Published on

IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com

Published in: Technology, Business
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide


  1. 1. Nirmal K. Das, B. Halder, P. Ray, B. Majumdar / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.711-716711 | P a g eDevelopment of Flow within a Diffusing C-Duct –ExperimentalInvestigation and Numerical ValidationNirmal K. Das1, B. Halder2, P. Ray3, and B. Majumdar41(Department of Mechanical Engineering, BCET, Durgapur, West Bengal, India)2(Department of Mechanical Engineering, NIT Durgapur, West Bengal, India)3(Department of Civil Engineering, NIT Durgapur, West Bengal, India)4(Department of Power Engineering, Jadavpur University, Kolkata, West Bengal, India)ABSTRACTExperimental investigation of flowdevelopment within a rectangular 90º curveddiffusing C-duct of low aspect ratio and arearatio of 2 was carried out and the three-dimensional computational results are thencompared with the experimental results fornumerical validation. All measurements weremade in a turbulent flow regime (Re = 2.35x105),based on the duct inlet hydraulic diameter (dh =0.0666m) and mass averaged inlet velocity of60m/s. Wall pressures were measured throughwall pressure taps. The mean velocities, staticand total pressures at six cross-sectional planes(along the centreline of the diffuser) wereobtained using a pre-calibrated three-holepressure probe in null mode. The flow wasnumerically simulated by using different viscousmodels available in the commercial CFDsoftware FLUENT 6.3. The experimental inletand boundary conditions were used as input forthe computation and validation of the numericalresults.Based on the comparison, it is found thatStandard k–ε turbulence model provides betterprediction of flow field in the diffusing C-duct.Numerical results of coefficient of mass averagedstatic pressure recovery (61.6%) and coefficientof mass averaged total pressure loss (9.1%) are ingood agreement with the experimentally obtainedstatic pressure recovery and total pressure losscoefficients of 52.48% and 15.5% respectively.Keywords - Aspect ratio, area ratio, coefficient ofstatic pressure recovery, C-shaped diffuser,secondary motion, wall pressure.NOMENCLATUREAR Area ratio (W 2/W1)AS Aspect ratio (b/W 1)b Height of diffusercc Concavecv ConvexCpr Coefficient of static pressure recovery [(Ps-Psavi)/Pdavi]ξ Coefficient of total pressure loss [(Pt-Ptavi)/Pdavi]De Dean number [Re/√ (Rc/dh)]dh Hydraulic diameter[2 W1b/( W1+b)]Pdavi Average dynamic pressure at inlet plane(0.5ρ Uavi2)Ps Average static pressure at a planePsavi Average static pressure at inlet planeRc Mean radius of curvatureRe Reynolds number (Uavi dh/ν)Uavi Inlet average velocityw Width at a measuring sectionW1 Inlet widthW2 Outlet widthν Kinematic viscosity2θ Total divergence angleΔβ Angle of turn of the centerline1. INTRODUCTIONIn many engineering applications, diffusersare used to convert the dynamic pressure into staticpressure. The importance of the diffuser as a single,useful, fluids handling element in wind tunnels,turbo-machinery, and as interconnecting flowpassage between the components of gas turbines hasbeen widely known. Understanding of diffuserflows, therefore, is of paramount importance to thedesign of fluid-flow systems. Diffusers are designedin different shapes and sizes to meet the specificapplication. Curved diffusers of different centre lineshapes find wide uses in the field of aircraftapplications to satisfy design compatibility andspace restrictions. Flow characteristics in curveddiffusers are most complicated due to the influenceof centerline curvature, different geometricalparameters like total angle of divergence (2θ), angleof turn (Δβ), area ratio (AR), inlet aspect ratio (AS),centerline shape, etc. as well as the dynamicalparameters like inlet Reynolds number, inletturbulence, etc.In curved channels the radial pressuregradient, resulting from the centrifugal force actingon the fluid due to the centerline curvature, canproduce significant secondary flows. In addition, theadverse stream wise pressure gradient, resultingfrom the diverging flow passage of curved diffusers,can lead to flow separation. The combined effectmay result in non-uniformity of total pressure andtotal pressure loss at diffuser exit, thus affecting thediffuser performance.
  2. 2. Nirmal K. Das, B. Halder, P. Ray, B. Majumdar / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.711-716712 | P a g eFrom the beginning of the last century, study ofinternal flow through ducts was the area of interestfor many researchers. With the advancement ofaircraft industry, the flow field characteristics andperformance of curved diffusers have been theresearch topics of interest.Williams et al., 1902, [1], were the first toinvestigate flow in curved pipes, where theyobserved the maximum axial velocity to shifttowards the outer wall of the curved pipe. Sincethen, numerous investigations have been carried outon curved pipes/ducts to establish the effect ofvarious parameters. According to these studies, thesecondary flows are generated in the curved pipeseither due to the non-uniform distribution ofvelocity at the inlet and/or due to the effect ofcentrifugal force.The complex nature of flow within curvedpipes was investigated experimentally by Eustice,1910, 1911, [2, 3], where he has pointed out theformation of secondary flow as one of the main flowcharacteristics in curved pipes.Probably, Fox & Kline, 1962, [4], carriedout the first systematic studies of flow regimes oncurved diffusers and additional studies of flowregimes in 2-D straight walled diffusers. Theydeveloped maps of flow regimes for stalled andunstalled curved diffusers for a range of flowturning angle from 0 to 90º, in steps of 10º, and withcurved diffuser geometry having a circular arccentre line and a linear area distribution normal tothe centre line.Bansod and Bradshaw, 1972, [5], madeexperimental investigation of flow through anumber of S-shaped ducts of different geometricparameters, each with a thin inlet turbulentboundary layer. From the measured total pressure,static pressure, surface shear stress and flow yawangle they have established that the large departurefrom axisymmetry of flow in the exit plane of the S-duct were due to the expulsion of boundary-layerfluid by a strong longitudinal pair of contra-rotatingvortices embedded in the boundary layer.Sajanikar et al., 1982, [6], studied the effect ofturning angle and Reynolds number on theperformance of curved diffusers in terms of pressurerecovery. Experiments were conducted on fixed arearatio curved diffusers by varying the turn anglebetween 0° to 90°, in steps of 15° and the Reynoldsnumber between 2x104to 5x105. From the obtainedresults they reported that, for a given curved diffuserthe pressure recovery decreased with increase inturn angle but increased with increase in Reynoldsnumber.Computational analysis of fully-developedturbulent flow of an incompressible viscous fluid incurved ducts of square section was reported by Hurand Thangam, 1989, [7]. They used finite volumemethod in their numerical study with a nonlinear K-lmodel to represent the turbulence. The results ofboth straight and curved ducts were presented,where in case of fully-developed turbulent flow instraight ducts the secondary flow was reported by aneight vortex structure. With introduction ofmoderate curvature authors observed a substantialincrease in the strength of secondary flow with achange in the pattern to either a double-vortex or afour-vortex configuration. Their computed results incomparison to available experimental data werereported to be in good agreement. In their opinionthe nonlinear K-l and K-ε models have the promiseto yield more accurate predictions for curvedturbulent flows than the more commonly used eddyviscosity models.By carrying out flow visualisation and wallstatic pressure measurements on an elliptical centreline 90º curved diffuser of large area ratio (AR =3.4), Agrawal and Singh, 1991, [8], detected largeseparation pockets on the convex wall. Theseparated flow affected the diffuser performanceseverely and the performance evaluation showedpoor pressure recovery and high losses.Majumdar and Agrawal, 1996, [9],performed an experimental study for air flow in alarge area ratio curved diffuser (AR = 3.4, Δβ = 90o,and AS = 0.685) by installing a row of vanes atdifferent angles to the diffuser inlet. Significantdevelopment of flow distribution inside diffuserwith increased pressure recovery coefficient wasachieved when the flow was deflected by 10otowards the convex wall.In a later study, Majumdar et al., 1998,[10], made detailed experimental investigation offlow in a high aspect ratio (AS = 6) 90ocurveddiffuser. The observed shift of stream wise bulkflow towards the concave vertical wall wasattributed to the centrifugal effect arising out of thecenterline curvature. The wall static pressureincreased continuously on both the convex andconcave walls due to diffusion and a pressurerecovery coefficient of 50% was obtained. They alsoobserved that except for a small pocket on theconvex wall at the exit there was no flow separationin the diffuser. Counter rotating vortices weredetected from 30º turn of the diffuser.Experimental and numerical studies ofturbulent flow in a rectangular curved diffuser (AR= 2, Δβ = 180o, 2θ = 9.12º, and AS = 2) were carriedout by Djebedjian, 2001, [11]. The test was carriedout at Re = 4.59x105, based on the inlet hydraulicdiameter of 0.0666m and mass averaged inletvelocity of 33.7m/s. The maximum velocity coreshifted gradually from the convex wall towards theconcave wall as the flow moved downstream. Thestream-wise velocity along the convex wall wasobserved to decrease rapidly up to the turn of 90º,approached to zero at 90º and in the downstreamtheflow was separated with expanding recirculationtowards the exit. Wall static pressure recovery along
  3. 3. Nirmal K. Das, B. Halder, P. Ray, B. Majumdar / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.711-716713 | P a g econvex wall was significantly more than that onconcave wall and the overall pressure recovery wasonly 27%. Time-averaged Navier-Stokes equationswere used for numerical analysis. The standard k-εturbulence model and a non-equilibrium k-ε modelwere tested and evaluated against the experimentaldata. Both the turbulence models could not predictat par with the experimental results, the separationposition and the reattachment length.Using the commercial CFD code FLUENT,Dey et al., 2002, [12], studied numerically the effectof area ratio and turning angle on the performanceof circular cross-section S-diffusers. Investigationwas done with steady incompressible flow throughdiffusers, keeping the centre line length and inletvelocity (40m/s) fixed. Computations were madewith different values of area ratios of 1.5, 2.0, 2.5,and 3.0 and different turning angles of 15°/15°,22.5°/22.5°, and 30°/30°. Counter rotating vorticeswere observed to be present throughout the diffuserlength with increasing magnitudes at higher angle ofturn. Influence of turning angle on pressure recoverywas very less, only 1% fall in static pressurerecovery was predicted with the change in turningangle from 15°/15° to 30°/30°. Increase in area ratioregistered strong increase in static pressure recoverywith reduction in the magnitude of the cross-velocity. They predicted higher number of pairs ofvortices to generate for an area ratio of 3.Sedlář, and Příhoda, 2007, [13], modelednumerically the flow phenomena occurring inturbulent flows through rectangular curved diffusersof AR = 1.5, Δβ = 90o, with straight inlet and exitparts. To find the relationship of the hydraulic losseswith the ratio of inner wall radius to the diffuserheight (Ri/Wi) and the Dean number (De),investigation was carried out at various inletReynolds numbers (1x105to 1x106) and differentRi/Wi (1 to 6) keeping the diffuser height constant.Authors used the ANSYS CFX package to solve thethree-dimensional Reynolds-averaged Navier-Stokes equation together with different suitableturbulence models. Regions of flow separation weremarked and were found to be dependent both ondiffuser geometry and Reynolds number. Secondaryflows of different structures of multiple vorticeswere found to form which were influenced mainlyby the nature of flow separation.2. EXPERIMENTAL ANDCOMPUTATIONAL FACILILITYThe experiments were conducted in theAerodynamics Laboratory, National Institute ofTechnology, Durgapur, India. The measurementswere carried out in an open-circuit wind tunnel(Fig.1). The rate of mass flow of air through the testrig was controlled by throttling the blower suctionwith the help of a flap gate mounted on it. The90ºcurved diffuser was designed with circular arccentre line, as suggested by Fox and Kline [4]. Therectangular cross-section was increased in area bylinearly varying the width from 0.05m at inlet to0.1m at the exit, over the total centre line length of0.6m while the height was kept constant at 0.1m.The curved part of the test diffuser was fabricated infour segments each of 0.382m centre line radius andsubtending an angle of 22.5º at the centre ofcurvature. Two constant area straight ducts, onepreceded while the other succeeded the curved partof the diffuser. The test diffuser was fabricated fromtransparent acrylic sheets. One hundred eightyevenly distributed wall pressure taps were provided(at six measuring sections as elaborated later) formeasurement of wall pressures on the four walls.The mean velocities, static and total pressures at sixcross-sectional planes, namely Inlet, A, B, C, D, andOutlet were obtained using a pre-calibrated three-hole pressure probe in null mode. The geometry ofthe curved diffuser along with the measuringsections is shown in Fig.2.The geometry of the diffusing C-duct wasdeveloped in GAMBIT and subsequently it wasmeshed with the solver FLUENT 5/6. Flow throughthe diffuser was numerically simulated by usingdifferent viscous models available in thecommercial CFD software FLUENT 6.3. Theexperimental inlet and boundary conditions wereused as input for the validation of the viscousmodels. The 2-D presentations of the computed wallpressure distribution and mean velocity contourswere developed by using the graphics softwareTecplot’10.3. RESULT AND DISCUSSIONThe flow development in the diffusing C-duct was investigated through measurements of thewall pressure distribution, velocity distribution,pressure distribution and pressure recovery.However, due to the space restriction, the importantparameters like the wall pressure distribution, meanvelocity distribution, and the pressure recovery arepresented here. For validation of different viscousmodels available in the commercial CFD software‘FLUENT 6.3’, the computed results werecompared with the experimental results.3.1 WALL PRESSURE DISTRIBUTIONThe measured static pressures on therespective walls were normalized by inlet dynamicpressure and using the software ‘Surfer’, 2-Disobars were plotted as shown in Fig.3. Fig. 3(a) and(b), referring to top and bottom walls respectively,show the existence of higher pressure along theconcave edge with respect to lower pressure alongthe convex edge at any angle of turn. It reveals thatthe fluid movement has taken place from concaveedge towards the convex side. The pressuredistribution on the convex wall at any section up tothe 45º turn (Fig.3(c)), shows lower pressure at thecentre and higher towards the top andbottom which
  4. 4. Nirmal K. Das, B. Halder, P. Ray, B. Majumdar / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.711-716714 | P a g esignifies a movement of fluid towards the centrefrom top and bottom. A reverse trend is observed onthe concave wall, (Fig.3 (d)). These directions offluid movement across the flow passage indicate aprobable generation of secondary motion in the formof two contra-rotating vortices at least up to 45ºturn. The wall pressure distribution on the four wallsclearly exhibits a continuous rise in static pressurealong the primary direction of flow, as expected incase of a diffuser.3.2 MEAN VELOCITY DISTRIBUTIONThe normalised (by the inlet averagevelocity) velocity profiles at the six measuringsections are presented in Figs. 4 (a)–(f). The flow atthe diffuser entry (Fig. 4(a)) is almost uniform andsymmetric except in the vicinity of the top andbottom parallel walls, which appeared to be due toaccumulation of low momentum fluid over the wallsand the corner effect. Refer Fig. 4(b) for Section-A(angle of turn 11.25º), the flow is seen to accelerateon the convex wall side to balance the relativelyslower flow on the other i.e. concave wall. This isdue to the higher pressure on the concave wallcompared to the convex wall, as a result of strongcentrifugal force. At Section-B (angle of turn33.75º), referring to Fig. 4(c), non-uniformity in thevelocity profile at station-5 near the convex wall isobserved. This can be attributed to the migration oflow momentum fluid to the convex wall due togenerated secondary flow. Similar fluid movementwas seen from the wall pressure distribution. AtSection-C (angle of turn 56.25º), referring to Fig.4(d), the non-uniformity in the velocity profiles isseen to spread over more area on the convex wallside (station-6 and 7) as compared to the previoussection. It indicates migration of more lowmomentum fluid along the top and bottom wallstowards the convex wall. This may be attributed tothe combined effect of secondary flow and greateradverse pressure gradient on the convex wall. As theflow moves further down stream, at Section-D(angle of turn 78.75º), greater area on the convexwall side (up to station 5) is infected with lowmomentum fluid, as seen from the velocity profilesin Fig. 4(e), which may be attributed to the strongadverse pressure gradient on this wall. About themid horizontal plane, between station-7 and theconvex wall, the flow velocity may have beenreduced to zero or even the flow may have reversed,giving rise to flow separation on the convex wallover a small area. Due to reduction in velocity, asexpected due to gradual expansion of flow passagethe centrifugal force on the concave wall issignificantly reduced and fluid with higher velocitymoves along this wall. Though the non-uniformityin velocity profiles near the convex wall still persistsat the Outlet Section (Fig. 4(f)) like that in theprevious section, the waviness is reduced to someextent indicating a significant improvement inquality of flow (regarding uniformity). This may beattributedto the effect of flow through straight ductin this part. The overall reduction in flow velocityreaches its peak value at this section. The generalpattern of flow is seen to be symmetrical about themid horizontal plane.3.3 MEAN VELOCITY CONTOURSTo obtain more detailed information offlow development inside the diffuser, the normalised(by the inlet average velocity) velocity contours inthe form of 2-D presentation (Figs. 5(a)-(f)) havebeen drawn by using the graphics software packageSURFER. Fig. 5(a) referring to the Inlet Section,depicts uniform flow throughout the entire cross-section except at some portion close to the top andbottom corners. A similar observation was madefrom the velocity profiles also. At Section-A, fromFig. 5(b), the flow is seen to accelerate on theconvex wall side and move with maximum velocityoccupying major portion of the flow cross-sectionadjacent to the convex wall while the flow on theconcave wall is retarded. Similar differences invelocities on the opposite curved walls were alsodetected from the velocity profiles, the cause ofwhich was discussed in details in the previousarticle. The velocity contours in Fig. 5(c) clearlydepict that low momentum fluid has accumulated onthe convex wall and the bulk flow with maximumvelocity is pushed towards the opposite wall.Migration of this low momentum fluid affects theuniformity of flow along the convex wall whichcorroborates the earlier observation of the non-uniformity in the velocity profile near the convexwall (station 5) at Section-B. The bulk flow withmaximum velocity still occupies the area betweenthe mid vertical plane and the convex wall. Thedeveloped flow at this section is symmetric aboutthe mid horizontal plane. In further down stream atSection-C, refer Fig. 5(d), more and more lowmomentum fluid is seen to accumulate along theconvex wall which affects the uniformity of flowover the wall. Similar trend was observed in thewaviness of the velocity profiles at station 6 and 7on the convex wall side, as discussed in the previousarticle. Bulk flow with maximum velocity is seen toshift further towards the concave wall, compared tothat at the previous section. Flow on the convex wallis retarded more compared to that on the oppositewall; however, stagnation or reversed flow on anypart of the convex wall is not observed. At Section-D, Compared to the previous section, more area(about 40% of the total area) near the convex wall isoccupied by low momentum fluid and affects theuniformity of flow over there. Similar non-uniformity was reflected through the velocityprofiles, spread over large area between station 5and the convex wall. Bulk flow with maximumvelocity is further pushed towards the concave wall,and located almost centrally about the mid vertical
  5. 5. Nirmal K. Das, B. Halder, P. Ray, B. Majumdar / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.711-716715 | P a g eplane. Velocity gradient at the mid horizontal planebetween the curved walls indicates the possibilityofflow stagnation or even reversed flow on a finite(may be small) area on the convex wall as discussedthrough mean velocity distributionse. At OutletSection in Fig. 5(f), the velocity contours depictgreater area on the convex wall to be affected bylow momentum fluid and the core flow withmaximum velocity to be further shifted towards theconcave wall and aligned between the mid verticalplane and the concave wall. This is due to the inertiaof flow; the fluid tries to maintain its upstream flowdirection and moves closer to the concave wall. Thepattern of the velocity contours indicate a significantimprovement in quality of flow (regardinguniformity) compared to that in the previous sectionand justifies the installation of the straight duct atthe exit of flow.3.4 PRESSURE RECOVERY AND LOSSCOEFFICIENTThe variation of normalised (by the inletdynamic pressures) average static pressure recoveryand total pressure loss, based on average staticpressures at different sections and the centre linelength of the diffuser is presented graphically in theFig. 6. The static pressure in the direction of flow isobserved to increase continually, with maximumgrowth in between Section A and B. The totalpressure loss in the direction of flow is observed toincrease, with maximum increase in betweenSection A and B. The effectiveness of the diffuser,ξo, defined as the pressure recovery capacity of thediffuser in comparison with ideal pressure recoverycoefficient (75% for the present diffuser) is found tobe 70%.3.5 NUMERICAL VALIDATIONThe flow through the present diffusing C-duct was numerically simulated by using differentviscous models available in the commercial CFDsoftware FLUENT 6.3. The experimental inlet andboundary conditions were used as input for thecomputation and validation of the numerical results.Based on the comparison, it is found that Standardk–ε turbulence model provides better prediction offlow development in the present diffusing C-duct.However, a little deviation between thecomputational and experimental results may beobserved on the concave wall side, velocitydistributions obtained from computational analysisare in good agreement with the experimental results;both qualitatively and quantitatively. In Fig.7, thenormalised velocities obtained from computationalanalysis and experimental results are presented incontour form for validation. Numerically obtainedcoefficient of mass averaged static pressurerecovery (61.6%) and coefficient of mass averagedtotal pressure loss (9.1%) are in good agreementwith the experimentally obtained static pressurerecovery and total pressure loss coefficients of52.48% and 15.5 % respectively. For comparison ofthe computed results with the experimentallyobtained pressure recovery(Cpr) and loss (ξ)coefficients, the same are presented in Fig.8 andFig.9 respectively. These agreements corroboratethat the CFD code using Standard k-ε model canpredict the flow and performance characteristicsreasonably well for similar geometries with sameboundary conditions.4. CONCLUSIONSFlow development in a 90º curved C-shaped diffuser, of rectangular cross-section, lowaspect ratio and area ratio of 2, was experimentallyinvestigated in a turbulent flow regime (Re =2.35x105), based on the duct inlet hydraulicdiameter of 0.0666m and mass averaged inletvelocity of 60m/s. The following conclusions maybe drawn from the study.1. Down the stream, the wall static pressureincreased continuously on the four walls dueto diffusion.2. The duct curvature induced strong pressuredriven secondary motion, evolved into pair ofcounter-rotating vortices, ultimatelyconveyed down stream and broke into largepairs (at least two).3. In the down stream beyond 33.75º turn, thestreamwise bulk flow shifted graduallytowards the concave wall side under theinfluence of centrifugal force with a fasterdiffusion on the convex wall.4. Strong adverse pressure gradient developedon the convex wall, leading to probabledevelopment of flow stagnation or flowreversal on a small pocket on the convex wallnear the outlet.5. The induced secondary flows convected thelow momentum fluid of the boundary layerstowards the centre of the duct, degrading thequality of flow by disturbing the uniformityof velocity distribution.6. Coefficient of static pressure recovery of52.48º and effectiveness of the diffusercompared with an ideal one of 70 percent wasachieved.7. The CFD code using Standard k-ε model canpredict the flow and performancecharacteristics of C-diffusers in reasonableagreement with the experimental flow fieldfor similar geometries and with sameboundary conditions.REFERENCES[1] G.S. Williams, C.W. Hubbell, and G.H.Fenkell, Experiments at Detroit, Michiganon the Effect of Curvature upon the Flowof Water in Pipes, Transaction of ASCE,1902, vol. 47, pp. 1-196.
  6. 6. Nirmal K. Das, B. Halder, P. Ray, B. Majumdar / International Journal of EngineeringResearch and Applications (IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.711-716716 | P a g e[2] J. Eustice, Flow of Water in Curved Pipes,Proc. Royal Society London, 1910, SeriesA84, pp. 107-118.[3] J. Eustice, Experiments of StreamlineMotion in Curved Pipes, Proc. RoyalSociety London, 1911, Series A85, pp.119-131.[4] R. W. Fox, and S. J. Kline, Flow Regimesin Curved Subsonic Diffusers, Trans. ofASME, Journal of Basic Engineering, Sept.1962, pp. 303-316.[5] P. Bansod, and P. Bradshaw, The flow inS-shaped Ducts, Aeronautical Quarterly,May 1972, pp. 131-140.[6] M. B. Sajanikar, S Kar, and U. S. Powle,Experimental Investigation ofIncompressible Flow in Curved Diffusers,Proc. 11thNational Conference of FMFP,Hyderabad, India, 1982.[7] N. Hur, S. Thangam, and C. G. Speziale,Numerical Study of Turbulent SecondaryFlows in Curved Ducts, NASA ContractorReport 181830, ICASE Report No. 89-25,1989.[8] D. P Agrawal, and S. N Singh, FlowCharacteristics in a Large Area RatioCurved Diffuser, Proc. 18thNationalConference of FMFP, Indore, India,December 1991.[9] B. Majumdar, and D. P. Agrawal, FlowCharacteristics in a Large Area RatioCurved Diffuser, Proc. Instn. Mech.Engrrs., 1996, vol. 210, pp. 65-75.[10] B. Majumdar, R. Mohan, S. N. Singh, andD. P. Agrawal, Experimental Study ofFlow in a High Aspect Ratio 90 DegCurved Diffuser, Trans. of ASME, Journalof Fluids Engineering, 1998, vol. 120, pp.83-89.[11] B. Djebedjian, Numerical andExperimental Investigations of TurbulentFlow in 180º Curved Diffuser, Proc.FEDSM, ASME Fluids EngineeringDivision Summer Meeting, May 29-June 1,2001, New Orleans, Louisiana, pp. 1-9.[12] R. K. Dey, S. Bhavani, S. N. Singh, and V.Seshadri, Flow Analysis in S-shapedDiffusers with Circular Cross-section, TheArabian Journal of Science andEngineering, 2002, vol. 27, No. 2C, pp.197-206.[13] M. Sedlář, and J. Příhoda, Investigation ofFlow Phenomena in Curved Channels ofRectangular Cross-section, Journal ofEngineering Mechanics, 2007, vol. 14, No.6, pp. 387-397.