experimental performance analysis of an annular diffuser with and without struts

Experimental performance analysis of an annular di�user with andwithout struts

Stefano Ubertini, Umberto Desideri *

Department of Industrial Engineering, University of Perugia, Via G. Duranti 1A/4, 06125 Perugia, Italy

Received 8 May 2000; received in revised form 15 July 2000

Abstract

In this paper, the performance analysis of an annular di�user is presented. In a typical industrial gas turbine di�user, a certain

number of structural members, called struts, serve both as load bearings support and as passages for cooling air and lubricant oil.

Measurements were made in a 35% scaled down model of a PGT10 gas turbine exhaust di�user with and without struts in order

to determine the total and static pressure development and the e�ect of struts on both the local phenomena and the overall per-

formance. More realistic ¯ow conditions are made available by a ring of 24 axial guide vanes at inlet, which represent the last turbine

rotor. The model has been tested on a wind tunnel facility developed at the University of Perugia with inlet speed around 80 m/s,

allowing satisfactory accuracy for ¯ow measurements and similarity with the PGT10 di�user in terms of Reynolds number. Static

pressure taps located at various streamwise positions on the hub and the casing allowed the estimation of pressure recovery de-

velopment. A Pitot tube and a hot split-®lm anemometer were used to determine static and total pressure inside the di�user at

di�erent axial positions. The comparison between the two cases, with and without the struts, was made also by the use of global

parameters, which correlate static and total pressure.

In a previous paper, a detailed three-dimensional analysis of the ¯ow path inside the di�user was presented and the detrimental

e�ect of the struts, in terms of ¯ow separation and unsteadiness, was discussed. The stationary ¯ow measurements and the in-

vestigation of the di�user without the struts are presented in this paper. The whole research project represent a complete di�user

investigation available to develop an optimal design and to advance the computational and design tools for gas turbine exhaust

di�users. Ó 2000 Elsevier Science Inc. All rights reserved.

1. Introduction

The exhaust di�user of an industrial gas turbine re-covers the static pressure by decelerating the turbinedischarge ¯ow. This allows an exhaust pressure lowerthan the atmospheric one, thus increasing the turbinework. The Mach number at the modern turbine exhaustis around 0.4±0.45, with a resulting velocity of about250 m/s and a kinetic energy of about 30 kJ/kg. Con-sidering that the energy produced by a gas turbine isaround 350 kJ/kg, the exhaust ¯ow kinetic energy can be10% of the entire turbine work. It is thus clear that theexhaust di�user is a critical element in turbomachineenvironment in terms of e�ciency and stability.

A considerable amount of experimental and numer-ical investigations on simple di�users [1,2] can be foundin literature and the factors in¯uencing their perfor-mance are predominantly the area ratio and the lengthof the ¯ow path over which di�usion occurs. Globalparameters, relationships between static and totalpressure and performance maps are generally used todetermine di�usersÕ performance [3,4].

In di�users situated downstream a turbomachine, theinlet ¯ow presents a swirl component and a high level ofturbulence. Therefore, the di�user cannot be treated asan isolated element but it has to be seen as a componentof the whole system, included the turbomachine. Theincreased turbulent mixing at inlet, which result in alater onset of ¯ow separation, can sometimes improvedi�usersÕ performance [5]. Dominy et al. [6] showed thatin a S-shaped duct, the wakes created by the upstreamturbine lead to total pressure distortion and signi®cantlocal yaw and pitch angles. Moreover di�users behind aturbomachine have struts supporting loads and passages

Experimental Thermal and Fluid Science 22 (2000) 183±195

www.elsevier.nl/locate/etfs

* Corresponding author. Tel.: +39-075-5852736; fax: +39-075-

5852736.

E-mail addresses: [email protected] (S. Ubertini), umberto.desideri

@unipg.it (U. Desideri).

0894-1777/00/$ - see front matter Ó 2000 Elsevier Science Inc. All rights reserved.

PII: S 0 8 9 4 - 1 7 7 7 ( 0 0 ) 0 0 0 2 5 - X

for engine cooling and lubrication systems and thesestructural members, that extends radially from the innerto the outer annulus wall, act as blu� bodies and con-sequently cause unsteady wakes [7]. As is well known,within di�users, which are characterized by strong ad-verse pressure gradients, boundary layers grow rapidlyand tend to separate. To avoid unacceptable weightpenalties, the di�usion must occur in the shortest pos-sible length, while preventing ¯ow separation requiressmall divergence angles. The presence of a row of strutsinevitably causes a blockage and an acceleration of the¯ow, thus reducing the di�usion achieved. Ubertini andDesideri [8] showed how the interaction between the lastturbine rotor and the struts, supporting the shaft, pro-duce the onset of ¯ow separation. Experimental andnumerical investigations in a di�using S-shaped ductmade by Norris et al. [9] showed a 28% e�ciency re-duction, an almost doubled pressure loss coe�cient anda signi®cant rise of the separation bubble size when arow of struts are present in the duct.

Anyway very few examples of experimental analysisconcerning annular di�users downstream a turbine[10,11] or a compressor [12,13] can be found in theliterature survey. In a previous paper [8], the authorspresented a detailed three-dimensional investigation ofa scaled down model of an annular exhaust di�user,showing the turbulent ¯ow ®eld, the growing and theseparation of the boundary layer and the evolution of

the strutsÕ wakes along the duct. The aim of this paperis to determine the overall performance of the di�userand how it is achieved by local and detailed mea-surements of static and total pressure. The analysishas also been made in the di�user without struts inorder to understand and quantify their detrimentale�ect.

2. Experimental apparatus

Measurement were made in a di�user model built byNuovo Pignone S.p.A. reducing in 0.35:1 scale the an-nular exhaust di�user of the gas turbine PGT10 (Fig. 1).The model was designed to operate in geometric andReynolds number similarity with the PGT10 exhaustdi�user. The GT di�user operates with a Reynoldsnumber exceeding 106 and the modelÕs Reynolds numberis higher than 6� 105, that is high enough to assumesimilar ¯ow conditions with the real di�user.

A funnel-shaped inlet prevents ¯ow separation fromthe walls in the ®rst part of the di�user. The inlet sectionfeatures 24 axial guide vanes, which provide both as ameans of introducing swirl into the test section as well asof producing wakes representative of those produced bythe last turbine rotor of the real PGT gas turbine. Thedi�user length is 450 mm and the inlet and outer di-ameters are 190 and 320 mm, respectively. To keep

NomenclatureA1 cross-surface at inlet, m2

A2 cross-surface at outlet, m2

Cp pressure recovery coe�cient, dimensionlessparameter

Cpi ideal pressure recovery coe�cient,dimensionless parameter

Pst1 static pressure at inlet, kPa

Pst2 static pressure at outlet, kPaPt1 total pressure at inlet, kPaPt2 total pressure at outlet, kPaK pressure losses coe�cient, dimensionless

parameterR20 electric resistance at 20°C, Xa20 temperature coe�cient at 20°C, °Cÿ1

g di�user e�ciency, dimensionless parameter

Fig. 1. Di�user model.

184 S. Ubertini, U. Desideri / Experimental Thermal and Fluid Science 22 (2000) 183±195

geometric similarity of the model with the GT di�user,the area ratio is maintained as the same along the axis inboth the model and the GT di�user. The length to wideratio is 2.01.

The transparent plastic shell allows checking theposition and the movement of the probes.

The PGT10 di�user has six high solidity struts, whichsupport one of the shaft bearings and have piping oilsupply inside for bearings lubrication. These struts ex-tend radially from the hub to the shell and are repro-duced in the model with ®ve 35% scaled down plasticstruts of the same shape. The inner hub is a stainlesssteel cylinder with a diameter of 190 mm and it is ro-tatable over 360°.

The di�user is connected to the suction side of asubsonic wind tunnel, located at the University of Pe-rugia. A 21 kW centrifugal fan provides ¯ow prevalenceand the ¯ow rate is controlled by a sluice gate positioned3.4 m before the exhaust of the wind tunnel. Thetransparent plastic di�user was designed and made in-tegral with the pipe section of the wind tunnel to avoid alip at the connection of pipe and di�user.

3. Instrumentation

A Pitot tube was used to measure static and totalpressure along the di�user. The static and the total holeshave a nominal internal diameter of 0.7 mm. The probecould be traversed into the di�user through openingslocated along the di�user wall at various axial locations.Since cross-¯ow velocity components prevent the for-mation of a true stagnation point at the total pressurehole, the Pitot tube have been designed for misalignmentinsensitivity with less than 1% error at 10°. The e�ect ofviscosity is negligible since the ReynoldsÕ number is over105 in the duct [14].

The misalignment of the velocity vector lead to errorfor the static pressure measurement too, since the tapsare exposed to some components of velocity. Therefore,static pressure is measured by four taps equally spacedaround circumference, thus, if we consider that the ve-locity vector is the same in those four positions, theerrors due to misalignment are compensated [14]. Anoverall error of 1.5% at 10° can be considered. The in-¯uence of the stagnation point on the tube supportleading edge causes an increase of the static pressureupstream if the static taps are too close to the support.The ratio between the diameter of the tube and the statictaps to support distance is 13, producing an error ofaround 0.5% [14]. This error tends to be cancelled by thee�ect of the presence of the tube in the ¯ow. In fact,since streamlines next to the tube must be longer thanthose in undisturbed ¯ow, there is an increase in velocityand then a reduction of static pressure. For our Pitottube, the ratio between total and static taps distance andthe tube diameter is 7.4, thus producing an error ofaround 0.8% [14].

Flow angles through the di�user were measured by atwo-dimensional hot split ®lm probe, operating in con-

stant temperature mode with a mean wire temperatureof 200°C:· DANTEC 55R56 split ®lm probe with R20 � 4:50 X

and 4.80 X and a20 � 0:0039°Cÿ1 and 0.0041°Cÿ1.The probe consists of two parallel, electrically separatedNickel ®lms of 3 lm thickness, deposited on a 200 lmdiameter quartz ®ber. The active sensor length is 1.25mm. An uncertainty of 5% on the ¯ow angle has to betaken into account [8]. Four automated carriages lo-cated inside the hub allow probe radial movement insidethe di�user. The probes could be traversed into thedi�user through four openings made along the hub.

Wall static pressure was measured by wall taps of1.5 mm inside diameter mounted at di�erent stationslocated along the di�user shell and the di�user hub.Special care was taken to properly align the hole of thetaps with the inside surface of the di�user wall for cor-rect measurements. For static pressure measurements atthe hub, the stainless steel hub was substituted with aplastic one of the same shape. A stepping motor allowedthe rotation of the hub in order to measure static pres-sure at di�erent circumferential stations. A two channelMSTEP ®ve motor management board permits thecontrol of the stepping motor by PC. The pneumaticprobe and the taps readings were recorded into a PCthrough a 32-channels pressure scanner by pressuresystems. A 16-channel fast sampling board (100 kHz),connected to a PC and featuring a 12 digit A/D con-verter reads voltage output of the split-®lm probe. Thesampling frequency of sensors acquisition has been setto avoid the in¯uence of pressure ¯uctuations in themeasurements of average quantities.

4. Results and discussion

4.1. Introduction to measurements

This paper presents the performance of the exhaustdi�user of the PGT10 industrial gas turbine, in terms ofpressure recovery and total pressure losses measured in ascaled down model. The real di�user features six struts,that support one of the shaft bearings and contains lu-bricant supply. Measurements were taken in the modelwith and without the struts, in order to understand andquantify their detrimental e�ect. In the next graphs andcontour plots, we will refer to the position of the mea-suring point in terms of axial position, radial positionand angle or circumferential position considering that:· axial position of 0 mm is the leading edge of the inlet

guide vanes;· radial position is measured from the hub to the shell;· results were produced in the di�user model with struts

for one 60° sector containing one strut in the middleand in the di�user model without struts for one 7.5°sector with one of the inlet guide vanes located at 0°.

Pressure is presented as relative to atmospheric pressureand kilopascal is used as the measuring unit.

The di�userÕs performance have been also determinedby the following parameters:

S. Ubertini, U. Desideri / Experimental Thermal and Fluid Science 22 (2000) 183±195 185

· pressure recovery coe�cient

Cp � pst1 ÿ pst2

pt1 ÿ pst1

;

· ideal pressure recovery coe�cient

Cpi � 1ÿ A1

A2

� �2

;

· di�user e�ciency g � Cp=Cpi;· pressure losses coe�cient K � Cpi ÿ Cp;where subscripts 1 and 2 are the inlet and the outletsections, respectively.

4.2. Results and discussion ± di�user with struts

Fig. 2 shows the wall static pressure developmentalong the shell, measured by two rows of 11 wall staticpressure taps at 0°, the midpoint between the struts, andat 30°, behind one of the struts. The axial locations ofthe wall taps at both the casing and the hub are shownon Table 1. In the 30° row, the three wall taps from the

®fth to the seventh are located around the strut asshown in Fig. 3.

In the ®rst part of the duct, the pressure rise at 30° ismuch higher than that at 0°. This is due to the stagna-tion zone produced by the downstream strut, wheredynamic pressure is converted into static pressure. Onthe other hand, the presence of the struts reduce thecross-passage section between them and the pressuregradient decreases, as can be detected by the staticpressure development at 0°. The sudden collapse ofstatic pressure at the fourth wall tap, the ®rst onearound the strut, is due to the ¯ow separation fromthe strut. Pressure recovery has the same trend at both0° and 30° from around 10 cm before the exhaust to theexhaust itself, since the ¯ow returns to be axial andthe in¯uence of the strutÕs wake is reduced. In the lastthree wall taps, a sudden decrease and a even negativepressure recovery gradient can be detected, probablydue to the ¯ow separation from the shroud, that reducesthe ¯ow passage section and anyway causes errors inpressure measurements.

Wall static pressure data at the hub were acquired fora 30° sector with steps of 5°, since the ¯ow ®eld repeatsitself every 30° and results are reported for a 60° sectorwith the strut in the middle. The hub has been rotated bysteps of 5°. The pressure distribution along the ductmeasured at the hub (Fig. 4) clearly demonstrates thee�ect of the strut, since static pressure around 30° in-creases rapidly in the ®rst 8 cm and then collapses tovalues close to the initial one. The most rapid di�usionoccurs in the last 15 cm, where static pressure increasesfrom ÿ2.9 kPa to around ÿ2.0 kPa. The hub and casingwall pressure development are compared in Fig. 5. In thevery ®rst part of the di�user, pressure measurements atthe hub and at the casing are quite di�erent and this isprobably a consequence of the non-axial component ofvelocity, due to the inlet guide vanes. In the last part ofthe duct the static pressure at the hub is slightly lowerthan that at the casing.

Total and static pressure and ¯ow angles measure-ments have been made at three di�erent sections named

Fig. 2. Static pressure development along the shell.

Table 1

Axial distance (mm) of the wall taps from the leading edge of the inlet

guide vanes

Casing Hub

5 5

39 40

73 74

109 109

139 140

172 171

204 201

237 231

267 262

297 293

327 323

Fig. 3. Wall taps arrangement.


as sections 1, 2 and 3 (Fig. 6). Results are produced for a60° sector, but measurements were made in a 30° sectorwith steps of 5° for pressure measurements and 2° for¯ow angles measurements. The probes have been movedin radial direction by steps of 10 mm starting at 5 mmfrom the hub. The distance between the static and thetotal taps in the Pitot tube is 1 cm; therefore, the axial

position of the static pressure distribution at each sec-tion is 1 cm before the axial position of the total pressuredistribution. The axial positions for total and staticpressure measurements at the three sections are sum-marized in Table 2.

As shown in Fig. 7, the ¯ow angles for the threesections are less than 10° in all the ¯ow path of eachmeasurement section, thus producing a misalignmenterror less than 1%. As shown in a previous work [8], theradial component of velocity can be neglected. Staticpressure and total pressure contour plots for sections 1,2 and 3 are shown in Figs. 8 and 9.

At section 1, just behind the inlet guide vanes staticpressure varies from around ÿ4.0 kPa to around ÿ3.2kPa. This value is higher than that measured by the walltaps at the same axial position both at the hub and at thecasing. The measurements made by the pneumatic probeis probably more accurate since the inlet guide vanescould produce a separation bubble in the very ®rst partof the di�user, as mentioned before. Going from the sideof the measuring sector to the center the static pressurerise, due to the slowing down of the ¯ow behind the inletblades, is clearly detectable. Moreover, the in¯uence ofthe downstream strut produces a static pressure rise inthe region around 30°. Inlet guide vanes wakes areshown in the total pressure distribution at section A,where getting closer to the blades locations total pres-sure collapse from around ÿ0.36 kPa to values betweenÿ0.6 and ÿ1.3 kPa. Total pressure tends to decreaseapproaching to the hub and to the casing.

The strut wake is emphasized in section 2 totalpressure distribution. At 30°, in fact, a stagnation zonewith total pressure almost equal to static pressure isshown. Total pressure losses are also detectable close tothe endwall regions, due to the development of theboundary layer at the hub and the onset of ¯ow sepa-ration from the shroud. Anyway, since turbulence in-tensity behind the strut reaches high values, errors inpressure measurements behind the strut are to be takeninto account. Total pressure increases between thestruts.

Static pressure tends to decrease in the regions closeto the hub and to the shroud between the strutÕs and theinlet guide vanes circumferential position. As shown in aprevious work [8], in fact, the interaction between thestrutÕs and the inlet guide vanes wakes causes the onsetof ¯ow separation from the walls and then those regionsare characterized by a very high turbulence intensity.Away from the strutÕs wake static pressure is almost

Fig. 6. Measurement sections arrangement.

Fig. 5. Hub and casing static pressure development.

Fig. 4. Static pressure development at the hub.

Table 2

Axial distance (mm) of the total and static pressure measuring point

from the leading edge of the inlet guide vanes

Total pressure Static pressure

Section 1 42 52

Section 2 234 244

Section 3 311 321


constant with values between ÿ2.3 and ÿ2.4 kPa, ac-cordingly with the wall static pressure at 0° at the sameaxial position. Once again, going from the shroud to thehub static pressure tends slightly to increase.

Just before the exhaust, at section 3, the e�ect of thestrut wake can still be detected, since total pressure de-creases approaching 30°. Wider regions of high-pressurelosses close to the endwalls are shown, due to separationfrom the hub and from the shroud. The separated regiongets wider and wider approaching the strut wake. Awayfrom the walls, between the struts total pressure lossesare negligible. Static pressure is quite constant at the endof the duct, with values between ÿ1.92 kPa close to thecasing and ÿ2.15 kPa close to the hub, in perfectagreement with wall static pressure which, at almost thesame axial location, is around ÿ1.9 kPa at the casingand around 2.12 kPa at the hub.

The data obtained by the pneumatic probe allows tocalculate the pressure recovery coe�cient and the e�-ciency of the di�user (Table 3).

The average static pressure measured by the Pitottube rises from ÿ3.44 kPa at section 1 to ÿ2.01 kPa

at section 3, with a resulting pressure recovery coe�cientof 0.526, calculated by using the dynamic pressuremeasured at section 1. Since the ideal pressure recoverycoe�cient between sections 1 and 3 is 0.663, a di�usere�ciency of 79% is obtained. The pressure recoverycoe�cient measured through the wall taps at the shell is0.581 and that measured at the hub is 0.512. Consideringthat the ideal pressure recovery coe�cient between the®rst wall tap and the last one is 0.678 at the casing and0.674 at the hub, a di�user e�ciency of 85% consideringthe static pressure at the shell and of 76% consideringthe static pressure at the hub are obtained.

4.3. Results and discussion ± di�user without struts

A second set of measurements was made in the dif-fuser model without the struts. The Pitot tube was tra-versed inside the di�user through three rows of nineholes at di�erent axial locations (Table 4) and covering asector of 7.5°, since the ¯ow path in the model withoutthe struts repeats itself every 7.5°. The angle of 0° cor-responds to one of the inlet guide vanes circumferential

Fig. 7. Flow angles contour plots.


position. The measuring point was radially moved fromthe hub to the casing by steps of 10 mm.

Fig. 10 shows the total pressure development throughthe di�user at three di�erent radial position for 0°, 3.75°and 7.5°. Behind the inlet guide vane, we have a totalpressure rise through the duct, since the e�ect of thebladeÕs wake tends to decrease. This is only a local e�ectand cannot be considered a total pressure recovery,which is obviously not possible. The losses due to thewake, in fact, that are concentrated just behind the bladeat the di�user inlet tend to spread out along the duct.This means that even if the local e�ect is a total pressurerise, the global e�ect, as it will be shown, is a totalpressure loss.

The region between the blades is instead character-ized by a regular pressure loss through the duct. Nearthe hub, as expected, a higher pressure loss is detected,due to the growth of the boundary layer and to aprobable ¯ow separation.

The average total pressure at 0°, 3.75° and 7.5°,shown in Fig. 11, con®rms that behind the inlet bladethere is a slight total pressure rise from around ÿ1.4 toÿ1.2 kPa and between the blades an evident pressureloss occurs. Making an average for the three angular

positions a pressure loss of around 0.4 kPa characterizesthe di�user without struts.

Fig. 12 shows the development of total pressurepro®le at di�erent axial locations at each of the threesample circumferential locations and non-uniformity ofthe ¯ow is clearly evident. As expected, total pressuregradient gets higher approaching the casing and the hub.Moreover, unless behind the inlet blade, total pressurepro®le in the center of the traverse plane is almostconstant, revealing a fully developed turbulent ¯ow.Total pressure data in the endwall regions must becarefully considered, since the high unsteadiness of the¯ow and probable ¯ow reversals can easily lead to errorsfor pneumatic probes.

Static pressure axial developments, shown for 0°,3.75° and 7.5° in Fig. 13 at di�erent radial positions,emphasize the regularity of static pressure recoverythrough the duct. No signi®cant di�erences can bedetected between one and any other radial position.Moreover we see from Fig. 14 that the development ofthe mean static pressure at 0°, 3.75° and 7.5° is veryclose to the ideal one. The pressure recovery coe�cientcalculated from the mean static pressure at the ®rstand the last axial location is 0.616. Since the ideal

Fig. 8. Static pressure contour plots


pressure recovery coe�cient is 0.676, a di�user e�-ciency of 91% is obtained. It is useful to remind thatthe ideal pressure recovery was calculated from the®rst axial position, just behind the inlet guide vanes,and then the losses due to the inlet blades were nottaken into account.

The static pressure pro®les (Fig. 15) con®rm thatstatic pressure is quite constant along radial direction,even if there is a slight reduction approaching to thewalls. The reduction of the static pressure gradient ap-proaching the exhaust is clearly detectable and it isprobably caused by the reduction of the cross-passagesection produced by ¯ow separation from the walls.Therefore, even without a row of struts, the di�usionrate in the last part of the duct is too high for the ¯ow toremain attached.

Wall static pressure was measured at the casing in thesame 11 axial stations used for the di�user with struts at0°, and then behind one of the inlet guide vanes. Asshown in Fig. 16, where static pressure development iscompared with the ideal one, the presence of the inletblade causes a very low static pressure in the ®rst twostations. In fact, the interaction between the boundarylayer and the inlet guide vanes wakes produces a sepa-

Fig. 9. Total pressure contour plots.

Table 4

Axial distance (mm) of the Pitot tube measuring points from the

leading edge of the inlet guide vanes

Total pressure Static pressure

30 40

63 73

95 105

125 135

157 167

188 198

221 231

252 262

284 294

Table 3

Global parameters for the di�user with struts

Sections 1±3 Casing Hub

Cpi 0.663 0.678 0.674

Cp 0.526 0.581 0.512

g 0.79 0.85 0.76

K 0.137 0.097 0.162


ration bubble in the very ®rst part of the di�user andthen a collapse of the static pressure. From the secondstation to the exhaust of the duct, the static pressuredevelopment is almost the same of that measured by thePitot tube. The pressure recovery coe�cient calculatedthrough the wall taps measurements is 0.629, while theideal one is 0.678. The di�user e�ciency is thereforearound 93%.

Wall static pressure at the hub was measured at 0°,just behind one of the inlet guide vanes, and at 7.5°,centered between the two blades. The static pressuredevelopment at the hub, shown in Fig. 17, con®rms thee�ect of the interaction between the inlet guide vaneswakes and the wall boundary layers, that causes a lowerstatic pressure at 0°. The static pressure recovery coef-®cient calculated at the hub is 0.61, and the resultingdi�user e�ciency is around 90%.

The performance coe�cients calculated for the dif-fuser without struts are summarized in Table 5.

4.4. Results and discussion ± comparison

The two di�erent situations examined can be com-pared through the performance coe�cients summarizedin Tables 3 and 5. The overall performance of the dif-fuser is highly in¯uenced by the presence of the struts,since passing from the situation with struts to the situ-ation without struts, the e�ciency increases of about 10±5% and the pressure losses coe�cient passes from amean value of about 0.14 to a mean value of about 0.06.The losses due to the struts can be clearly detected bylooking at the total pressure distribution at section 2(Fig. 9).

Apart from the obvious di�user performance rise inthe di�user without struts, the comparison of thepressure recovery coe�cient development through thedi�user is very interesting in some situations with andwithout struts, and is shown in Fig. 18. In the ®rstpart of the di�user negative Cp can be noticed, becauseCp was calculated through the mean values of thestatic and the dynamic pressure measured by the Pitottube and the wall taps. Therefore for all the axialstations located before the ®rst axial position of thePitot tube, we obtain a negative Cp and, moreover,locally static pressure can be lower than the meanstatic pressure at the ®rst station of the Pitot tube.Pressure recovery in the model without struts increasesmore rapidly in the ®rst part of the duct, and this isdue to both the strutsÕ e�ect of causing losses and ofreducing the ¯ow passage. However, after the strutaxial position this situation is reversed and Cp has ahigher gradient in the model with struts. In the regionbetween the struts, the reduction of the ¯ow passage,due to the struts presence, reduces di�usion and dy-namic to static pressure conversion is slackened, ascan be detected combining total and static pressure

Fig. 10. Total pressure developments at 0°, 3.75° and 7.5°.

Fig. 11. Mean total pressure development at 0°, 3.75° and 7.5°.


measurements shown in Figs. 7 and 8. The kineticenergy not recovered for the struts blockage can beconverted into potential energy in the last part of thedi�user. This explains the pressure recovery gradientrise behind the struts.

The ¯ow, just beyond the struts, ®nds a sudden riseof the cross-passage section and from an ideal point ofview the dynamic pressure not converted until thenshould become static pressure. Getting farther from thestruts, however, the ¯ow passage is reduced by thestrutsÕ wakes, the e�ect of which tends to reduce whileapproaching the exhaust. After the struts, since thecross-passage section rise is not only due to the di�userslope but also due to the disappearance of the strutsÕwakes, the static pressure gradient is higher. This

Fig. 13. Static pressure developments at 0°, 3.75° and 7.5°.

Fig. 12. Total pressure pro®les at 0°, 3.75° and 7.5°.

Fig. 14. Mean static pressure development at 0°, 3.75° and 7.5°compared with the ideal one.


means that in the di�user with struts the highestdi�usion occurs in the last part of the duct. On theother hand, the di�user end is characterized by ¯owseparation from the walls, as shown in a previouswork. Therefore, the presence of the struts in an an-nular di�user causes a ¯ow blockage and a slackeningof the di�usion. Part of the kinetic energy can be re-covered after the struts but the presence of the strutsÕ

Fig. 15. Static pressure pro®les at 0°, 3.75° and 7.5°.

Fig. 16. Static pressure development at the shell compared with the

ideal one.

Fig. 17. Static pressure development at the hub.

Table 5

Global parameters for the di�user without struts

Pitot measurements Casing Hub

Cpi 0.676 0.678 0.674

Cp 0.616 0.629 0.61

g 0.91 0.93 0.90

K 0.06 0.049 0.064

Fig. 18. Pressure recovery coe�cient development with and without

the struts.


wakes and ¯ow separation causes losses and reducespressure recovery, since they attenuate the e�ective riseof the cross-passage section.

5. Practical signi®cance and usefulness

As outlined in Section 1, di�users behind a turb-omachinery have a deeply di�erent behaviour fromsimple di�users, where the lack of obstacles and inletdisturbs allow an easier computational analysis [1±4].On the other side, theoretical analysis of di�users inturbomachine environment cannot be put aside from theexperimentation [5±14].

In a previous work [8], the distortion of the ¯owproduced by the struts and the inlet guide vanes and theseparation from the walls in a gas turbine di�user modelhave been presented. In this paper, the detrimental e�ectof the struts has been quanti®ed in terms of overallperformance and pressure losses. Moreover, the detailedinvestigation with pneumatic probes has allowed theidenti®cation of the local phenomena that in¯uence theentire system behaviour.

The experimental results presented in this paper arean important and detailed set of data that can be used tovalidate computational predictions on the duct with andwithout the struts. Moreover the e�ects, both local andglobal, of the inlet guide vanes and of the struts on theperformance of the di�user showed in this work can helpthe development of new and more performant designs ofGT di�users.

The di�user have been studied with and without thestruts to outline their detrimental e�ect. A possible fu-ture application could be the investigation of the dif-fuserÕs behaviour with struts of di�erent shapes in orderto develop an optimal struts design.

6. Conclusions

An experimental study has been performed to in-vestigate the static and total pressure developmentthrough an industrial gas turbine exhaust di�user. Theanalysis allowed the observation of both the local phe-nomena and the overall duct performance. The attentionhas been focused on the e�ect of a ring of struts, which isquanti®ed showing the di�erences between the duct withand without the struts.

From the above discussion the following conclusionscan be made:· the di�usion in the duct with struts is interrupted by

the reduction of the cross-passage section, due tothe struts and their wakes; this means that the ¯owpotentially has more di�usion to achieve; consequent-ly, a higher pressure recovery gradient is observedbehind the struts;

· even in the empty duct, a reduction of the pressurerecovery gradient is observed in the very last partof the di�user, due to the separation of the ¯ow fromthe walls; this e�ect is more evident in the duct withstruts where the interaction between the inlet guide

vanes and the struts anticipate the separation pro-cess;

· in both cases, e�ciency calculated at the shroud ishigher than that calculated at the hub and this hap-pens probably because the hub diameter is constantalong the duct and then in the upper part of the ductan higher di�usion occurs;

· e�ciency in the duct with struts is 10±15% lower thanin the empty duct; this reduction lead to a signi®cantloss in the whole turbomachinery system;

· pressure recovery development in the di�user withoutstruts is very close to the ideal one;

· comparing the two cases, with and without the struts,it is clear that the overall di�user loss is signi®cantlyincreased by the struts and this loss rise mainly occursin the axial region of the struts and in the endwall re-gions, where ¯ow separates from the hub and thecasing.

The experimental results presented in this paper are animportant and detailed set of data that can be used tovalidate computational predictions on the duct with andwithout the struts. Moreover, the e�ects of the inletguide vanes and of the struts on the performance of thedi�user showed in this work can help the developmentof new and more performant designs of GT di�users.

References

[1] R.K. Singh, R.S. Azada, Measurement of instantaneous ¯ow

reversals and velocity ®eld in a conical di�user, Experimental

Thermal and Fluid Science 17 (1995) 100±106.

[2] P.W. Runstadler, F.X. Dolan, R.C. Dean, Di�user Data Book,

Creare, TN-186, 1975.

[3] D. Japikse, A new di�user mapping technique ± Part 1: Studies in

component performance, ASME Journal of Fluids Engineering

108 (1986) 148±156.

[4] S.J. Kline, D.E. Abbott, R.W. Fox, Optimum design of straight-

walled di�users, ASME Journal of Basic Engineering 91 (1959)

321±330.

[5] T. Zierer, Experimental investigation of the ¯ow in di�user

behind an axial ¯ow compressor, ASME Journal of Turboma-

chinery 117 (1995) 231±239.

[6] R.G. Dominy, D.A. Kirkham, A.D. Smith, The in¯uence of

blade wakes on the performance of inter-turbine di�users, ASME

paper 94-GT-207, 1994.

[7] T.F. Fric, R. Villarel, O. Auer, M.L. James, D. Ozgur et al.,

Vortex shedding from struts in an annular exhaust di�user,

ASME paper 96-GT-475, 1995.

[8] S. Ubertini, U. Desideri, Flow development and turbulence

length scales within an annular gas turbine exhaust di�user,

Experimental Thermal and Fluid Science, accepted for publica-

tion, paper no. 99-11-GP-T153, 2000.

[9] G. Norris, R.G. Dominy, A.D. Smith, Strut in¯uences within a

di�using annular S-shaped duct, ASME paper 98-GT-425,

1998.

[10] U. Desideri, G. Manfrida, Flow and turbulence survey for a

model of gas turbine exhaust di�user, ASME paper 95-GT-139,

1995.

[11] B. Djebedjian, J. Renaudeaux, Numerical and experimental

investigation of the ¯ow in annular di�user, Proceedings of

FEDSMÕ98, Washington, DC, FEDSM98-4967, 1998.


[12] H. Pfeil, M. Going, Measurements of the turbulent boundary

layer in the di�user behind an axial compressor, ASME Journal

of Turbomachinery 109 (1987) 405±412.

[13] H. Harris, I. Pineiro, T. Norris, A performance evaluation of a

three splitter di�user and vaneless di�user installed on the power

turbine exhaust of a TF40B gas turbine, ASME paper 98-GT-284,

1998.

[14] E.O. Doebelin, Measurement Systems Application and Design,

fourth ed., McGraw-Hill, New York, 1990.


experimental performance analysis of an annular diffuser with and without struts

Documents