor blade wake measurements, Naval Postgraduate School, Monterey, CA, USA (1986). [20] C. C. Koch and L. H. Smith, Loss sources and magnitudes in axial-flow compressors, Journal of Engineering for Power, 98 (3) (1976) [21] H. Schlichting and K. Gersten, Boundary-Layer Theory, McGraw-hill, New York, USA, 7 (2003). Appendix Procedures for roughness measurement and characterization from Sohn [16] are briefly cited here. The centerline average roughness, R a, (Eq. (A.1)) of both sides of a real compressor blade (Fig. A.1) had been measured with a contact stylus system. Each side of the blade had been divided into nine regions 3 regions from hub to tip and from the leading edge to the trailing edge. Among many correlations which convert the R a value to the equivalent sandgrain roughness, k s, the Koch and Smith [20] correlation (Eq. (A.2)), appropriate for a sandgrain surface type, has been adopted. Furthermore, to account for the roughness effects under different flow conditions, the roughness Reynolds number, k +, (Eq. (A.3)) suggested by Schlichting [21] has been adopted. By matching the roughness Reynolds number, different equivalent sandgrain roughness, k s, would have the same effects on flow behavior at different Reynolds numbers. Following these procedures, the roughness Reynolds number distributions on the blade surface have been attained for both suction and pressure sides. Table A.1 shows the detailed roughness measurement results of the rough blade pressure side, suction side, and averaged quantities (c). More details can be found in Sohn [16]. R s a 1 N = Y (A.1) N i = 1 a i k = 6.2 R (A.2) k k + s = Rec c c f c c f = log. k s (A.3) Table A.1. Roughness measurement results of the rough blade. Region R a k s k + PS/LE/Tip PS/LE/MS PS/LE/Hub PS/MC/Tip PS/MC/MS PS/MC/Hub PS/TE/Tip PS/TE/MS PS/TE/Hub PS/Average Region R a k s k + SS/LE/Tip SS/LE/MS SS/LE/Hub SS/MC/Tip SS/MC/MS SS/MC/Hub SS/TE/Tip SS/TE/MS SS/TE/Hub SS/Average (c) R a k s k + Average Fig. A.1. Surface roughness of a real compressor blade. Dongjae Kong received B.S. in the School of Mechanical and Aerospace Engineering from Seoul National University. He is currently M.S. candidate at Turbomachinery Laboratory from Seoul National University. His current research interest is unsteady flow phenomena induced by impeller-diffuser interaction inside a centrifugal compressor.