experimental and numerical investigation of secondary flow...

Experimental and Numerical Investigation ofSecondary Flow Structures in an Annular LPT Cascadeunder Periodical Wake Impact – Part 1: ExperimentalResultsMartin Sinkwitz1*, Benjamin Winhart1, David Engelmann1, Francesca di Mare1, Ronald Mailach2

SYM

POSI

A

ON ROTATING MACHIN

ERY

ISROMAC 2017

InternationalSymposium on

Transport Phenomenaand

Dynamics of RotatingMachinery

Maui, Hawaii

December 16-21, 2017

AbstractExperimental studies have been conducted on a modi�ed T106 low pressure turbine (LPT) pro�lein a 1.5 stage annular axial turbine rig at the Chair of �ermal Turbomachines, Ruhr-UniversitatBochum. �e rig setup allows the highly resolved measurement of unsteady wake-stator �owinteraction in both space and time. Incoming wakes are generated by a variable-speed driven rotorequipped with cylindrical bars.

In the present paper an experimental approach to the investigation of unsteady phenomena isproposed. Time-averaged as well as instantaneous measurement data from 2D �ow �eld traversesat the stator exit are provided for the analysis of the periodically unsteady vortex formation,displacement and suppression. Additional time-accurate blade pressure data is used to study therelationship between the detected secondary �ow downstream of the stator row and the immediatewake impact on the blade pro�le.KeywordsAxial Turbomachinery — Secondary Flow — Wakes — Unsteady Flow Measurements

1Chair of Thermal Turbomachines, Ruhr-Universitat Bochum, Bochum, Germany2Chair of Turbomachinery and Flight Propulsion, Technische Universitat Dresden, Dresden, Germany*Corresponding author: [email protected]

INTRODUCTION

Blade count reduction in the low pressure part of the aero en-gine can provide a substantial contribution to the reductionof the overall weight of an engine. However, this must becompensated through the design of high-li�, heavy-loadedblades which in turn are exposed to increased pressure gra-dients, a�ecting in particular blade pro�le and end wall �owwhich is already prone to separation. Intensi�ed boundarylayer separation at end walls and on the pro�le suction sideresults in pronounced secondary �ow systems. �ese in turninduce high loss, a loss of li� and main �ow perturbation[1, 2]. Di�erent secondary �ow models have been proposed,extended and modi�ed, resulting in a basic, broadly accepted,secondary �ow model representation (e.g. [2]). �is conceptmainly consists of passage vortices (PV), horse shoe vortices(HSV), tip leakage vortex, corner vortices (CV) and a vortexstreet that is shed from the blade trailing edge (TEWV). Afew of the models incorporate yet another vortex structureoriginating from the separated suction side boundary layer,labeled as the concentrated shed vortex (CSV) [3], resultingfrom end wall boundary layer �uid (PV) impinging on theblade suction side and leading to a separation of the suctionside boundary layer itself close to the end wall. Analysisand potential reduction of these vortical structures are com-plicated by the multistage environment in turbomachines,

where rotor stator interaction a�ects the transport and shed-ding of blade wakes and vortices, resulting in an extremelynon-uniform, distorted and time-dependent �ow �eld. �us,it is di�cult to separate the distinct interacting �ow struc-tures from each other and to determine their degree of in�u-ence on aerodynamic loss generation mechanisms. Numer-ous tests have been conducted under simpli�ed conditions inlinear cascades. To include the in�uence of periodically un-steady rotor blade wakes in linear setups, linear wake gener-ators have been employed, using cylindrical bars to simulatewakes [3, 4, 5, 6, 7, 8, 9, 10]. An important common �ndingof these investigations is that unsteady wake in�ow can sta-bilize boundary layer �ow. �us, periodic rea�achment ofa boundary layer, which would separate under unperturbedin�ow conditions, is promoted [8, 9, 11]. However, the inves-tigation of secondary �ow within linear cascades neglectsseveral essential e�ects present in real turbomachinery �ow.

In this study an experimental setup for the time-resolvedanalysis of wake-stator �ow interaction, using an annulargeometry instead of a linear cascade, is applied. �is way,the in�uence of curvilinear end walls, non-uniform, radiallyincreasing pitch and radial �ow migration is incorporated.Incoming wakes are generated by a variable-speed drivenrotor disk equipped with cylindrical bars. Special emphasisis put on wake-induced, time-dependent weakening and dis-

Experimental Investigation of Secondary Flow in an Annular LPT Cascade under Periodical Wake Impact — 2/9

Figure 1. Experimental Test Facility. Sectional view (a), 3D illustration (b).

placement of particular vortices of the secondary �ow system.Furthermore, wake-boundary layer �ow interaction and �owseparation along the blade suction surface is studied. �edistinction between time-averaged and time-resolved anal-ysis of secondary �ow phenomena and the importance oftime-resolved measurement is emphasized and the temporalevolution of characteristic �ow �eld quantities is utilized toanalyze the periodic impact on the vortices. By relating mea-surement data of two-dimensional �ow �eld traverses andtime-accurate blade pro�le pressure signals to each other, adeeper understanding of the complex interaction between barwake, boundary layer and secondary �ow can be obtained.

1. EXPERIMENTAL TEST FACILITYFor this study an existing large scale axial �ow turbine testrig installed at the Chair of �ermal Turbomachines of Ruhr-Universitat Bochum was retro��ed to allow highly resolvedmeasurements of unsteadywake-stator �ow interaction. Pivot-mounted casing elements with multiple probe accesses havebeen constructed which are placed between the IGV and thestator row as well as downstream of the stator. Stepwise ro-tation of these elements and radial translation of the probesenable the automated recording of two-dimensional �ow tra-verses. �e test facility is operated continuously in an opencircuit, with ambient air. Flow is induced by a 150 kW vari-able speed engine coupled to a radial blower with a m = 13kg/s mass �ow capacity. �e blower is placed downstream ofthe test section, so that the rig is operated in suction mode.�e large dimensions of the �ow channel allow detailed �owmeasurements with negligible blockage and perturbation byinstalled probes.

�e a�-loaded blade pro�le under investigation, labeledas T106RUB, is an in-house modi�cation of the well-knownT106 LPT blade. It was developed for matching the char-acteristics of the T106 pro�le at the rig’s low Mach num-ber �ow. �e development and underlying principles of thetransformation procedure are described in [12]. �e pro�le ischaracterized by a cylindrical geometry with a chord lengthof C = 0.1 m and an aspect ratio of H/C = 1.7. �is aspect

ratio ensures that near end wall vortices do not merge andas a consequence quasi two-dimensional �ow conditions atmid-span. �e stator row consists of 60 blades with a staggerangle of λT106 = 30.7° and a pitch-to-chord ratio at mid-spanof g/C = 0.78.

An inlet guide vane (IGV) row ensures proper in�ow an-gles to the stator. It is made up of 60 vanes (NACA 8408pro�le) and was developed for correct �ow turning whilstleaving the T106RUB in�ow as far as possible una�ected bywakes and secondary �ow structures. �e choice of 60 NACA8408 pro�les is the result of an elaborate numerical study,comparing 42 di�erent pro�le geometries and several bladecounts. �e IGV is placed 2.61 C upstream of the bar planeutilizing the maximal possible distance the axial test rig di-mensions allowed.

To simulate blade wakes, the rotor disk was equippedwith radially stacked steel bars (bar diameter DB = 0.002 m,bar length LB = 0.168 m, bar pitch gB = 0.039/0.078/0.117 m).�us, it is possible to isolate both velocity defect and turbu-lence increase of rotor blade wakes without the additionalsecondary �ow system, associated with �ow turning in ablade passage. Wakes are generated in a plane parallel to thestator leading edges (LE), located at an axial distance of 0.33C upstream of the LE, which represents a typical axial gapwidth in a LPT. �e investigations were carried out for threedi�erent values of bar pitch. �e same stator and IGV bladecounts (60) as well as the number of wake generator bars(40/60/120) allow time-e�cient numerical studies.

A 15 kW AC engine controlled by a frequency converterdrives the rotor sha�. �e setup of IGV, wake generator andT106RUB stator is shown in Figure 1.

2. MEASUREMENT TECHNIQUESAdjustment and monitoring of the operating point was real-ized with help of multiple Prandtl-probes (pt, ps and hence c)at di�erent characteristic planes, a stationary �ve hole probe(5HP) (pt, ps, c, �ow angles α and γ) in the stator exit and acombined temperature and relative humidity sensor at thetest rig inlet. �e ambient pressure (reference pressure, abso-


lute) was measured by a Rosemount 3051 absolute pressuretransducer (accuracy±0.15 % of calibrated transducermeasur-ing range). All remaining pressures (relative to ambient) wererecorded by a 64-channel Scanivalve ZOC33/64Px miniaturepressure scanner module (accuracy±0.10 % of transducer full-scale range) controlled by a Scanivalve ERAD4000 unit. �e64 individual high-sensitivity low-range (70 mbar) pressuresensors enable simultaneous data acquisition of all incorpo-rated pneumatic pressure probes. By recording pressurespractically without temporal o�set between one and another,the systematic measurement uncertainty – e.g. of 5HP, wallor pro�le pressure measurements - is reduced considerably.

Table 1. Main geometric test rig properties and turbinestage parameters

Test rig

Outer diameter (Casing) DC 1.660 mInner diameter (Hub) DH 1.320 mMid-span diameter Dm 1.490 m

Turbine stage

Blade height IGV, T106RUB H 0.170 mChord length IGV CIGV 0.137 mStagger angle IGV λIGV -25.5°Chord length T106RUB CT106=C 0.100 mStagger angle T106RUB λT106 30.7°Blade count IGV, T106RUB nIGV,T106 60Design �ow angles, mid-span:IGV inlet α0 90.0°IGV outlet = T106RUB inlet α1=α2 52.3°T106RUB outlet α3 153.2°Axial distances:IGV TE - bar plane x/CT106 2.61bar plane - T106RUB LE x/CT106 0.33Bar diameter DB 0.002 mBar length LB 0.168 mBar pitch at mid-span gB 0.039 m,

0.078 m,0.117 m

Operating point, Design point

Mass �ow m 12.8 kg/sReynolds number (exit, th.) Reexit, th 200,000Mach number (exit, th.) Maexit, th 0.091Strouhal number range Sr 0.45–3.15Flow coe�cient range φ 0.81–2.84

For this study, two-dimensional (2D) (in radial and cir-cumferential direction) �ow �eld traverses have been carriedout downstream of stator TE (Figure 1, plane 3.1) to quantifythe bar wake impact on the downstream secondary �ow sys-tem. For time-averaged traverses, miniature 5HP (head diam-eter of 2 mm, bore diameter of 0.3 mm) were employed. Hotwire anemometry measurements (CTA mode) provided time-

resolved traverse data. For acquisition of the time-resolved3D �ow vector these were conducted with two di�erent typesof Split Fiber Probes (SFP), fabricated by Dantec Dynamics.SFP probes of type 55R56 were used for measuring axial andradial velocity components (for computing �ow angle δ) andprobes of type 55R57 for measuring axial and circumferentialcomponents (for computing �ow angle α). Hot wire signalshave been recorded at a rate of 50 kHz (125 times higher thanthe maximum bar passing frequency) for at least 50 samples,triggered by a one/revolution signal from an inductive en-coder on the rotor sha�. �e minimum of 50 samples wasde�ned as the best trade-o� between time consumption (of2D traverses) and validity of the phase-averaged results. Forthe discussed investigations, where unsteady �uctuations inthe wake �ow exceed the angular range of ±45° and whererecording rates of 50 kHz are su�cient, SFP have shownsuperior performance over conventional X-array probes. As3D SFP are not available and conventional triple-sensor wire-probes are too large in diameter for the test rig probe accesses,acquisition of all three velocity components requires the com-bined use of the two di�erent types of SFP. Measurementdata of both probes were correlated and post-processed toobtain 3D �ow vectors. For SFP data acquisition a StreamLine90N10 CTA anemometer by Dantec Dynamics (incorporatingthree 90C10 CTA modules) in combination with a NationalInstruments NI 9215 A/D converter was employed. Highspatial traverse resolution was achieved by 2D measurementgrids of up to 46 radial and 49 circumferential positions. �ecircumferential positions were evenly distributed over twostator pitches, whereas the radial increments were re�nedclose to the end walls. In the case of time-consuming SFP-measurements the spatial resolution was decreased slightly.At the start of every measurement, probes were aligned tothe main �ow direction at mid-span. 5HP measurementswere carried out with fully calibrated probes in a non-nullingmode.

Figure 2. T106RUB blade, instrumented with Kulite LQ-125sensors for time-resolved pro�le pressure measurements.Pressure side view (le�) and arrangement of sensors (right).

For additional data acquisition within the T106RUB statorpassage, blade pro�le pressures were measured on variousT106RUB stator blades. 84 pressure taps, evenly distributedto 3 positions of constant span (near hub at R/H = 10 %, at


mid-span at R/H = 50 %, near casing at R/H = 90 %) enabledtime-averaged, pneumatic pressure measurement. Further-more, time-resolved, piezo-electric pressure measurementswere conducted at mid-span, using 10 �ush-mounted Kulitetype LQ-125 sensors (see Figure 2). An IMC CRONOS�exsystem was incorporated for recording Kulite signals with anacquisition rate of 50 kHz. At every measurement position200 samples were collected, also triggered once per revo-lution. A phase-locked ensemble averaging technique wasapplied to the recorded time-resolved data samples to obtainreliable quantities from SFP and Kulite sensor measurements.�is facilitates the separation of periodically unsteady sig-nals originating from wake impact from stochastic turbulent�uctuations without losing the time-dependent character. Allshown time-resolved results have been averaged with thismethod.

3. EXPERIMENTAL RESULTS AND DISCUS-SIONFor all investigations discussed in this paper, the theoreti-cal exit Reynolds number was kept constant at Reexit, th =200,000 (based on T106RUB chord length C, with the �uid’sdynamic viscosity µ, the theoretical exit velocity cexit, th andthe corresponding density ρexit, th):

Reexit, th =cexit, th · ρexit, th · C

µ(1)

�is is a typical value for LPT operation at high alti-tude where suction side boundary layer separation can occur.�e theoretical exit state (using upstream total pressure anddownstream static pressure) is a common way to account foraltered losses when bar velocity or bar count are modi�edwhile maintaining comparability. As the main dimension-less quantities for characterizing unsteady �ow disturbances,the Strouhal number Sr and �ow coe�cient φ have beenidenti�ed [6, 12]:

Sr = f BP ·Ccax=

cBgB·

Ccax

(2)

φ =caxcB

(3)

To study the e�ect of di�erent periodically unsteady in-�ow conditions (disturbance by incoming wakes), these havebeen varied in wide ranges including values for typical LPToperation. As for all discussed measurements the axial veloc-ity cax is approximately held constant, bar wake frequencyf BP and thus Sr and φ were modi�ed by altering bar speedcB and/or bar pitch gB. �e results presented in this articleare based on �ndings from previous work by the authors[12] where detailed analysis of 2D �ow �eld traverses wereconducted for di�erent combinations of Sr and φ. By in-cluding time-resolved blade pressure data for the presentwork, a causality relationship between the blade passage anddownstream �ow can be established.

3.1 Stator exit flow2D �ow �eld traverses have been conducted in multipleplanes downstream of the T106RUB stator row. Measurementdata from these planes permit an evaluation of the secondary�ow system emerging upstreamwithin the stator passage. Bycomparing �ow �elds of di�erent stator exit �ow planes, theprocesses based on the mixing of stator wakes and vorticeswith the free passage �ow can be described. By consideringdi�erent operating points - de�ned by combinations of Srand φ - and comparing them with the unperturbed case, thee�ect of bar wakes and the temporal evolution of secondary�ow structures can be derived (for detailed underlying anal-ysis see [12]).

3.1.1 Time-averaged measurement dataFigure 3 shows time-averaged (obtained with 5HP) distribu-tions of axial vorticity (AVO) in plane 3.1 (Figure 1a), 0.15 Cdownstream of stator TE in the range of two stator pitches.

Figure 3. Time-averaged results 0.15 C downstream of TE.AVO distribution for steady undisturbed case (le�) and for acase with signi�cant wake disturbed in�ow (Sr = 1.55, φ =0.83) (right).

�e AVO can be derived from those components of thevelocity vector, tangential to the axial direction (cy and cz)and the geometry of the measurement grid:

AVO =∂cz∂y−∂cy∂z

(4)

�e �ow �eld is given for the undisturbed (le�) and aperturbed (right) case, whereas the direction of view is up-stream. Relevant vortices of the stator secondary �ow system,the estimated stator wake position as well as the positionof suction and pressure side of the stator wake have beenlabeled to aid interpretation. �e particular vortices havebeen identi�ed by their direction of rotation and their loca-tion in the vicinity of characteristic �ow regions, such as thesuction side or the end walls. �us, the passage vortex (PV)pair could be detected near the end walls on the suction sideof the stator wake. �e PV pair is a result of low momentumend wall boundary layer �uid, following the cross passage


pressure gradient from the pressure side to the suction side ofthe neighboring blade. Impinging at the pro�le suction side,this cross�ow is diverted across the blade towards mid-span,consequently shaping the PV. Furthermore, near the pro�leLE the horse shoe vortex (HSV) system is formed, when theend wall boundary layer from upstream hits the LE and splitsdue to the local adverse pressure gradient. �e suction sidelegs (HSV-SL) propagate close to the suction side as theyare pushed against the blade by the cross passage pressuregradient. As the suction side legs are concentrated within thecorner between end wall and blade suction side, they do notcontain much material, stay small and potentially merge withthe respective nearby corner vortex (CV). �e pressure sidelegs (HSV-PL), instead, are diverted from the blade pressureside to the suction side of the adjacent blade, also followingthe cross passage pressure gradient. For the steady, unper-turbed case the HSV-PL can clearly be identi�ed among thePV and the respective end wall. Despite the same directionof rotation, these have not merged entirely with the PV, asreported for other setups e.g. in [2, 10]. Instead they push thePV towards mid-span. In close proximity, a spacious systemcan be detected, rotating in opposite direction. �is systemis supposed to contain the concentrated shed vortex (CSV),the described combination of CV and HSV-SL, as well as thetrailing edge wake vortex street (TEWV). �e origin of theCSV is a direct consequence of the low momentum PV �owimpinging at the blade suction side. It induces separationof the suction side boundary layer, which then rolls up intoa counter rotating vortex next to the PV [12]. At TE theadverse directions of radial blade �ow migration on pressureand suction side form a shear layer (TEWV) that unites withthe CSV.

In the right part of Figure 3, with Sr = 1.55, φ = 0.83 a casewith signi�cant periodical in�ow perturbation is displayed.In direct comparison with the unperturbed case an appar-ent impact on several vortical structures can be observed.Near the casing both a decrease in size and a displacementof the PV + HSV-PL system towards the end wall can beseen. Furthermore, a signi�cant weakening of the CSV +TEWV takes place. �is can be justi�ed by the impact ofthe bar wakes, which interact with boundary layer �ow onthe blade and on end walls. �e wakes, acting as negativejets, promote boundary layer transition and leave the wall�ow in a temporarily unperturbed condition a�er the wakehas passed [11, 12, 13]. In this state the boundary layer isnot as prone to separation and de�ection as in the undis-turbed case. Intermi�ently, cross passage transport of endwall �ow is reduced and less lowmomentum �uid is providedto the PV, reducing its dimensions. Further on, separation ofboundary layer �ow at the pro�le suction side is periodicallyreduced, as well. As a consequence, particularly the CSV isweakened, as it results from boundary layer material bothfrom the end wall and from the pro�le suction side. Nearthe hub, impact on the secondary �ow system seems slightlydi�erent. �e PV appears even larger, but also more diluted.A weakening of the CSV+TEWV system is observable, but

not as pronounced as near the casing. �us inspection ofthe time-averaged exit �ow �eld already indicates the spatialimbalance (hub/casing) of bar wake impact in an annularcascade contrary to the uniform �ow in linear cascades.

3.1.2 Time-resolved measurement dataFor amore detailed insight into the intermi�ent �ow phenom-ena downstream of the T106RUB stator passage, time-accuratedata were recorded with SFP. �e probes were traversed sim-ilarly to 5HP measurements in the same plane 0.15 C down-stream of stator TE. �ree observation stations have beenselected at various spanwise positions (R/H = 20 %, 50 %, 85%) and the corresponding �ow �eld temporal evolution overthree bar passing periods of c and AVO are represented inFigure 4. �e stations capture the secondary �ow systemsnear the hub and the shroud and the �ow at mid-span. �eoperating point (Sr = 1.55, φ = 0.83) matches the one chosenin Figure 3 (right). �us, Figure 3 (right) represents the time-averaged picture of the time-resolved phenomena depictedin Figure 4.

Figure 4. Time-resolved results 0.15 C downstream of TE.Time-space evolution of ∆c (top) and AVO (bo�om) forthree slices at constant span, Sr = 1.55, φ = 0.83.

In the upper part of Figure 4 the temporal evolution ofvelocity c is given, in the lower the same can be seen for theAVO. For the velocity evolution, the respective local valuesat undisturbed condition have been subtracted from the time-resolved, phase-averaged velocity values:

∆c (R, θ, t) = c (R, θ, t) − c (R, θ) steady (5)

For the AVO, absolute values are shown. Approximatelocations of the stator wakes have been marked with do�ed


vertical lines, also suction side (SS) and pressure side (PS)�ow. �e velocity �eld is perturbed periodically with thebar (wake) passing frequency (tBP = 1/ f BP correlates to theduration of one bar passing). However, the situation that canbe observed downstream of the stator is not just a periodicalperturbation by the passing bar wake, but a superposition ofseveral e�ects, which can be traced back to wake kinematicsand the displacement of vortices. While propagating throughthe stator passage, the incoming bar wakes are cut, stretched,bowed, undergo mixing and interact with the passage �owas well as with the near wall �ow at the blade suction andpressure side and at the end walls. �e bar wakes, acting aslocal regions of low pressure, momentum and velocity can bedescribed as negative jets [11, 14], pointing in the oppositedirection than the main �ow. When impinging on the pro�lepressure side and passing over it, the bar wake induces twocounter rotating, vortical-like structures (radial rotationaldirection) (compare e.g. [13] or [15]). �is results in a �owdeceleration in front of the wake and a �ow accelerationimmediately a�er the wake has passed. On the suction sidethe impact on �ow is reversed. Here, the wake pushes aregion of accelerated squeezed passage �ow in front of it andis followed by a low-speed region. �is interaction betweenbar wake and pro�le �ow will be discussed in more detail insection 3.2 and in [15]. �e described local �ow accelerationand deceleration within the stator passage �nally result infour characteristic, periodically passing parcels of alteredvelocity that can still be observed in the stator exit �ow �eld.For the �ow at mid-span (R/H = 50 %) these four parcels closeto the bar wake have been labelled with A-D in Figure 4 (topcenter). �e distorted bar wakes have been highlighted, aswell. Immediately before the wake arrives, decelerated vol-ume A and accelerated volume C are present. When the barwake has passed the observed plane, parcels B and D follow.Based on their origin, these four volumes are located near theblade suction and pressure side respectively and thus closeto the stator pro�le wakes. Consequently, especially close tothe pro�le wake a pulsating �ow �eld is induced, whereasthe free passage �ow is li�le disturbed. Downstream of thepassage, the accelerated parcels B and C from previouslyseparated, adjacent passages merge. Interestingly, there isnearly no change at all regarding the AVO (bo�om center),remaining fairly constant over time, indicating that no addi-tional vortices with axial sense of rotation are generated. �iscorresponds to the observation that the vortical structures,induced by the negative jet like propagation of the bar wake,generate vortices with primary radial rotational direction.

Near the end walls, the situation is far more complex thanat mid-span. Beyond the pulsating bar wake impact, the �ow�eld here is a�ected by the stator secondary �ow structures.Both at R/H = 20 % and R/H = 85 % the velocity distribu-tions undergo a similar periodical perturbation as alreadydescribed for the �ow at mid-span. But although the charac-teristic parcels can be found again, an apparent di�erence isnoticed regarding structures C and D, as they are broken upand displaced from the suction side into the passage. �is can

be traced back to the periodical interaction between bar wakeand the nearby secondary �ow system. It has to be noted thatthe altered velocity is a result both of the passing bar wakevelocity �eld and of displaced vortical structures carryinglow velocity �uid. By including the evolution of the AVO intothe analysis, this interaction can be observed more clearly. Inthe range of -0.5° ≤ θ ≤ 0.5° a periodical weakening of the PV(within the regarded plane) becomes visible. �is weaken-ing can be linked to the approaching bar wake carrying theaccelerated �ow regime C. When the wake has passed, thedecelerated structure D is split, as is wraps around the PV.At the same time, the nearby CSV (θ ≈ 1.5°) is periodicallyweakened as well. As the comparison with adjacent span-wise slices (below and above, not shown here) has revealed,additionally to the intermi�ent weakening and strengthen-ing of PV and CSV, they are also displaced radially. Bothweakening as well as radial displacement out of the observedplane reduce the local magnitude of the AVO. As the CSV isfar more displaced (in radial direction) than the PV, its coreis intermi�ently located within the observed plane, above orbelow. �is wavelike in-and-out of plane movement mustnot be misinterpreted as a pa�ern occurring at a doubledfrequency. �e in�uence on the secondary �ow structures inturn can be traced back to a combination of a direct and anindirect bar wake impact. When the bar wake passes, its localvelocity �eld (parcels A-D) has a direct impact on the exit�ow �eld including the secondary �ow system. Additionally,the vortex development has already been a�ected upstreamby the bar wake passing through the stator passage, but backin time. As only the temporal o�set consequences of thise�ect can be observed in the exit �ow, the authors term thisthe indirect impact. Based on the interaction between the barwake (periodical altered velocity �eld and injection of mo-mentum and turbulence) and the �ow through the passagenear the end walls, the cross passage transport is decreasedand vortex formation and propagation are seriously in�u-enced. �e authors assume, that initially the propagation ofthe HSV-PL is massively perturbed by the passing wakes. Itsimpingement position at the adjacent blade suction side isshi�ed downstream, reducing impact on suction side �owand thus on PV and CSV. Under unperturbed in�ow bothPV and CSV are pushed towards mid-span by the intenseHSV-PL. Under wake impact, they are periodically shi�edback towards the end wall. Due to slower propagation of thevortical secondary �ow structures than the main �ow, thee�ects of passing bar wake (direct) and a�ected secondary�ow structures (indirect) arrive at di�erent instants of timein the observed plane in the stator exit �ow. Furthermore, thedi�erent propagation velocities at mid-span, hub and casingalso induce a temporal shi� between the impact of the barwakes at di�erent spanwise positions.

3.2 Profile pressure distributionEvaluation of the stator exit �ow �eld allows an analysisof bar wake-secondary �ow interaction. To examine theimmediate impact of periodically unsteady bar wakes onstator blade pro�le �ow, experimental data from within the


passage is necessary. �us, multiple positions on variousT106RUB stator blades were equipped with pressure taps. At�rst, time-averaged, pneumatic pressure readings will bediscussed, followed by time-accurate measurement data frompiezoelectric sensors, embedded in the blade surfaces.

3.2.1 Time-averaged measurement data

Figure 5. Time-averaged cp distributions at T106RUBmid-span for three di�erent operating points.

Figure 5 shows time-averaged, non-dimensionalized staticblade pressure distributions (cp) at mid-span (R/H = 50 %) fordi�erent operating points, whereas cp was de�ned as follows:

cp =p (x) − p1pt,1 − p1

(6)

For reasons of clarity, error bars have been omi�ed, seepart 2 [15] for those values including error bars. �e undis-turbed steady situation is given by the black curve, the greencurve (Sr = 0.45, φ = 2.84) shows a lightly disturbed case,whereas the red curve (Sr = 1.55, φ = 0.83) represents the al-ready introduced case with heavy disturbance. �e disturbedoperating points exhibit constant bar pitch (gB = 0.078 m) butdi�ering bar speed and thus wake frequency. At �rst sightthe distributions are very similar and do not feature any strik-ing di�erences. On closer examination, variations con�nedto the regions close to LE (details A and B for 0 ≤ x/Cax ≤ 0.2)and close to TE (detail C for 0.8 ≤ x/Cax ≤ 1)) become visible.Due to the wake generator rotation closely upstream of LE,for the perturbed operating points the T106RUB stator pro�lesexperience in�ow with considerable periodical incidence. Asshown in prior work by the authors [12] this results in max-imal deviation angles from design in�ow of ∆α = -9°. �enegative incidence in�ow shi�s the stagnation point at LEtowards the suction side. �is decreases the velocity on thesuction side (detail B) and increases it on the pressure side(detail A) close to LE. Although this is a time-accurate e�ect,its time-averaged impact can be observed in Figure 5. NearTE only suction side �ow is in�uenced by the unsteady in�ow(detail C). �is di�erent behavior of suction side �ow nearTE can possibly be traced back to the interaction betweenthe passing bar wake and a separation bubble in the rear partof the suction side.

From comparison of these three operating points it be-comes obvious that the impact of unsteady in�ow on thetime-averaged blade pressure distribution is very small. As aconsequence, the use of pneumatic pressure taps alone givingtime-averaged blade pressure distributions is not suitable fordetailed analysis of the discussed unsteady phenomena.

3.2.2 Time-resolved measurement dataFor deeper insight into the unsteady interaction between barwakes and stator passage �ow, time-accurate measurementdata of T106RUB blade pressures have been conducted. InFigure 6 the temporal evolution of local blade pressure �uctu-ations over two bar passing periods are given for pro�le �owat mid-span. Pressures on suction side (0 ≤ x/Cax ≤ 1) and onpressure side (-1 ≤ x/Cax ≤ 0) are shown for the two alreadyintroduced unsteady operating points with light and heavydisturbance and a third one with intermediate perturbation(Sr = 0.90, φ = 1.45). Similarly to equation (5), the respectivevalues at undisturbed condition have been subtracted fromthe time-resolved, phase-averaged pressures.

Figure 6. Time-space evolution of T106RUB pro�le pressure�uctuations at mid-span for three periodically disturbedoperating points.

For all operating points, periodical pressure �uctuationsat the blade surface caused by impinging bar wakes can beobserved. By analyzing the �uctuations (deviation from theundisturbed condition), the propagation of the wake distur-bance from LE across suction and pressure side towards TEcan be tracked. Wake kinematics, derived in the context ofFigure 4, can be applied to the pro�le �ow, as well. However,it must be noted that, as bar wake �ow around one singlepro�le is now considered in relation to the �ow of multiplebar wakes through a passage of adjacent pro�les (chapter3.1.2), bar wake paths and the con�guration of the induced�ow structures look di�erent. Furthermore, as static bladepressures are considered, the color scale is inverted. Parcelsof high velocity correspond to decreased static pressure andvice versa. Before the bar wake impinges at LE, it wrapsaround it and is cut by the LE. Volume C of squeezed passage�ow (high velocity and low static pressure) becomes visibleat the suction side and is transported downstream towardsTE. Immediately a�er the wake has passed, volume D (low


velocity, high static pressure) follows. �is corresponds tothe observations from the stator exit �ow �eld (chapter 3.1.2).

Under low disturbance frequency (Sr = 0.45, φ = 2.84), theperiod of time between individual wake events is su�cient forblade �ow to switch back to the unperturbed condition. Withraising disturbance frequency and thus Sr , the amplitude oftime-resolved pressure �uctuations increases signi�cantly,reaching maximum oscillations of ±10 % (relative to cascadeoutlet dynamic pressure) during one bar wake passing. �isresults in a highly periodically unsteady blade loading underwake impact. Furthermore, for the lightly disturbed case anadditional periodical oscillation can be observed close to TE(0.8 ≤ x/Cax ≤ 1), pointing toward the periodical suppressionof a laminar separation bubble. Under steady in�ow, the suc-tion side boundary layer is assumed to separate towards TE,forming a separation bubble. With an increasing disturbancefrequency the oscillations connected to this phenomenon arereduced. �is is indicative of the bubble not switching backto its undisturbed condition for su�ciently high bar wakefrequencies.

Figure 7. T106RUB pro�le pressure �uctuations at mid-span(Sr = 1.55, φ = 0.83). Temporal evolution for one bar passing,divided into 4 time instants.

�e impact of the passing bar wake on the pro�le �owcan furthermore be investigated using Figure 7, showingthe blade pressure �uctuations for four representative timeinstants of one bar wake passing, again for Sr = 1.55, φ =0.83. One sensor position on the suction side (x/Cax = 0.5)is highlighted with a red circle. For t/tBP = 0 this positionis already facing the approaching bar wake (parcel C) withdecreasing pressure, which further decreases at time instantt/tBP = 1/4. At t/tBP = 2/4 the wake has passed and parcel Dwith maximum local pressure is following. For t/tBP = 3/4the wake has nearly le� the pro�le. On the pressure sidesimilar behavior can be recognized with lower �uctuationamplitudes.

4. CONCLUSIONSWithin the present work experimental investigations havebeen performed, focusing on periodically unsteady �ow phe-nomena in a LPT stator passage of a 1.5 stage annular cascade.Conventional time-averaged measurement data is completedand combined with phase-averaged, highly time-resolveddata, thus simplifying analysis within the time-domain. �roughcombination of �ow �eld data in the stator exit �ow and bladepressure data a link could be established between secondary�ow structures downstream of the passage and underlyinge�ects directly at the blade. �e well-known secondary �owmodel could be applied to the stator exit �ow �eld. Analyz-ing the temporal velocity and vorticity evolution in threecharacteristic slices of constant span, vortex displacementand weakening could be described. Furthermore, periodi-cally unsteady structures could be identi�ed downstreamof the passage. �ese represent the remaining portion ofthe bar wakes, carrying a velocity �eld of four characteris-tic volumes of decreased and increased velocity. Finally theimmediate interaction between the bar wakes and T106RUBstator pro�le �ow was presented. Similarly to observationsmade in the passage exit �ow, on the blade suction side pres-sure �uctuations are enhanced considerably for the high Srcase. �e interaction of bar wake and pro�le boundary layer�ow could be characterized by the passing negative jet andits alternating sharp decrease and increase of local pressure,respectively velocity. �e negative jet e�ect, not only de-scribes a �uctuating velocity �eld, but also an injection ofturbulence at various scales into the near wall �ow.

ACKNOWLEDGMENTS�e investigations reported in this paper were conductedwithin the framework of the joint research project “UnsteadyFlow and Secondary Flow in Compressor and Turbine Cas-cades” (PAK-530). �e authors wish to gratefully acknowl-edge funding and support by the Deutsche Forschungsge-meinscha� (DFG). �e experimental investigations wouldnot have been possible without the e�orts of our technicalsta�. �e responsibility for the contents of this publicationlies entirely by the authors.

NOMENCLATURE

Latin and Greek Symbolsc [m/s] velocitycp [-] non-dimensional pressure distributionC [m] blade chord lengthD [m] diameterg [m] pitch (at mid-span)H [m] blade heightL [m] lengthMa [-] Mach numberp [Pa] pressure


Latin and Greek Symbols (continued)Re [-] Reynolds numberSr [-] Strouhal numbert [s] timex [m] axial direction

α [°] �ow angle in circumferential directionδ [°] �ow angle in radial directionθ [°] circumferential directionλ [°] stagger angleµ [Pa·s] dynamic viscosity of a �uidρ [kg/m3] density of a �uidφ [-] �ow coe�cient

Subscriptsax axial directionB barBP bar passingC casingH hubm mid-spanr radial directionth theoretical

Abbreviations5HP �ve hole probeAVO axial vorticityCTA constant temperature anemometryCSV concentrated shed vortexCV corner vortexHSV horse shoe vortexHSV-PL horse shoe vortex - pressure side legHSV-SL horse shoe vortex - suction side legIGV inlet guide vaneLE leading edgeLPT low pressure turbinePS pressure sidePV passage vortexSFP split �ber probeSS suction sideTE trailing edgeTEWV trailing edge wake vortex

REFERENCES[1] P. Puddu, C. Palomba, and F. Nurzia. Time-space evo-

lution of secondary �ow structures in a two-stage low-speed turbine. Number GT2006-90787. Proceedings ofASME Turbo Expo 2006: Power for Land, Sea, and Air,Spain, May 08-11 2006.

[2] L. S. Langston. Secondary �ows in axial turbines—areview. Annals of the New York Academy of Sciences,(934):11–26, 2001.

[3] S. Kang and C. Hirsch. �ree dimensional �ow in alinear compressor cascade at design conditions. Num-ber 91-GT-114. Proceedings of ASME 1991 International

Gas Turbine and Aeroengine Congress and Exposition,Orlando, Florida, USA, June 03-06 1991.

[4] R. J. Goldstein and R. A. Spores. Turbulent transporton the endwall in the region between adjacent turbineblades. ASME J. Heat Transfer, (110):862–869, 1988.

[5] W. R. Hawthorne. Rotational �ow through cascades. Q.J. Mech. Appl. Math., (8(3)):266–279, 1954.

[6] R. Ciorciari, I. Kirik, and R. Niehuis. E�ects of unsteadywakes on the secondary �ows in the linear T106 turbinecascade. ASME J. Turbomach., (136(9)), 2014.

[7] A. Krug, P. Busse, and K. Vogeler. Experimental investi-gation into the e�ects of the steady wake-tip clearancevortex interaction in a compressor cascade. ASME J.Turbomach., (137(6)), 2015.

[8] R. J. Volino. E�ect of unsteady wakes on boundary layerseparation on a very high li� low pressure turbine airfoil.ASME J. Turbomach., (134(1)), 2011.

[9] D. Lengani, D. Simoni, M. Ubaldi, P. Zunino, andF. Bertini. Time resolved PIV measurements of the un-steady wake migration in a LPT blade passage: e�ectof the wake passing frequency. Number ETC2017-324.Proceedings of 12th European Conference on Turboma-chinery, Fluid Dynamics and �ermodynamics, ETC12,April 3-7, 2017; Stockholm, Sweden.

[10] P. Lampart. Investigation of endwall �ows and lossesin axial turbines. part I. formation of endwall �ows andlosses. Journal of �eoretical and Applied Mechanics,(47(2)):321–342, 2009.

[11] H. P. Hodson and R. J. Howell. �e role of transition inhigh-li� low-pressure turbines for aeroengines. Progressin Aerospace Sciences, (41(6)):419–454, 2005.

[12] M. Sinkwitz, D. Engelmann, and R. Mailach. Experimen-tal investigation of periodically unsteady wake impacton secondary �ow in a 1.5 stage full annular LPT cascadewith modi�ed T106 blading. Number GT2017-64390. Pro-ceedings of ASME Turbo Expo, Charlo�e, NC, USA, June26-30 2017.

[13] M. Berrino, D. Lengani, D. Simoni, M. Ubaldi, P. Zunino,and F. Bertini. Dynamics and turbulence characteristicsof wake-boundary layer interaction in a low pressureturbine blade. Number GT2015-42626. Proceedings ofASME Turbo Expo 2015, June 15-19, Montreal, Canada.

[14] R. D. Stieger and H. P. Hodson. �e transition mecha-nism of highly loaded low-pressure turbine blades. Num-ber 126(4), pages 536–543, 2004.

[15] B. Winhart, M. Sinkwitz, D. Engelmann, F. di Mare, andR. Mailach. Experimental and numerical investigationof secondary �ow structures in an annular LPT cascadeunder periodical wake impact – part 2: Numerical results.Proceedings of ISROMAC, Maui, Hawaii, December 16-21 2017.

experimental and numerical investigation of secondary flow...

Documents