International Journal of Heat and Fluid Flow 50 (2014) 467–478

Contents lists available at ScienceDirect

International Journal of Heat and Fluid Flow

journal homepage: www.elsevier .com/ locate/ i jhf f

Effect of boundary layer thickness on secondary structures in a shortinlet curved duct

http://dx.doi.org/10.1016/j.ijheatfluidflow.2014.10.0160142-727X/� 2014 Elsevier Inc. All rights reserved.

⇑ Corresponding author.E-mail address: amitam@rpi.edu (M. Amitay).

1 Department of Mechanical, Aerospace & Nuclear Engineering, 110 8th Street,Troy, NY, USA.

2 James L. Decker’45 Endowed Chair of Aerospace Engineering, and Director of theCenter for Flow Physics and Control (CeFPaC), 110 8th Street, Troy, NY, USA.

Jeremy Gartner 1, Michael Amitay 2,⇑Rensselaer Polytechnic Institute, Troy, NY 12180, USA

a r t i c l e i n f o

Article history:Received 4 February 2014Received in revised form 15 October 2014Accepted 17 October 2014Available online 13 November 2014

Keywords:S-shape inletSecondary flowsThree dimensional separationSecondary flow structures

a b s t r a c t

The flow pattern in short inlet ducts with aggressive curvature has been shown to lead, in some cases, toan asymmetric flow field at the aerodynamic interface plane. In the present work, a two-dimensionalhoneycomb mesh was added upstream of the curved duct to create a pressure drop across it, and there-fore to an increased velocity deficit in the boundary layer. This velocity deficit led to a stronger stream-wise separation, overcoming the instability that can result in an asymmetric flow field at theaerodynamic interface plane. Experiments were conducted at Mach numbers of M = 0.2, 0.44 and 0.58in an expanding aggressive duct with rectangle to a square cross section with area ratio of 1.27. Steadyand unsteady pressure measurements, together with Particle Image Velocimetry (PIV), were used toexplore the effect of the honeycomb on the symmetry of the flow field. The effect of inserting a honey-comb was tested by increasing its height from 0 to 2.2 times the boundary layer thickness of the baselineflow upstream of the curve. Using the honeycomb, flow symmetry was achieved for the specific geomet-rical configuration tested with a negligible decrease of the pressure recovery.

� 2014 Elsevier Inc. All rights reserved.

1. Introduction

Heightened interest in short and curved inlet ducts for aircrafthas led us to further explore the flow field existing in such devicesas well as to a better knowledge of the issues associated with theirdesign and methods of mitigating such issues. Several factorsrelated to engine and aircraft performance and operation drivethe use of short inlet duct designs, such as the overall airframelength reduction enabled by a shorter duct and reduction of frontalplanform by burying the engine into the airframe. These factors ledto a reduction in weight and fuel consumption, and allowed forinnovative external and integrated aerodynamics, such as BlendedWing Bodies (Dagget et al., 2003). Another factor that must betaken into consideration is the stability margin for operation of ajet engine following the duct, where uneven pressure distributionand secondary flow structures can lead to engine stall at the fan/compressor stages (surge stall) (Scribben et al., 2006; Mattinglyet al., 2002).

A considerable body of work is available in the literature con-cerning the analysis of the flow field in short inlet ducts (Bansodand Bradshaw, 1972; Launder and Ying, 1972; Enayet et al.,1982; Wellborn et al., 1992, 1993; Whitelaw and Yu, 1993a,b; Nget al., 2008, 2006). These previous research efforts have shed lighton the main features of the flow field existing in aggressivelycurved ducts, where the rapid curvature in the duct results in pres-sure gradients in the direction normal to the turn, leading to theonset of secondary flow structures in the form of two counterrotating vortices. Other structures were also noticed to co-existwith these counter-rotating vortices, such as cross stream flow atthe internal surfaces, which invade the local boundary layer lead-ing to further flow detachment (Wellborn et al., 1992, 1993; Nget al., 2008, 2006; Chen, 2012) disrupting the flow and creatingrecirculation zones in the duct. The symmetric counter-rotatingvortices can be described by inviscid flow equations, caused solelyby the turning of the flow. These pressure driven counter-rotatingvortices convect the low momentum fluid of the boundary layertowards the center of the duct impacting flow uniformity and pres-sure recovery at the face of the engine located downstream, at theaerodynamics interface plane.

Implementation of passive and active flow control techniques inshort inlet ducts has been an active field of research (Scribbenet al., 2006; Ng et al., 2008, 2006; Chen, 2012; Vaccaro, 2011;Debronsky, 2012; Amitay et al., 2002; Gissen et al., 2011; Jirasek,

Nomenclature

AIP aerodynamic interface planeD width of the duct (mm)h honeycomb height (mm)d boundary layer thickness (mm)h/d relative height of the honeycombh momentum thickness (mm)L length of the duct (mm)Minlet Mach number at the inlet

PR pressure recoveryPinlet static pressure at the inlet (atm)c specific heat ratio of airCp pressure coefficientP0 total pressure (atm)P static pressure (atm)

468 J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478

2006; Reichert and Wendt, 1994). The predominant forms of actu-ation are vortex generators, steady and unsteady jet blowing tan-gent to the surface, synthetic jet actuators, and many more.Recent work has studied multiple actuation devices (Vaccaro,2011; Gissen et al., 2011) including combination of flow controltechniques. Although most of the previous work was focused oncircular cross section ducts (Wellborn et al., 1992, 1993;Whitelaw and Yu, 1993a,b; Gissen et al., 2011; Jirasek, 2006),emphasis was also given to rectangular cross section ducts(Launder and Ying, 1972; Ng et al., 2008, 2006; Chen, 2012;Vaccaro, 2011; Debronsky, 2012; Amitay et al., 2002). All of thework performed with flow control had the objective of improvingthe pressure recovery and pressure distribution at the exit of theduct.

Recent experiments (Vaccaro, 2011; Debronsky, 2012) andnumerical simulations (Chen, 2012) have shown that, under somegeometrical conditions, the flow can become asymmetric, whereone of the counter-rotating vortical structure supersedes the other.In the case of the rectangular ducts, the secondary flow structureswere shown to move towards one of the corners of the duct (Chen,2012; Vaccaro, 2011; Debronsky, 2012).

As noted by Chen (2012), the secondary flow phenomenon(i.e., a turbulent flow with mean streamwise vorticity) is attrib-uted to two mechanisms: (i) the skew induced, inviscid mecha-nism, which is caused by any bend in the flow path of ductswith any cross sectional shape as shown by Miller (1991) (andFig. 1 below), and (ii) a stress-induced mechanism occurring inany non-circular ducts, straight or not, due to anisotropy of theReynolds stresses. A more in-depth description of secondary flowcan be found in Perkins (1970) and Bradshaw (1987). Also notethat further complexity in the flow structures is due to swirldevelopment in the second bend of the s-duct. This reverse inthe curvature is accredited with the crossover of the transversevelocity component near the side walls, an essentially inviscid

Fig. 1. Development of secondary flows in a pipe bend showing the (a) presence of an advMiller, 1991).

process. Another feature of short inlet ducts is the adverse pres-sure gradient caused by the opposite curvature of the secondbend. Therefore, the secondary flow generated by the first bendis attenuated, being reversed depending on the aggressivenessof the turn (i.e., the aspect ratio L/D, the offset and area ratiobetween the inlet and the exit sections).

Fig. 2a (taken from Wellborn et al., 1993) shows a three-dimen-sional perspective of the owl face separation topology with a coun-ter-rotating pair of vortices orientated with upwelling along thecenterline. The skeleton drawing of Fig. 2b shows the schematicof a symmetric but unstable owl face separation of the first kind.Perry and Hornung (1984) suggested that this unstable symmetricdistribution could exist due to slight variations in the flow field.However, they stated that it was a special condition and that anyasymmetry in the flow would cause the streakline pattern to shiftto one side as shown in Fig. 2c.

Based on previous experiments (Ng et al., 2008, 2006; Vaccaro,2011) and numerical analysis (Chen, 2012), it was shown that acritical length to diameter ratio exists in a rectangular cross sectioncompact inlet duct, controlling the asymmetry of the secondarystructures. For ducts longer than this critical length the flow pat-terns are asymptotically stable. With the reduction of duct lengthbelow this critical value, the streamwise pressure gradientincreases and interacts with the transverse invasion. Simulta-neously, the forward moving main flow confronts the backflowto the streamwise separation. The symmetric pattern becomesunstable due to the saddle–saddle connections existing in thetopology of the flow, leading to a flow bifurcation (Tobak andPeak, 1982).

This asymmetric configuration is the starting point for the cur-rent work, which also derives from the observation (Enayet et al.,1982; Chen, 2012) that the secondary flow structures forming onthe inside of duct bends have their strength dependent on inletflow conditions, specifically the momentum thickness of the

erse pressure gradient, and (b) direction and orientation of the secondary flow (from

Fig. 2. Owl face separation of the first kind topology showing the (a) three-dimensional perspective of the separation, (b) symmetric skeleton schematic of streaklines, and (c)skeleton schematic of asymmetric streakline pattern (from Wellborn et al., 1993).

Fig. 3. The experimental facility.

Fig. 4. Exploded view of the inlet duct assembly.

∗

Fig. 5. Inlet duct with the Instrumentation Can for AIP measurements.

Fig. 6. Pressure sensors distribution at the AIP.

J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478 469

Honeycomb MeshQuarter of a dollar

Fig. 7. Compact inlet with the honeycomb inserted.

Fig. 8. Internal view with the honeycomb inserted.

Fig. 9. Inlet total pressure field measured at M = 0.44; (a) across the duct height,and (b) across the duct span.

Fig. 10. Cross-stream distributions of the streamwise velocity profiles upstream ofthe turn at M = 0.44.

Fig. 11. Effect of the honeycomb height on the shape of the velocity profile at x/D = �0.58.

Fig. 12. Effect of the honeycomb height on the shape of the velocity profile at x/D = �0.14.

Table 1Momentum displacement at x/D = �0.58.

h/d0 h (mm)

0 0.550.74 0.951.47 1.042.2 1.16

470 J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478

incoming boundary layer. Therefore, the objective of the currentwork is to show that proper manipulation of the incoming bound-ary layer can lead to modification of the flow and pressure fieldswithout incurring any significant pressure loss. The main gainexpected from the manipulation of the incoming boundary layeris on the cross flow symmetry of the pressure distribution at theAIP.

Fig. 13. Streamwise pressure coefficient distribution for the three different honeycomb actuator heights at M = 0.2.

J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478 471

2. Experimental setup

The high subsonic facility at the Center for Flow Physics andControl (CeFPaC) at Rensselaer Polytechnic Institute was utilizedin the current experiments. The high subsonic facility is a blowdown, open return wind tunnel. The inlet duct geometry allowedfor Mach numbers up to 0.6 although most data presented wasconducted at a Mach number of 0.44 (mass flow rate up to1.76 kg/s). Fig. 3 shows the flow path of the facility and labels eachcomponent. The air begins in the blower and transitions throughthe blower diffuser into the settling chamber. Next, it enters con-traction followed by the inlet duct (i.e., the test section), and finallyexits the facility through a diffuser.

The blower used is a Cincinnati Fan model HP-12G29 run by a100 HP motor that is controlled by a variable frequency drive.The blower can produce a volumetric flow rate up to 170 m3/min. The air exits the blower and enters the diffuser section thattransitions the circular cross-section of the blower to the squarecross-section of the settling chamber. The air is slowed as it entersthe settling chamber where the fluid is conditioned through a setof screens and honeycomb. In addition, a thermocouple and a staticpressure ring were monitored for all experiments and were used inthe calculation of the Mach number. The air then enters the con-traction section with a contraction ratio of 142:1 and a conven-tional 5th order polynomial curvature.

The air then enters the inlet duct, which has a constant cross-section with a length of 0.3048 m for boundary layer growth as

well as to measure the inlet Mach number. Another static pressurering, as well as a thermocouple, was instrumented in this constantcross-section section. Utilizing a one-dimensional isentropic flowassumption, the inlet Mach number was found from the staticpressure ring in the inlet as well as the static pressure ring of thesettling chamber. This assumption was validated in the previouswork done by Vaccaro (2011) utilizing a total pressure Kiel probein the constant cross-section region of the inlet duct. The totalpressure of the inlet nearly matched the pressure measured inthe static pressure ring of the settling chamber, where the airvelocity is small enough to be assumed to be zero allowing thepressure and temperature measured in the settling chamber tobe taken as the total quantities. From the total quantities measuredin the settling chamber and the static pressure measured in theinlet, the Mach number from the isentropic flow assumption is:

Minlet ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiP0

Pinlet

� �c�1c

� 1

" #� 2

c� 1

� �vuut ð1Þ

Following the constant area section of the inlet is the diffusingS-shaped section. After the air goes through the inlet duct, it exitsthe wind tunnel through a diffuser. This diffuser angle is 3� toreduce flow speed as it exits into the open room.

The inlet duct has a length-to-diameter ratio of 1.6, where theinitial rectangular cross-section area of 90 mm tall by 114.3 mmwide transitions to a square cross-section 114.3 mm by114.3 mm resulting in an area ratio of Aexit/Ainlet of 1.27. The design

Fig. 14. Streamwise pressure coefficient distribution for the three different honeycomb actuator heights at M = 0.44.

472 J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478

was modular to allow for easy removal and exchange of parts ascan be seen in Fig. 4, which shows an exploded view of the inletduct. For example, two sets of windows were fabricated: one setmade of aluminum for most of the experiments and the otherout of optical grade acrylic solely for the use of PIV. This reducedthe chance of scratches occurring on the windows that would havea negative impact on the PIV quality and potentially causing glaresand/or inaccuracies in the collected data.

The inlet duct was instrumented with 126 static pressure tapsalong the lower surface of the duct. The spanwise density of statictaps increased towards the side-wall with an array on the oppositeside for checking symmetry (see Fig. 4). The static pressure tapswere sampled by the means of four pressure scanners (ScanivalveDSA3217, 16 channels each, ±5 psid full scale and accuracy of±0.05% of full scale, or ±0.0025% psi). Because only 64 ports couldbe sampled at a time, two runs were required to sample all ofthe static pressure taps.

The static pressures measured from these taps were used to cal-culate the pressure coefficient CP at each location. The definition ofCP used for this calculation is:

CP ¼2

cM2inlet

PPinlet

� 1� �

ð2Þ

The coordinate systems is defined such the origin is at the mid-dle of the cross-section plane at the beginning of the curvature. x isthe streamwise direction, y is the cross-stream direction (i.e., nor-mal to the surface) and z is along the span. Another axis was

defined as y⁄ in order to quantify the boundary layers, where y⁄

is aligned with y, but is shifted such that it is defined to be zeroon the floor (see Fig. 5).

In addition to the static surface pressure sensors, an Instrumen-tation Can was mounted at the exit of the inlet duct to acquire thetotal pressure field at the aerodynamic interface plane (AIP). Fig. 5shows the Instrumentation Can mounted to the inlet duct witheight rakes each containing five sensors (a total of 40 sensors). Eachsensor location on the rake has a high frequency pressure trans-ducer (Kulite model XCQ-062, ±5 psid full scale and repeatabilityof ±0.5% of full scale, or ±0.025% psi) as well as a steady total pres-sure tap that can be sampled with the four Scanivalves. The num-bering scheme of the Kulite sensors at the AIP is shown in Fig. 6.The perspective of Fig. 6 is the same orientation as Fig. 5 wherethe observer is looking downstream toward the AIP.

The 40 Kulites at the AIP were utilized to measure the steadyand unsteady pressure. The steady pressure was used to plot con-tours maps of the pressure recovery, which is the total pressure atthe AIP normalized by the total pressure at the inlet. The averagepressure recovery, PRavg, and the lower half pressure recovery,PRlowerhalf, are defined respectively as

PRavg ¼ðP40

1 P0;iAIPÞ=40P0 inlet

ð3Þ

PRlowerhalf ¼ðP35

21P0;iAIPÞ=15P0 inlet

ð4Þ

Fig. 15. Streamwise pressure coefficient distribution for the three different honeycomb actuator heights at M = 0.58.

J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478 473

where the Kulites are numbered as represented in Fig. 6.Fig. 7 shows the compact inlet with the honeycomb inserted

inside it. In order to provide some proportional perspective ofthe experiment’s scaling, a quarter of US dollar is situated nextto the honeycomb mesh. An additional internal view of the tunnelis presented in Fig. 8, where the flow direction is out to the pagewith the same quarter displayed in Fig. 7.

In this work, Particle Image Velocimetry (PIV) measurementswere conducted in order to collect quantitative flow measurementof the velocity field at different streamwise and spanwise planeswithin the duct. All PIV experiments conducted in this work uti-lized a commercial LaVision System of software, and TSI hardware.The hardware included two 120 mJ Nd:YAG lasers and a single1000 � 1016 pixel resolution TSI CCD camera. Additional hardwarewas a Martin Magnum 850 fog machine utilized for seeding theflow with O(1 lm) smoke particles. An array of different opticsincluded a cylindrical lens utilized to create the laser light sheet,focal lens to focus the sheet within the measurement field, andcamera optics for focusing the camera at different focal lengthsand desired fields of interest. Also, the camera and the laser weremounted on computer-controlled traverses to provide a preciselocation of the measurement planes.

The velocity components were computed by the DaVis 7.1 soft-ware using the cross-correlation technique of pairs of successiveimages with 50% overlap between the interrogation windows.The successive images were processed using a multi-pass methodin which the initial and final passes were 32 � 32 pixels and16 � 16 pixels, respectively. Lastly, the averaged velocity field in

each plane was averaged over 500 instantaneous velocity fields.The maximum velocity (approximately 150 m/s) corresponds toan average displacement of about 4 pixels with an approximatemaximum error of ±0.1 pixel, which corresponds to an error of±2.5% of the inflow velocity.

3. Results

The results section is divided into three sections: (i) upstream ofthe curved duct, (ii) along the curved duct, and (iii) at the AIP. BothPIV and pressure measurements were conducted and are presentedin this section.

3.1. Upstream of the curved duct

Prior to the analysis of the effect of the boundary layer thicknesson the flow field downstream, we must ascertain that the incomingflow is uniform and symmetric. Thus, the pressure distributionsand the boundary layer velocity profiles were measured upstreamof the curved section, i.e. prior to the inlet plane of the duct (x/D = 0).

Two techniques were utilized to quantify the incoming flowfield. The first method utilized a total pressure Kiel probe, whichwas traversed across the height and span upstream of the inletplane at x/D = �1.56. The traversing of the probe was computercontrolled and stepped in 0.5 mm increments close to the wall,and 2.5 mm intervals towards the center of the duct. The heightand span were normalized by the inlet width (D = 114.3 mm),

Fig. 16. Pressure recovery at the AIP for the baseline and the three different honeycomb heights at M = 0.2.

474 J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478

and the pressure measured by the Kiel probe was normalized bythe total pressure in the settling chamber of the flow facility.Fig. 9 shows the total pressure at Mach 0.44, which is nearly uni-form across the height as well as the span of the duct, except closeto the duct walls, where a boundary layer exists.

The second method utilized was the PIV, which captured thevelocity flow field before the turn (x/D = �0.58) at three spanwiselocations (z/D = 0, ±0.25) and enabled the extraction of the bound-ary layer profiles. As presented in Fig. 10, the flow field upstream ofthe turn is two dimensional.

As was mentioned above, the main goal of the present work isto change the weighted contribution of the flow separation toaffect the secondary flow structures. Thus, a honeycomb wasinserted upstream of the turn at various heights (into the flow)to create a pressure drop across it which generated a velocity def-icit. As the honeycomb was inserted deeper into the flow (i.e.,increase h/d0) the velocity deficit of the boundary layer increased,which made it more susceptive to separation. Fig. 11 shows thevelocity profiles at a streamwise location of x/D = �0.58. Thevelocity was normalized by the freestream velocity, U1, and theheight was normalized by the boundary layer thickness(d0 = 0.99U1) of the baseline. The value of d0 is defined at x/D = �0.58 for M = 0.44 and is kept the same throughout the paperfor all the different Mach numbers. Similar trends were seen forM = 0.2 and 0.58 but not shown here for brevity.

The velocity profiles were also measured at a streamwise loca-tion of x/D = �0.14 and are shown in Fig. 12. The same trend seenupstream is also present at this streamwise location, except thatthe flow accelerates due to the proximity to the turn.

In order to quantify the results presented in Fig. 11 and to showthe momentum deficit, which increases due to the insertion of thehoneycomb, the momentum thickness, h, is presented in Table 1for each velocity profile.

3.2. Along the curved surface

Next, the distributions of the pressure coefficient, Cp, along thecurved surface are presented at the three Mach numbers of 0.2,0.44 and 0.58 (Figs. 13–15, respectively), at three spanwise loca-tions of z/D = 0, ±0.25 for four different honeycomb heights. Theseplots are divided into three regions:

� Region 1, where the flow accelerates around the turn and threedimensionalities develop for the baseline case. However, whenthe honeycomb is introduced into the flow it results in quasitwo-dimensional flow acceleration, where the asymmetrydecreases as h/d0 increases.� Region 2 represents the region where the baseline flow is sep-

arated as indicated by the constant pressure. The presence ofthe honeycomb affects the separation region by making itmore two-dimensional and extends it farther downstream, asis expected since the velocity deficit of the incoming boundarylayer was increased (as discussed above).� Region 3, where without the honeycomb, the flow on left side of

the duct (z/D = �0.25) reattaches first. By increasing the honey-comb height, the difference in the pressure coefficient betweenz/D = �0.25 to the other two spanwise locations (z/D = 0 and0.25) decreases. This means that the reattachment point is far-

Fig. 17. Pressure recovery at the AIP for the baseline and the three different honeycomb heights at M = 0.44.

Fig. 18. Pressure recovery at the AIP for the baseline and the three different honeycomb heights at M = 0.58.

J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478 475

Table 2Pressure recovery at the AIP.

h/d0 M = 0.2 M = 0.4 M = 0.58

PRave PRlower half PRave PRlower half PRave PRlower half

0 0.98904 0.98185 0.94934 0.91631 0.92183 0.875020.74 0.98916 0.98217 0.94905 0.91563 0.92115 0.873261.47 0.98879 0.98126 0.94690 0.91068 0.91731 0.865052.2 0.98857 0.98070 0.94517 0.90656 0.91560 0.86134

476 J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478

ther downstream with the honeycomb than without it. There-fore, elevating the honeycomb increases the extent of the sepa-ration region.

3.3. At the AIP

Next, the pressure recovery at the AIP was measured and isdiscussed in this section. Figs. 16–18 present the time-averagedpressure recovery at three Mach numbers and four honeycombheights. The red region represents a pressure recovery of 1,meaning that there are no pressure losses, and the blue regionrepresents pressure losses. As can be seen, without the honey-comb, the asymmetry, which is a result of the superposition ofthe separation and the secondary flow structures, is clearly pres-ent for all three Mach numbers. Here, the lowest pressure recov-ery is on the bottom right side. However, increasing thehoneycomb height decreases the asymmetry, with minimal effectof the overall pressure recovery.

Table 2 presents the pressure recovery values averaged over theentire AIP plane, PRave, and over only the lower half of the AIPplane, PRlowerhalf, for the different honeycomb heights. An importantresult from this study is the low impact of the honeycomb mesh onthe pressure recovery while improving the symmetry of the pres-sure distribution. By looking at the values of Table 2, the largestimpact on the pressure recovery is at M = 0.58 between the base-

Fig. 19. Standard deviation of the pressure recovery at the AIP at M = 0.2, 0.44 andconfiguration achieved in the current experiments (h/d0 = 2.2).

line case (without honeycomb) and the case with the honeycombinserted at h/d0 = 2.2, where the difference is 0.68%.

In addition to the time-averaged pressure recovery, the effect ofthe honeycomb on the fluctuating total pressure was also explored.Fig. 19 presents the standard deviation of the pressure recovery forthe same cases showed in Fig. 18. For the baseline cases (i.e., with-out the honeycomb present) at all three Mach numbers, there is alarge concentration of pressure fluctuations in the lower portion ofthe AIP, which is the region where the PR is the lowest. When thehoneycomb is inserted into the duct (where h/d0 = 2.2), the fluctu-ating pressure fields show symmetric distribution of the pressurefluctuations field.

In addition to a global evaluation of the fluctuating pressurefield, more detailed examination was conducted where the powerspectra at three spanwise locations (sensors 23, 28, 33) at the AIPwere calculated and are presented in Fig. 20. For all three Machnumbers, without the honeycomb in the flow, the power spectraat the three spanwise locations are different, whereas the spectralcontent on the left side and the middle of the AIP (sensors 23 and28, the red and green lines, respectively) contains higher energythan that on the right side (sensor 33, blue line). When the hon-eycomb is inserted into the flow there is a negligible change onthe energy content in the left side of the AIP, meanwhile theenergy on the right side increases considerably. This modificationin the amplitude of the power spectra further suggests that theintroduction of the honeycomb induce a more symmetrical flowfield (both the time average and the fluctuating). Also, the mag-nitude of the peak, which represents the dominant shedding fre-quency of the separated flow, increases with the honeycombinserted into the flow, corroborating the results shown previ-ously. The increase in amplitude, especially at the sheddingenergy, is another indication that the separation becomes moredominant and by doing so helps push the saddle–saddle pointfarther downstream in the duct. This saddle–saddle point, asmentioned previously, causes the onset of the instability respon-sible for the asymmetry encountered in the flow without thehoneycomb actuator.

0.58. Upper row: baseline flow (h/d0 = 0), and lower row: the most symmetrical

Fig. 20. Power spectra on the bottom part of the AIP at three spanwise locations and for the three Mach numbers. Upper row: h/d0 = 2.2, and lower row: h/d0 = 2.2.

J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478 477

4. Conclusion

The present work showed that, under some conditions, anasymmetric flow field could develop inside a short inlet duct. Inorder to correct this asymmetry, a two-dimensional honeycombmesh was inserted upstream of the curved duct to increase thevelocity deficit of the incoming boundary layer. It was shownexperimentally, using steady and unsteady pressure measure-ments and PIV, that the manipulations of the velocity deficit ledto a more promptly flow separation, which, as a result, modifiedthe secondary flow structures and corrected the flow asymmetryat the AIP. The experiments were conducted at Mach numbers of0.2, 0.44 and 0.58, and the effect of inserting a honeycomb wastested by increasing its penetration into the flow from 0 to 2.2times the boundary layer thickness of the baseline flow upstreamof the curve. Using the honeycomb, flow symmetry was achievedfor the specific geometrical configuration tested with a negligibledecrease of the pressure recovery at the AIP. Moreover, the span-wise uniformity of the flow increased as the height of the honey-comb was increased.

Acknowledgements

This work was funded by the Northrop Grumman Corporation(monitored by Ms. Florine Cannelle). The authors would also liketo thank the help of Dr. Israel Salvador and Mr. Brian Debronsky.

References

Amitay, M., Pitt, D., Glezer, A., 2002. Separation control in duct flows. J. Aircraft 39(4), 616–620.

Bansod, P., Bradshaw, P., 1972. The flow in S-shaped ducts. Aeronautical Quarterly23, 131–140.

Bradshaw, P., 1987. Turbulent secondary flows. Annu. Rev. Fluid Mech. 19, 53–74.

Chen, Y., 2012. Subsonic Flows through S-Ducts with Flow Control. Ph.D. Thesis.Department of Mechanical, Aerospace & Nuclear Engineering, RensselaerPolytechnic Institute, Troy, NY.

Dagget, D.L., Kawai, R., Friedman, D., 2003. Blended Wing Body Systems Studies:Boundary Layer Ingestion Inlets with Active Flow Control. Technical ReportNASA/CR-2003-212670, Hampton, VA.

Debronsky, B., 2012. Experimental Investigation of Flow Control Techniques in aCompact Inlet Duct. M.S. Dissertation. Department of Mechanical, Aerospace &Nuclear Engineering, Rensselaer Polytechnic Institute, Troy, NY.

Enayet, M.M., Gibson, M.M., Taylor, A.P., Yianneskis, M., 1982. Laser-Dopplermeasurements of laminar and turbulent flow in a pipe bend. Int. J. Heat FluidFlow 3 (4), 213–219.

Gissen, A.N., Vukasinovic, B., McMillan, M.L., Glezer, A., 2011. Dynamics of HybridFlow Control in a Boundary-Layer-Ingesting Offset Diffuser. AIAA Paper 2011-3096, 41st AIAA Fluid Dynamics Conference and Exhibit, Honolulu, HI.

Jirasek, A., 2006. Design of vortex generator flow control in inlets. J. Aircraft 43 (6),1886–1892.

Launder, B.E., Ying, W.M., 1972. Secondary flows in ducts of square cross-section. J.Fluid Mech. 54 (Part 2), 289–295.

Mattingly, J.D., Heiser, W.H., Preatt, D.T., 2002. Aircraft Engine Design. AIAAEducational Series.

Miller, D.S., 1991. Internal Flow Systems, 1990. BHRA, Cranfied, Bedford, UK.Ng, Y.T., Luo, S.C., Lim, T.T., Ho, Q.W., 2006. On swirl development in a square cross-

sectioned, S-shaped duct. Exp. Fluids 41 (6), 975–989.Ng, Y.T., Luo, S.C., Lim, T.T., Ho, Q.W., 2008. On the relation between centrifugal force

and radial pressure gradient in flow inside curved and S-shaped ducts. Phys.Fluids 20 (5), 055109–055109-15.

Perkins, H.J., 1970. The formation of streamwise vorticity in turbulent flow. J. FluidMech. 44 (4), 721–740.

Perry, A.E., Hornung, H., 1984. Some aspects of three-dimensional separation. II-Vortex skeletons. Zeitshrift fuer Flugwissenschaften und Weltraumforschung 8,155–160.

Reichert, B.A., Wendt, B.J., 1994. Improving Diffusing S-Duct Performance bySecondary Flow Control. NASA Technical Memorandum 106492, also AIAA 32ndAerospace Sciences Meeting and Exhibit, Reno, NV.

Scribben, A.C., Ng, W., Burdisso, R., 2006. Effectiveness of a serpentine inlet ductflow control technique at design and off-design simulated flight conditions. J.Turbomach. 128, 332–339.

Tobak, M., Peak, D.J., 1982. Topology of three-dimensional separated flows. Annu.Rev. Fluid Mech. 14, 61–85.

Vaccaro, J.C., 2011. Flow Control in a Compact Inlet. Ph.D. Thesis. Department ofMechanical, Aerospace & Nuclear Engineering, Rensselaer Polytechnic Institute,Troy, NY.

Wellborn, S., Reichert, B., Okiishi, T., 1992. An experimental investigation of the flowin a diffusing S-duct. NASA Technical Memorandum 105809.

478 J. Gartner, M. Amitay / International Journal of Heat and Fluid Flow 50 (2014) 467–478

Wellborn, S., Okiishi, T., Reichert, B., 1993. A study of the compressible flow througha diffusing S-duct. NASA Technical Memorandum 106411.

Whitelaw, J.H., Yu, S.C.M., 1993a. Velocity measurements in an S-shaped diffusingduct. Exp. Fluids 15 (4–5), 364–367.

Whitelaw, J.H., Yu, S.C.M., 1993b. Turbulent flow characteristics in an S-shapeddiffusing duct. Flow Meas. Instrum. 4 (3), 171–179.