m SThe Society shall not be responsible for statements or opinions advanced in papers or discussion at meetings of the Society or of its Divisions or

Transcription

1 THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 345 E. 47th St., New York, N.Y GT-260 m SThe Society shall not be responsible for statements or opinions advanced in papers or discussion at meetings of the Society or of its Divisions or Sections, or printed in its publications. Discussion is printed only if the paper is published in an ASME Journal. Authorization to photocopy for internal or personal use is granted to libraries and other users registered with the Copyright Clearance Center (CCC) provided $3/article or $4/page is paid to CCC, 222 Rosewood Dr., Danvers, MA Requests for special permission or bulk reproduction should be addressed to the ASME Technical Publishing Department. Copyright 1998 by ASME All Rights Reserved Printed in U.S.A. THE EFFECT OF INLET BOUNDARY LAYER THICKNESS ON THE FLOW WITHIN AN ANNULAR S-SHAPED DUCT Toyotaka Sonoda, Toshiyuki Arima, Mineyasu Oana HONDA R&D Co., Ltd. WAKO RESEARCH CENTER Saitama , JAPAN ABSTRACT al and numerical investigations were carried out to gain a better understanding of the flow characteristics within an annular S-shaped duct, including the effect of the inlet boundary layer (IBL) on the flow. A duct with six struts and the same geometry as that used to connect compressor spools on our experimental small two-spool turbofan engine was investigated. A curved downstream annular passage with a similar meridional flow path geometry to that of the centrifugal compressor has been fitted at the exit of S-shaped duct. Two types of the IBL (i.e. thin and thick IBL) were used. Results showed that large differences of flow pattern were observed at the S-Shaped duct exit between two types of the IBL, though the value of "net" total pressure loss has not been remarkably changed. According to "overall" total pressure loss, which includes the IBL loss, the total pressure loss was greatly increased near the hub as compared to that for a thin one. For the thick IBL, a vortex pair related to the hub-side horseshoe vortex and the separated flow found at the strut trailing edge has been clearly captured in the form of the total pressure loss contours and secondary flow vectors, experimentally and numerically. The high-pressure loss regions on either side of the strut wake near the hub may act on a downstream compressor as a large inlet distortion, and strongly affect the downstream compressor performance. There is a much-distorted three-dimensional flow pattern at the exit of S- Shaped duct. This means that the aerodynamic sensitivity of S- Shaped duct to the IBL thickness is very high. Therefore, sufficient carefulness is needed to design not only downstream aerodynamic component (for example centrifugal impeller) but also upstream aerodynamic component (LPC OGV). NOMENCLATURE A = local cross-sectional area Cp = static pressure recovery coefficient L = representative axial length of S-shaped duct, 98.6 mm (see Fig. 1) Pt = total pressure (kpa) Ps = static pressure (kpa) R = radius (mm) Tt = total temperature (K) V. = flow velocity X = axial distance (mm) = total pressure loss coefficient S* = displacement thickness (mm) 0 = momentum thickness (mm) H = shape factor Superscripts and Subscripts = time average hub = inner casing wall casing = outer casing wall max = maximum w = Wall 0,1,2 = streamwise position (see Figs. 1 & 2) INTRODUCTION A swan-neck duct is used to connect the low- and highpressure compressors of aircraft gas turbine engines. In a small gas turbine, a centrifugal high-pressure compressor is often used. In this case, the swan-neck duct is `S' shaped due to the aerodynamic design restriction of the centrifugal compressor. Presented at the International Gas Turbine & Aeroengine Congress & Exhibition Stockholm, Sweden June 2-5, 1998 This paper has been accepted for publication in the Transactions of the ASME Discussion of it will be accepted at ASME Headquarters until September 30, 1998

2 Within this duct, flow separation must be avoided to minimize the total pressure loss within the duct. In addition, a uniform flow field at the duct exit must also be achieved. However, it is very difficult to satisfy these various requirements in practice. The S-shaped duct also contains struts to support loads and passages for engine accessories and other systems, resulting in a highly complex flow field pattern due to interaction of the duct passage with the struts. Furthermore, the requirements must be fulfilled for a short axial length. Recently, some studies have been reported on S-shaped ducts (Britchford et al., 1994; Bailey et al., 1995; Bailey and Carrotte, 1996). Their long-term objective is to apply CFD methods to enable the optimum design of S-shaped ducts. Therefore, comprehensive measurements were done using an LDV system under ideal and typical actual engine inlet conditions. However, the S-shaped duct inlet velocity was relatively low and the duct itself was of constant-flow area. Only one radial strut was used in order to assess the effect on performance of placing radial struts within the duct. On the other hand, Sonoda et al. (1997) studied the flow within an S- Shaped duct with a more practical configuration and highspeed condition, experimentally and numerically. The emphasis of the earlier paper was to investigate the influence on the flow of downstream passage located at the exit of the S-Shaped duct. They showed that the total pressure loss is greatly increased near the hub in the case of the curved annular downstream passage when compared to the straight annular passage. Furthermore, a vortex related to casing-side horseshoe vortex was observed in the form of high loss region adjacent to the casing, experimentally. However, they could not detect a vortex related to hub-side horseshoe vortex as a spatial concentrated total pressure loss (as a vortex core). While, according to their numerical simulation results of the effect of IBL thickness, they qualitatively predicted the existence of a much higher loss region due to the hub-side horseshoe vortex, as the IBL thickness was thicker. However, these numerical results are not still acceptable for aerodynamic engineer as a design data, due to the uncertainty of simulation technique or the lack of quantitative estimation. In this stage of progress in research, it would be worthwhile to investigate not only the effect of the IBL thickness on the flow within an S-Shaped duct but also the limitation of in-house-developed density based Navier-Stokes code. Motivation of our study is that, according to performance test of our centrifugal compressor with an S-Shaped duct, the decrease in efficiency was much greater than expected. Therefore, project team was organized last year to investigate the flow mechanism within S-Shaped duct, and to clear the cause of efficiency drop on centrifugal compressor test. The objectives of this investigation are 1) to study the aerodynamic sensitivity of the flow within an S-Shaped duct to the IBL thickness, experimentally and numerically and 2) to evaluate the reliability of the three-dimensional density based Navier-Stokes time marching code, in more detail. EXPERIMENTAL METHODS al Facility An experimental apparatus of the S-shaped duct test rig is shown in Fig. 1. This is for the case of the thick inlet boundary layer (IBL). Air is supplied by laboratory compressors into a plenum chamber prior to passing through the S-shaped duct, and is exhausted to atmospheric pressure via an exhaust diffuser system set up in the test cell. A curved downstream annular passage, which is a similar meridional flow path geometry to that of the centrifugal compressor on our experimental small two-spool turbofan engine, has been fitted at the exit of S-shaped duct. A long inlet passage with about 200mm length has been newly fitted between the inlet contraction and the S-Shaped duct to investigate the effect of the IBL thickness on the flow within the S-shaped duct. A test rig shown in our earlier paper is for the thin IBL. That is, the long inlet passage is removed in the test for the thin IBL. The detailed configuration of the S- shaped duct and the curved downstream annular passage was described in our earlier paper. A passage (from about X/L = 0.2 to about + 0.1) upstream of the S-shaped duct leading edge has the same geometry as that of the LPC OGV. In Fig.1, Station 1 (X/L=0.0) corresponds to the trailing edge position of the LPC OGV, and Station 2 (X/L=1.0) corresponds to a leading edge of centrifugal impeller. The inlet Mach number, used in the case of the thick IBL, is ± 0.03, based on the inlet total pressure (Station 0) and the mean value of the hub and casing static pressures at Station 1. This Mach number is the same as that of the thin IBL. Static pressure taps are located at various streamwise positions on the hub and the casing, which allow for the estimation of the flow characteristic and the effect of the IBL thickness. The circumferential position of these taps corresponds to midpoint between the struts. In addition, on the strut surface, static pressure taps are located at 11%, 44%, and 89% of the strut span. The area ratio (A1/A2) of the S-shaped duct is about 1.2. The S-shaped duct has six struts with NACA 0021-profile geometry. Instrumentation Inlet total pressure and temperature are measured at the center of the plenum chamber (at Station 0 in Fig. 1). At the outlet, using the 3- axis traverse mechanism, traverse measurements of pressure were made using a miniature five-hole pressure probe with an overall diameter of 1.5 mm, which was calibrated in advance. The axial position of these measurements (X/L=1.03) is nearly corresponding to 2

3 Plenum Chamber IJ Boundary layer Pressure Probe Five-hole pressure probe m J11T Inlet Contraction X/L L.98. 6mm 1.0 Station 01 Station 0-t Station i Station 2 long Inlet Passage S-Shaped Duct Curved Downstre Annular Passage Fig. 1 al Apparatus for Thick Inlet Boundary Layer the centrifugal impeller leading edge (Station 2). The outer ring of curved annular passage moves in the circumferential direction by the 3-axis traverse mechanism. The data are mainly obtained at eight radial positions traversed along 20 circumferential points. The area traversed corresponds to half of the strut pitch. The measurements of the two types of the IBL profile (i.e. the thin and the thick IBL) are done at Station 1 using a miniature boundary layer pressure probe with 0.4mm (height) x 0.85mm (width). Furthermore, additional boundary layer measurements are done at Station 0-1 in order to get inlet boundary pressure conditions for CFD in the case of the thick IBL. The flows on the strut and the hub/casing surfaces have been visualized using a mixture of titanium dioxide and oleic acid. Data Reduction Static pressures along the hub, the casing and the strut surface are given in terms of the pressure recovery coefficient (Cp). While, the total pressure loss coefficient between the S- shaped duct inlet (Station 1: X/L=0.0) and the duct exit (X/L=1.03: near Station 2) is defined as. C Ps1. ^s P 1P,2 p! P,1 Ps,l P,1 Ps,l Ps,l a 2 (Ps,1,casing + Ps,l,hub ) Here, the total pressure at Station 1 (Pt, 1) has been assumed the same as the pressure at Station 0 (Pt,,) in the plenum chamber. The total pressure loss coefficient obtained from the above assumption is called "overall" total pressure loss, which includes the loss of the IBL. On the other hand, total pressure loss that does not include the loss of the IBL is called "net" total pressure loss. The "net" loss is obtained using the measured mass-averaged total pressure at Station 1(Pt, 1) instead of Pt,,. The five-hole pressure probe provides information on total and static pressure as well as flow direction (pitch and yaw angles) and velocity is calculated from these data. Estimate of al Error The inlet total pressure probe, the boundary layer pressure probe, the static taps and the five-hole pressure probe discussed above were connected to a pre-calibrated differential pressure transducer which had a range of ± 98 kpa. The output was read automatically from an integrating digital voltmeter. Total and static pressure measurements were reproducible to within ± 10 mm of water. Based on these values, it was estimated that the static pressure recovery and the total pressure loss coefficient were reproducible to within -t 0.02 and ± 0.01, respectively. COMPUTATIONAL METHODS A three-dimensional compressible density based Navier- Stokes code with a low Reynolds number k-c turbulent model, Arima et al. (1997), has been applied to the flow within the S- shaped duct. The computational body-fitted grid, for example, used for the thick IBL is shown in Fig. 2. Part of the casing 3

4 grid is omitted to allow a better view of the S-shaped duct with struts. The grid consists of 69 nodes in the strut-to-strut direction, 69 nodes in the spanwise direction, and 251 nodes in the streamwise direction. High grid density has been used because of the increase of computational accuracy. That is, the change of total pressure between S-Shaped duct inlet and outlet is a very small fraction of the inlet total pressure. Grid sensitivity study had been performed by focussing on the distribution of total pressure loss coefficient obtained from different grid systems. The coarse grid had sometimes gave locally a numerical error, such as a negative loss coefficient. With the increasing of grid points (in especially radial direction), the unreasonable physical phenomena had disappeared. Therefore, such a high-density grid, as used in this work (251x69x69), has been adopted. For proper resolution of the boundary layer, typical value less than 5 should be required. In this work, except in the immediate vicinity of the leading and trailing edges, the strut, hub, and casing surface boundary layers had at least 11 grid points with values of y+ < 20. This yielded values of y *al at grid points adjacent to the walls; and then its average value of whole grid points adjacent to the walls had become y < 2. s have been done for two types of IBL thickness. For the case of the thick IBL, as inlet boundary conditions, the measured pressure and temperature at Station 0-1 were used, and inlet flow direction was along the hub and the casing without swirling. On the other hand, for the case of the thin IBL, the grid was partially changed. That is, a grid with short inlet passage (from X/L= 0.62 to 0.0 (Station 1)) was used instead of the long inlet passage (from X/L= 0.92 to 0.0 (Station 1)). This is, 251 x 69 x 69 1? Station 1 X/L= (Station 0-1) Fig. 2 Computational Grid of S-Shaped Duct for Thick IBL because the experimental data as inlet boundary conditions for the thin IBL could not be gotten due to the restriction of test rig layout. Therefore, the calculation for the thin IBL was repeated until the calculated IBL thickness of Station 1 was in good agreement with that of the measurement, by changing the IBL thickness at X/L= The initial total pressure profile in the IBL was calculated automatically using a 1/7-th-power velocity profile law and was used as an inlet total pressure boundary condition. Downstream back pressure was fixed at hub and casing due to no swirling and adjusted to coincide with the corrected mass flow of Station 1. RESULTS & DISCUSSION Inlet Boundary Layer In order to estimate the characteristic of the IBL at Station I for the thin and thick IBL, the IBL measurements were done using the miniature boundary layer pressure probe. The obtained total pressure data were transformed to velocity profile using static pressure values measured on the hub and the casing. Figure 3 (a) and (b) show the total pressure profile and the velocity profile at Station 1, respectively. In both cases (the thin and the thick IBL), as the flow is affected by a curvature effect in this station, the velocity-shear flow field is generated. The thickness of IBL on the hub and casing is of the same order, and it is about 5% and about 30% of passage height for the "thin" and the "thick" IBL, respectively. Calculated results are also shown in Fig. 3. When contrasting both results (experiments and calculations), there is a difference regarding the boundary layer profile. This is due to the use of the 1/7-th-power velocity profile as an inlet boundary condition at Station 0-1, instead of the measured boundary layer total pressure profile. However, the tendency between measurements and calculations is quite the same, though there is a little deficit of inlet total pressure from about 30 to 60% span height. The boundary layer parameters (S*, 0, H) were obtained under the extrapolation of the main flow velocity (i.e. the velocity of no pressure loss region), as shown in the form of "straight lines" in Fig. 3 (b). Table 1 shows the measured boundary layer parameters. It is apparently clear that the IBL at Station I is turbulent flow, because the shape factor HE 6*/0 is sufficiently small as compared to 2.59 (laminar flow along a plate) or 1.30 (turbulent flow along a plate). b * _ 0 (1- V )dr V max 9 = f (1- V ) V dr V max V max 4

5 Table 1 Measured Boundary Layer Parameters for Thin and Thick IBL at Station Station 1 =60 CAL EXP E Thin IBL 0 : Thin IBL Thick IBL A : Thick IBL m a v Q^ - IBL - SIDE 8 *(mm) e (mm) H Thin CASING HUB Thick CASING HUB E 0 40 m a U) Pt / Pt, 0 (a) Total Pressure Profile V / V max (b) Velocity Profile Fig. 3 Boundary Layer Profiles for Thin and Thick IBL Flow Visualization at Station 1 Flow has been numeric visualized experimentally and ally to gain a qualitative understanding of the flow pattern within the S-shaped duct. The representative results are shown in Fig. 4. Figure 4 (a) shows the flow pattern along the strut and hub surface. A typical horseshoe vortex that forms ahead of the blunt-strut leading edge with a saddle point is observed. From this point, two limiting streamlines emerge and wrap around the strut. As the limiting streamline is traced downstream, it is seen to move toward the strut surface and to be migrating in the spanwise direction on the strut surface. A small separation region is observed on the strut surface at the hub corner of the trailing edge, experimentally and numerically. When contrasting the two IBL (i.e. the thin and the thick IBL), the difference of each saddle point's position is not so large. In the case of the thick IBL, the migration levels of low momentum fluid on the strut surface are smaller than that of the thin IBL, experimentally and numerically. This may be due to a developed counter-clockwise hub-side horseshoe vortex (see Figs. 5 and 9)) and due to a decreased radial pressure gradient (see Fig. 6). The calculated flow pattern on the hub surface has very small shear stress in the both cases, as compared to that for a straight downstream passage (Sonoda at al. 1997). This is mainly due to a high positive streamwise pressure gradient on the hub in the latter half of the S-Shaped duct. The pressure gradient measured for the straight downstream passage in our earlier paper is the limitation for the separation on the hub. The horseshoe vortex observed on the hub can also be seen on the casing as shown in Fig. 4 (b). The strut has been removed. This feature is, however, very different from that of the hub. That is, the distance between the limiting streamline and the strut surface is almost constant within the S-Shaped duct. while at the trailing, two-wake limiting streamlines form. In the two types of IBL, the feature of the flow pattern for the thick-ibl is somewhat different from that for the thin-ibl. First, the saddle point is moved toward the upstream direction as the IBL thickness has been increased. Second, though the shape of wake limiting streamline is almost same for the two types of IBL, the distance between the limiting streamline related the horseshoe vortex and the wake limiting streamline is significantly increased. On the whole, the calculated results are in good agreement with the experimental results, qualitatively. Figure 5 numerically shows the relation between the oil flow pattern and the hub-side horseshoe vortex. This is in the case of the thick IBL. As already described, though the limiting streamline is seen to be migrating in the spanwise direction on the strut surface (in Fig. 4 (a)), the hub-side horseshoe vortex keeps a constant distance from the strut surface. The spatial development of the hub-side horseshoe vortex becomes to be noticeable from the latter half of the strut, due to a positive pressure gradient on the hub, as shown later in Fig. 6. 5

6 Thin IBLI hick IB (a) Flow Pattern along Strut and Hub Surfaces hin IBLI hick IBLI (b) Flow Pattern along Casing Surface Fig. 4 al and Numerical Flow Visualization Results for Thin and Thick IBL 6

7 Flow Pattern yview S0.0 L 96. 6mm 1.0 \ Z^" t t 1),' ff 1^ 1 Total Pressure Loss Contours X/L(Axial Chord %) IIE: ^` Mt f ` C (58%) I r 0.76 (100%:T.E.) Hub, Casing and Strut Static Pressure The hub and casing wall pressure coefficient, Cp, is presented in Fig. 6 for two cases of the IBL. As the flow follows a curved path within the S-shaped duct, a modification to the static pressure field occurs due to the balance between centrifugal force and radial pressure gradient. Across the first bend ( Station 1), the pressure close to the casing is higher than that adjacent to the hub. However, this situation is reversed at the second bend since the flow follows the downstream curved passage fitted at the exit of S-Shaped duct. The flow experiences a significant positive streamwise pressure gradient in the latter half of the duct on the hub surface. In contrast, the pressure gradient is almost negative adjacent to the casing where Cp decreases along approximately 80% of the duct length. When contracting both cases (i.e. the thin and thick IBL), Cp profiles show significant, though not dramatic, differences in the latter half of the S- Shaped duct. Radial pressure gradient from hub to casing is decreased adjacent to the S-Shaped duct exit in the case of the thick IBL. The cause is the less curvature effect due to the thick boundary layer. As shown in Fig. 6, the calculated results are in good agreement with experimental results. The strut wall static pressure coefficient, Cp, is presented in Fig. 7 at 11%, 44% and 89% of the strut span. There is not a noticeable effect of the IBL thickness on the strut surface static pressure. The tendency between the measurement and the calculation is quite the same. Fig. 5 Comparison of Hub Side Oil Flow Pattern 2 with Hub Side Horseshoe Vortex Measurement 0 : Thin IBL 0 : Thick IBL : Thin IBL : Thick IBL 1.5 to to 1.5 n U C N U 0.5 a U^ 1 C U_ a) N 0j 0.5 a) 0, Co a a` 0 U 0 Hub Cs U Cs U) 11 % Span e o 44 % Span -0.5 Measurement O : Thin IBL :: Thick IBL : Thin IBL : Thick IBL Casing -0.5 O 89 %Span Axial Position X/L Axial Position X/L Fig. 6 Static Pressure Distribution on Hub and Casing for Thin and Thick IBL Fig. 7 Static Pressure Distribution on Strut Surface for Thin and Thick IBL 7

8 Thin IB hick IBL] Rno% A nnl 0 60% 80% 20% 30 Strut Mid-Position Mid-Position 0 5 nm 0 5 Mid- Position Mid - Position Fig. 8 Total Pressure Loss Contours at S-Shaped Duct Exit (-Station 2) for Thin and Thick IBL 8

9 Thin lblj hick IB LI 0,n 0 5,A Mid Position Mid Position Casing 30 Casing 30 t ts,.,, t,rt,,..,,#f t l s. t l l t t, r,..,, rn._..^sttr r, f ^ttrr,rr,, - fl ftlrrrr,, r^ fitrrr r,,, (f rfsrrr,,,, Mltttf r, - tt X111 ru.,. 1 11r t f r. l u+r 1 1tt,, Strut Hub Mid-Position Hub '` Mid-Position Fig. 9 Secondary Flow Vectors at S-Shaped Duct Exit ( Station 2) for Thin and Thick IBL 9

10 Total Pressure Loss Coefficient al and calculated contours of the total pressure loss coefficient are shown in Fig. 8 for thin IBL (left) and the thick IBL (right), obtained at the S-shaped duct exit (X/L=1.03). The calculated contours have an increment of 5%. According to the experimental results in the both cases (i.e. the thin and the thick IBL), the boundary layer near the hub at the midpoint between the struts is thicker than that near the casing. This is due to the streamwise positive pressure gradient along the hub in the latter half of the duct, as already shown in Fig. 6. When contrasting both cases (i.e. thin and thick IBL), the boundary layer thickness near the hub and casing region is greatly increased in the case of the thick IBL. At the same time, the higher loss region is observed, as a large total pressure hole, on either side of the strut wake near the hub in the case of the thick IBL. This "hole" seems to be related to the inlet hub-side horseshoe vortex. The calculated total pressure losses in Figure 8 are much larger than indicated in the experimental contours, especially for the thick IBL. This is probably due to the difference of inlet boundary layer total pressure profile at Station 1 between experiment and CFD (see Fig. 3(a)). That is, total pressure loss in CFD is larger than that of the experiment. In our CFD calculation, the boundary layer thickness is adjusted to be the same value of the experiment, using a 1/7-th-power velocity profile law. Furthermore, in the case of the thick IBL, the circumferential variation-region of the total pressure loss adjacent to the casing is more extended radially-inward. The maximum of total pressure loss at about 5 degrees corresponds to the "wake limiting streamline", as already shown in Fig. 4(b). This deformation of the total pressure loss is due to the casing side horseshoe vortex. Secondary Flow Vectors As already reported in our earlier paper, the secondary flow vectors indicated radially-outward flow due to the hub passage curvature at the S-shaped duct exit (at about Station 2: X/L=1.03). Therefore, the flow features are not clear due to the above high radially-outward flow. In this study, the radial velocity vectors at the various spatial points have been subtracted from that of each radial position at the mid-position in order to clear the flow feature. Figure 9 shows the experimental and calculated secondary flow vectors at the S-shaped duct exit. The flow features are very clear, especially for the thick IBL. A vortex related to a casing-side horseshoe vortex has been clearly captured for both thin and thick IBL in the form of velocity vectors, experimentally and numerically. Furthermore, a single vortex (clockwise) has been observed near the hub wall for the thin IBL. The hub-side single vortex is induced in the separated flow found at the strut trailing edge. On the other hand, in the case of the thick IBL, a vortex pair is noticeable in both experimental and numerical velocity vector results. Vortical motion with counterclockwise corresponds to the hub-side horseshoe vortex. It is the counter-rotating action of the two vortices that is pumping the low total pressure fluid from the near-hub wall boundary layer into the passage that causes the "hole". On the whole, the experimental and calculated results show good qualitative agreement. It is very clear that the effect of the IBL thickness on the flow within an S-Shaped duct is very high. In other word, the aerodynamic sensitivity of S-Shaped duct to the IBL thickness is very high. Especially, the hub-side horseshoe vortex may have an adverse effect on the downstream centrifugal compressor performance, such as a large inlet distortion. Since, in a real engine, the inlet flow upstream of S-Shaped duct may be much distorted and occupied with a poor IBL than our "thick" IBL used in this study. Therefore, sufficient carefulness is needed to design not only downstream aerodynamic component (for example, centrifugal impeller) but also upstream aerodynamic component (for example, LPC OGV). Overall and Net Total Pressure Loss Figure 10 shows the spanwise distribution of massaveraged "overall" total pressure loss coefficient. The total pressure loss near the hub is greatly increased in the case of the thick IBL, experimentally and numerically. As already described in "Data Reduction", the overall total pressure loss coefficient includes the pressure loss within the IBL developed along the inlet upstream passage. Therefore, it is worthwhile to subtract the upstream IBL total pressure loss from the results in Fig. 10. Table 2 shows not only the overall mass averaged total pressure loss coefficient but also the net total pressure loss coefficient for the thin and the thick IBL. The measurement neglects the high loss regions in the annulus boundary layer. Therefore, the calculations had been processed in the same way as the experiments. The "overall" loss of the measurements increases from 6.4% to 11.8% as the IBL thickness increases. While, the "net" total pressure loss increases from 5.4% to 7.2%. The rate of increase for the net E 0 c40 m a to 20 0 O ^' Measurement O : Thin IBL A : Thick IBL "" : Thin IBL :Thick IBL Total Pressure Loss Coefficient : A (%) Fig. 10 Spanwise Distribution of Overall Total Pressure Loss Coefficient at S Shaped Duct Exit ( Station 2) 10

11 Table 2 Comparison between Mass Averaged Overall and Net Total Pressure Loss Coefficient A (%) Thin IBL Thick IBL Overall Measurement Net Measurement loss is about 33%. Though there is no dramatic difference for the "net" pressure loss between the thin and thick IBL, it is apparent that the flow pattern within S-Shaped duct is remarkably affected by the IBL thickness. In comparing the experimental and calculated results, there are discrepancies in the spanwise distribution of overall total pressure loss and in the "net" total pressure loss values for the thin and the thick IBL. Both results, however, indicate the same overall behavior. CONCLUDING REMARKS al and numerical investigations were carried out to gain a better understanding of the flow characteristics within an annular S-shaped duct with the curved annular downstream passage, including the effect of the inlet boundary layer (IBL) thickness on the flow. The following conclusions were drawn. (1) A vortex related to a casing-side horseshoe vortex has been clearly captured for both thin and thick IBL in the form of velocity vectors, experimentally and numerically. Furthermore, a single vortex has been observed near the hub wall for the thin IBL. The hub-side single vortex is induced in the separated flow found at the strut trailing edge. (2) When contrasting both cases (thin and thick IBL), the spatial concentrated high loss region (i.e. the large total pressure "hole") has been generated near the hub wall for the thick IBL. This is due to a vortex pair. It is the counter-rotating action of the two vortices that is pumping the low total pressure fluid from the near-hub wall boundary layer into the passage that causes the "hole". At the same time, high total pressure loss region near the casing has been extended radially-inward. (3) Large differences of flow pattern have been observed at the S-Shaped duct exit between two types of the IBL, though the value of "net" total pressure loss has not been remarkably increased. Aerodynamic characteristic of the flow within an S-Shaped duct is greatly affected by the IBL thickness. (4) A vortex pair near the hub wall may have an adverse effect on the downstream centrifugal compressor performance, such as a large inlet distortion. Since, in a real engine, the inlet flow upstream of S-Shaped duct may be much distorted and occupied with a poor IBL than our "thick" IBL used in this study. Therefore, sufficient carefulness is needed to design not only downstream aerodynamic component (for example, centrifugal impeller) but also upstream aerodynamic component (for example, LPC OGV). (5) Calculated results using the three-dimensional Navier- Stokes code with a low Reynolds number k-c turbulent model are in good agreement with experimental results. However, quantitative discrepancies are still observed in the wake region. ACKNOWLEDGEMENTS The authors would like to acknowledge Dr. Jon Carrotte of Loughborough University of Technology for useful discussions. We also thank Mr. Junji Takado and Mr. Yasuyoshi Oke of HONDA R&D for their excellent support in obtaining the data presented here. REFERENCES Arima, T., Sonoda, T., Shirotori, M., Tamura, A., and Kikuti, K., 1997, "A Numerical Investigation of Transonic Axial Compressor Rotor Flow Using a Low Reynolds Number k-a Turbulence Model", ASME Paper No. 97-GT-82. Bailey, D. W. and Carrotte, J. F., 1996, "The Influence of Inlet Swirl on the Flow within an Annular S-Shaped Duct", ASME Paper No. 96-GT-60. Bailey, D. W., Britchford, J. F., Carrotte, J. F., and Stevens, S. J., 1995, "Performance Assessment of an Annular S-Shaped Duct", ASME Paper No. 95-GT-242. Britchford, K. M., Carrotte, J. F., Stevens, S. J., and McGuirk, J. J., 1994 "The Development of the Mean Flow and Turbulence Structure in an Annular S-Shaped Duct", ASME Paper No. 94-GT-457. Sonoda, T., Arima, T., and Oana, M., 1997"The Influence of Downstream Passage on the Flow within An S-shaped Duct" ASME Paper No. 97-GT