Experimental investigation on pump-intake-elbowsystems using refraction index matching andTR-SPIVSten Mittag*1, Martin Gabi1

SYM

POSI

A

ON ROTATING MACHIN

ERY

ISROMAC 2016

InternationalSymposium on

TransportPhenomena and

Dynamics ofRotating

Machinery

Hawaii, Honolulu

April 10-15, 2016

AbstractIntake elbow components to centrifugal pumps cause additional losses and secondary flows, hencenon-rotational velocity profiles at the pump intake. As a result the impeller vanes experience permanentchanges of the intake flow angle and with it transient flow conditions in the blade channels.

This paper presents the results of a project, experimentally investigating the consequences of non-rotational inflow to leading edge flow conditions of a centrifugal pump. Therefore two pump-intake-elbowsystems are compared, by only altering the intake elbow geometry: a common single bended 90◦ elbowand a numerically optimized elbow. In addition the same measurements are carried out at the referencesystem consisting of the same pump with straigth intake tube.

The experiments are performed, using time resolved stereoscopic PIV on a full acrylic pump withrefractions index matched (RIM) working fluid. This allows transient investigations of the flow fieldsimultaneously for all blade leading edges. Special focus is translated to the applied optical measurementsinside the operating impeller.

Keywordsnon-rotational symmetric inflow, centrifugal pump, intake elbow, PIV, RIM

1Institute of Fluid Machinery, Karlsruhe Institute of Technology, Karlsruhe, Germany*Corresponding author: mittag@kit.edu

INTRODUCTIONRequirements to small construction sizes most of the timeslead to short inflow pipes and consequently to distorted intakeprofile to turbo machines. Moreover for many industrial appli-cations the fluid needs to be transported continuously in onedirection, sometimes even in line. For centrifugal pumps thisleads to pump-elbow-systems, most of the times with compactintake elbow tubes upstream the pump.

Usually for pump-intake-elbow systems the demands forhigh efficiency and low cavitation affinity are tried to be ful-filled by optimizing each system component (intake elbowand pump) separately. Common elbow geometries are theresult of an optimization regarding low friction coefficients,to increase the elbow efficiency and decrease the risk of cavi-tation for downstream components.

Simply summing up the component efficiencies of thepump-intake-elbow system and neglecting any interaction,would lead to good system efficiencies.

The design for efficient impellers however, assumes arotational symmetric inlet velocity profile which cannot bedelivered by common and/or friction coefficient optimizedintake elbows. Consequently for a realistic system considera-tion additional losses have to be taken into account.

A 90◦ bended tube rather delivers a shifted, not rotationalsymmetric main flow profile and generates Dean vortex likestructures as secondary flow at the elbow outlet [1].

Many early investigations on the downstream velocity pro-files of single bended tubes have been carried out, e.g. byNippert [2] and Weske [3]. They show dependencies of theelbow flow phenomena mainly on the Reynolds number andthe bend characterizing ration δE = relbow/D−1 (radius ofthe elbow curvature divided by elbow tube diameter).

Studies on the interaction between upstream elbow com-ponent and centrifugal pump are comparatively more complexto execute, due to the difficulty of determining the distortedflow profile between elbow and pump in an assembled setup.

Siekmann and Schroter [4] investigated the effects ofdistorted inflow conditions to characteristics on a diagonalcentrifugal pumps impeller (nq = 100 [min−1]), includingswirl and deformations at the inflow and/or partial blockings.Moreover the compared the impacts of a common elbow to astraight intake tube. The velocity profiles between pump andupstream component were determined using LDV. They couldshow that the efficiency is mainly affected by the meridianmain flow profile or an artificial counter swirl. Moreover theauthors concluded that distortions at the intake profile lead toadditional shock losses.

Thamsen [5] carried out experiments on radial, diagonaland axial deep well pumps (nq = 40− 215 [min−1]) with in-take disturbances. The inflow characteristics were determinedby up scaled Reynolds similar models in air. He asserts lossesin pressure head and efficiency and increased fluctuations inthe power consumption. Furthermore Thamsen noticed a shift

Experimental investigation on pump-intake-elbow systems using refraction index matching and TR-SPIV — 2/9

List of SymbolsA [m2] Area λ [−] Power numbera, r, , ϕ [−] Cylindircal coord. (axial, radial circumf.) λPIV [nm] Wavelengh PIV laser light 532nmc [m/s] Absolut velocity ν [m2/s] Kinematic viscosityD [m] Diameter ρ [kg/m3] Densityl [m] Length τP [s] Particle relaxation timen [s−1] Rotational speed φ [−] Flow numberni [−] Refraction index ψ [−] Pressure numbernq [min−1] Specific speed ω [s−1] Angular velocity (ω = 2πn)rElbow [m] Radius elbow curvature ζ [−] friction factor (ζ = ∆ptot/(ρ/2 · c2m))P [−] Pressure measurement position∆p [Pa] Pressure head IndiciesT [Nm] Torque -1...3 Control section definitionu [m/s] Impeller circumferential speed A, B, C System definitionV [m3/s] Flow rate BP Best point (point of largest ηw [m/s] Relative velocity F Fluidy+ [−] Dimensionless wall distance Imp Impeller

P Seeding particleGreek letters ROI Region of interest

β [deg] Flow or blade design angle Sys SystemδE [−] Elbow radius to diameter ratio tot Totalεc [deg] Crit. incidence angle for tot. reflection vec PIV post processed vectorη [−] Efficiency

of the best points to lower volume flows and, like Siekmannand Schroter, an increased particularly local risk of cavita-tion.

Jager [6] investigated radial pumps , regular and withoutshroud (nq = 20− 60 [min−1]), with non-rotational symmet-ric meridian inflow and suppressed secondary flow swirl. Theintake profile was measured by LDV. Like the other authorsbefore him, he showed an increased cavitation risk, a reduc-tion of the pressure head and loss in efficiency, especially forhigher nq values.

The same results of deterioration of the pump characteris-tic, Roth-Kliem [7] examined in her study on a pump-intake-elbow system with different distances between elbow andimpeller as well the combination with a butterfly valve at theelbow intake.

Summarized the consequences of a non-rotational sym-metric intake velocity profile to centrifugal pumps are ad-ditional losses, especially shock losses at the leading edges.Moreover the distorted inflow conditions lead to an increasedrisk of cavitation, which is not considered in the NPSH value.

This paper presents the results of a project, investigat-ing the consequences of non-rotational symmetric inflow toleading edge flow conditions of a centrifugal pump. Thereforetwo pump-intake-elbow systems are compared experimentally,by only altering the intake elbow geometry: a common singlebended 90◦ elbow and a numerically optimized elbow.

The elbow optimization is performed regarding a rota-tional symmetric main flow velocity profile without secondaryflows at the outlet and a low friction coefficient, to improvedsystem efficiency.

Furthermore the impeller characteristics in the elbow sys-tems are compared to an impeller in an ideal system with astraight intake tube.

1. METHODSThe results aim to visualize and quantify the flow at all lead-ing edges and inlet area of all blade channels simultaneously,while determining the pump and system characteristics. There-fore time resolved stereoscopic particle image velocimetry(TR-PIV) combined with refraction index matching (RIM)techniques are use, to sufficiently resolve appearing flow phe-nomena in time and space.

1.1 Experimental SetupThe experimental setup is built as a common cycle test rig forfluid machines and illustrated in figure 1.

DUT

1

2 34

5

67

8

910

11

P−1

P0

P3

P4

K1

K2

Tn

1 Booster pump2 Straighteners3 Pre measurement section4 Intake elbow5 Investigation pump6 Reservoir7 Throttle8 Filter bypass

9 Heat exchanger10 Venturi11 Fluid supplyPi Pressure sensorKi Temperature sensorT Shaft Torquen Rotational speed

Figure 1. Test rig cylce

An intake tube with straighteners and a length of 40 ·D−1

(compare figure 10, 2→3) ensures a fully developed turbulenttube profile to the elbow. Moreover this tube includes an

Experimental investigation on pump-intake-elbow systems using refraction index matching and TR-SPIV — 3/9

acrylic section for optical access, to check this profile at theelbow inlet and deliver boundary conditions for the numericalmodels.

The investigated system contains of exchangeable elbows,invariable impeller and pump casing. Since the focus is trans-lated to the effects of distorted inflow conditions, any non-rotations interactions from downstream should be avoided.Consequently a simple cylinder with a diameter ofD3 = 3·D2

is chosen as impeller casing. A detailed sketch of the deviceunder test (DUT) including all pressure measuring points isshown in figure 2.

P−1

P0z/D2 = 0, 57

P3

P0.1

P0.4

P0.3

P0.2

inside outsie

60◦

12◦

P3.1P3.3

P3.5

P0

P3

zx

yx

yx

V

V

Figure 2. Device under test

The radial impeller is designed to Gulich [8] with fivecylindrical, two circle arch blades. The specific rotationalspeed is nq = 79 [min−1] and the specific diameterDq = 1.4[−]. The design is used to achieve a radial impeller withproportionally large width b2 to allow measurement planes indifferent levels parallel to the hub. The impeller’s meridiangeometry is illustrated in figure 3.

z

xM4 : z/D2 = −0, 025M3 : z/D2 = −0, 05M2 : z/D2 = −0, 075M1 : z/D2 = −0, 1

M0 : z/D2 = −0, 36

Figure 3. Scaled impeller geometrie including measurmentplanes

1.2 System DeclarationsFor the continuation of the paper the following definitions arechosen:

Table 1. System setup declaration

System A Common intake elbow upstream the pumpSystem B By optimization improved intake elbow

upstream the pumpSystem C Straight intake tube upstream the pump-1 Elbow intake / DUT inlet0 Intake profile to impeller1 Blade leading edge2 Impeller outlet3 Casing outlet / DUT outlet

1.3 Measurement TechnicsAs mentioned before, TR-PIV is used to detect the vecocitycomponents in all three directions. The measurement planesparallel to the hub resolve an area of AROI = 0.45D2 ·0.45D2 (ROI=Region of Interest) and containing around84 ·62 vectors and at a temporal resolution of fvec = 2 [kHz].Four planes are positioned equally spaced inside the bladechannel (see figure 3, M1−4) and one measuring the intakeprofile to the impeller respectivly outlet profile of the elbow(M0). An addtional meridian planes is installed along theelbow meridian orientation and resolved equally (see figure4).

hub parallel meridianD2

D1

x

y

D2

D0

D1

xz

Figure 4. ROI for the measurement planes parallel to thehub and meridian

The entire DUT including camera point of view perspec-tives is shown in figure 5. All parts of the DUT are custommanufactured of acrylic glass (PMMA).

zx

yx

Laser

Meridian view Top view

Scheimpflug

Cameras

Figure 5. Measurement setup - hub parallel planes

Experimental investigation on pump-intake-elbow systems using refraction index matching and TR-SPIV — 4/9

Meridian view

zx

yx

Top view

Figure 6. Measurement setup - meridian plane

To realize the almost distortion free complex optical ac-cess to the measurement planes (cf. figures 3, 5), RIM fluid toPMMA is applied. Using water or air as working fluid leadsto larger areas of total reflection (critical incidence angle fortotal reflection e.g. for water-acrylic glass: εc = 60.6◦) andrefraction while observing the targeted RIO. Moreover thedistortions would need to be dewarped for each record sepa-rately since the accurate positions of the blades can hardly bepredicted for transient measurements.

In literature many detailed studies for different types ofrefraction matching fluids for acrylic glass can be found (e.g.[9], [10]). The ideal solution would be to achieve the refrac-tion index of acrylic glass and obtain the fluid properties ofwater as well. However the necessary compositions of fluidsfulfilling these requirements are difficult to handle (hazardousand/or unstable in case of contact with air and/or light, e.g.NaI, ZnI, NaSCN, NH4SCN or KSCN), in particular for thelarger experimental setup like presented in this paper (fluidcapacity around 130 liter).

The RIM for this paper is implemented by using the min-eral oil Shell Gravex 917. Its optical and rheological prop-erties are determined in pre studies and summarized in table2.

Table 2. RIM fluid Gravex 917 probe properties at 20.5◦Cand a wavelength of λPIV = 532 [nm]

Refraction index deviation to acrylic glass ∆ni 0.2%Critical incidence angle for total reflection (avoided) 87 degMaximal distortion angle (dewarped) 3 degDensity deviation ρOil/ρWater 0.89Kinematic viscosity deviation νOil/νWater 30

By controlling the fluid temperature (see heat exchangerand K1, 2 in figure 1) and respecting the critical incidenceangle εc at the interface between acrylic glass and oil in theexperimental setup design (see figure 2) an almost distortion-free map (except the perspective distortions due to the stereoPIV setup) can be achieved (see figure 7, the showed oil is afirst test oil with a refraction index deviation of ∆ni ∼ 1% ata temperature of 24◦C).

in air in water in RIM oil at 24◦C

zx

targetb2/2

zx

D2 D2

b2

viewing perspective

Figure 7. Distorted image of a control plane in air, waterand in RIM test oil

For the experimental setup design pre studies are carriedout to identify the regions of critial incidence angles ε at thefluid solid interface (ε > εc). An excerpt of the analysis andits resulting design adaption is shown in figure 8

Geometry adaptations:

Increased shaft diameter

Reduction of radiusbetween shaft and hub

2 1 0

Distortion Error [Pixel]

Figure 8. Adaptation of the impeller geometry to minimizedistorion and total reflection

The drawback of using this non-harmful, stable and com-parably cost efficient oil, is the decrease of the Reynoldsnumber due to the higher kinematic viscosity νOil:

Re =D2 · u2νOil

≈ 16000 [−] (1)

De =cm,(−1)

νOil

√D3

−1

relbow[−] (2)

For seeding silver hollow coated glass spheres are used(dP = 15µm, ρP = 1.6 g/cm3). Here the relaxation timeτP (particle tendency to attain velocity of the fluid) improvesby applying the RIM oil’s higher viscosity [11].

τP =ρP

νFluid · ρFluid[s] (3)

Experimental investigation on pump-intake-elbow systems using refraction index matching and TR-SPIV — 5/9

1.4 Intake ElbowThe common elbow and finally selected geometry of the bestoptimization individual are shown in a meridian section infigure 9. For this paper the optimization will only be summa-rized.

The elbow is optimized in steady mode and detached fromthe pump elbow system. However its goal functions are for-mulated regarding systems efficiency:

Table 3. Goal functions for the elbow optimization

Goal function 1 Pressure loss in elbowGoal function 2 Rotational symmetric outflow velocity

profile with no secondary flow

Since this paper is only focusing on the flow phenomenaand their impact, cavitation effects and corresponding goalfunctions are not regarded.

In a second step additional transfer functions, which areadjusted to each optimization run, are applied to derive theexpected system efficiency ηSys from the goal funcfunctionvalues. These transfer function are developed from selectedelbow individuals, which are simulated fully transient andassembled in the entire pump elbow system. The selectiondemand for these elbows is a complete and preferably uniformdistribution of their goal function values in the sample spaceof the optimization. More details about the optimization canbe found in previous publications [12, 13].

D−1

D0

z

x

V

rElbow

Elbow A

Elbow BQ1

Q2 Q3

Q4

Q5

Figure 9. Measurement setup

2. RESULTS AND DISCUSSION

All shown results, including the optimization, are producedat the flow rate of the best point for the impeller in system A:φBP, A = 0.11.Moreover this chapter refers to the previus published paper ofthis work [14].

2.1 Characteristic CurveAs a consequence of the distorted intake profiles the character-istic curves shown in figure 10 result for system and impeller.

Hereby the following non dimensional definitions are use:

φ =V

π2

4 D32n

ψ =∆ptot

ρπ2

2 D22n

2

λ =Tω

ρπ4

8 D52n

3η =

∆ptotV

Tω

(4)

0.05 0.1 0.150

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0

0.05

0.1

0.05 0.1 0.15

0 0.05 0.1 0.150

10

20

30

40

50

60

φ

ψ

λ

η

Operation Point (PIV)

ElbowOptimization

ImpellerSystem

System

Impeller

CBA

ImpellerSystem

Figure 10. Characteristic curves for system and impeller

The slightly lower pressure head ψ for system B is ex-plained by the larger losses due to the more complex elbowgeometry. However concering the pressure head of the im-pellers only, system B is better than system A. The greatestdifference is detected for the power consumption λ. Here animprovement of around 10% is measured. Finally the effiencyη shown an improvement for system B and impeller B, whilethe system mainly benefits of the improved impeller charac-teristics. Furthermore needs to be mentioned that best point isslightly shifted (figure 10, black circles). The improvementcan still be found comparing the operation points at constantflow number φ or pressure number ψ.Overall the more uniform intake profile of system B achivesaround 50% of the possible improvement potitial by compar-ing to the configuration with perfect intake conditions (C).

Experimental investigation on pump-intake-elbow systems using refraction index matching and TR-SPIV — 6/9

y/D0

x/D0

cmcm,0

System CSystem BSystem A

0 0.4-0.4

0

0.4

-0.4

0

0.5

1.0

1.71.5

0 0.4-0.4 0 0.4-0.4

Figure 11. Flow profiles at the impeller intake (M0)

2.2 Intake Profiles at M0

Figure 11 shows the main flow velocity profiles in measure-ment plane M0. As discussed in the previous publication [14]the secondary flow is a order of magnitude smaller than themain flow velocity and has less impact on the flow behaviorin the impeller.

The elbow component losses are characterized by thefriction factor ζ:

ζElbow, A = 0.5 [−] ζElbow, B = 0.67 [−] (5)

As expected, for the common elbow (system A) a crescent-shaped maximum of the flow rate forms at its outside cur-vature, corresponding to elbow orientation. For system Bthe rotational symmetry is considerably improved, but notcompletely accomplished. Pre studies have shown that it ispossible to further improve this rotational symmetry, howeverat the expense of the elbow’s pressure loss hence the system’spressure head and efficiency.

As described, the discussed elbow geometry considersthe demand for a low pressure loss as well. Yet, due to themore complex geometry of the improved elbow (additionalcurvatures, increase of the elbow length and surface) morelosses are cause.

2.3 PIV ResultsFigure 12 shows the PIV results of the normalized volumeflow distribution cm inside the impeller (measurement planesM1, M3 and M4, see figure 3).

The crescent-shaped of the intake profile (see figure 11)is transported recognizably in the blade channels for systemA. Significant partial and over load regions develop, depen-dent on the elbow orientation. In the measure plane M4 thevelocities are reduced due to the expected stagnation point atthe hub. Here new secondary flow structures appear, tryingto compensate the local partial and over load regimes in theimpeller.

Furthermore the flow almost separates at the over loadshroud side, visualized in plane M1. Here the flow is notcapable to follow the shroud radius and starts separating, dueto increased mass flow rate and consequently larger axial ve-locities at the elbow outside curvature.

Overall the same effects can be detected for the system Bwith improved elbow, but in considerably declined existencealso compared the measurements for system C.

For system A and B the distibution of cm seems symmtricto the y/D2 = 0 plane so also symmetric the elbow orienta-tion. Further investigations need to be carried out, to determinif the relatively small deflection of the choosen impeller bladedesign affects that the evolving structures are not transprotedfurther with the impeller rotation but symmetric to y/D2 = 0.

Since the plane is orientated along the elbow meridian cutas well, the local partial and overload regions are visualizedclearly. Moreover the plotted streamlines show secondaryflow like behavior close to the hub, trying to compensate par-tial and overload region, especially for system A.

To receive more detailed information on the evolving sec-ondary/compensating flow structures close to the hub thestreamlines of the relative velocity are plotted in figure 14.The visualizations show an eccentric stagnation point. Analogto the already described phenomena, here the asymmetry isamplified developed for system A as well. The shift of theflow rate maximum to the outer curvature seems to interactwith the rotation of the impeller and finally results in the posi-tion of the stagnation point.Again the shift of the stagnation point is distinct for system Aand reduced for system B.

The result of the compensating secondary flow and/or theshift of the stagnation point are significant deviations of theintake flow angle consequently strong incidence angle ∆β1 atthe leading edge.

∆β1 = βBlade, 1 − βFluid, 1 (6)

Here figure 15 shows the incidence angle ∆β1 for the mea-surement plane M4 close to the hub. Phase averaging is usedto correct the results of fluctuations due to impeller rotation.The appearing incidence angles reach up to ∆β1 > 50 deg.This significantly will affect the impeller torque (compareEuler equation [14]), hence the system efficiency η, as thecharacteristic curves show (see figure 10).

Experimental investigation on pump-intake-elbow systems using refraction index matching and TR-SPIV — 7/9

cmcm0

0

1.5

1.0

0.5

0

0.4

-0.4

0

0.4

-0.4

Syst

emB

Syst

emA

M1 M3 M4

D1

D0

D2

y/D2

0 0.4-0.40 0.4-0.4 0 0.4-0.4

x/D2

0

0.4

-0.4

Syst

emC

Figure 12. Measurement results of the distribution of the flow rate inside the impeller passage

0 0.4-0.4

0

-0.2

0 0.4-0.4

0 0.5 1.0 1.5 1.7cmcm,0

z/D2

x/D2

System A System B

Figure 13. Measurement results of the distribution of the flow rate inside the impeller passage

Experimental investigation on pump-intake-elbow systems using refraction index matching and TR-SPIV — 8/9

0 0.4-0.4

0.4

-0.4

0

M4

ωy/D2

x/D2

V

Elbow

System ASystem BSystem C

Figure 14. Streamlines of relative velocity close to the hub(M4)

-40

-20

0

20

40

-1 -0.5 0 0.5 1ϕ[π]

∆β1

stagnationpointover load over load

y/D2

x/D2

0ϕ

0.4

-0.4

0

0 1.71.51.00.5cm/cm

0 0.4-0.4 0.40-0.4

partial load

System ASystem B

E4

Figure 15. Phase averaged incidence angle ∆β1 resultingof the evolving secondary flow at M4

3. CONCLUSIONBy using PIV combined with RIM the impact of non-rotationalsymmetric intake profiles to the flow in an impeller of pump-intake-elbow systems could be visualized and quantified. Sys-tem A, consisting of a common intake elbow and pump, isexperimentally compared to system B with improved intake el-bow. The elbow improvement is the result of an optimizationregarding pump-intake-elbow system efficiency by improvingthe rotational symmetry of the elbow outflow and respectingthe pressure loss of the elbow simultaneously. Moreover bothsystems A and B are compared the ideal system C representedby the pumpe with straight intake tube. All presented resultsrefer the flow at the impeller best point in system A.

As working fluid the haszard-free, comparably well-pricedmineral oil Shell Gravex 917 is chosen. The unchanged probehits the refraction index of acrylic glass by only a small de-viation of ∆ni = 0.2% at a temperature of 20.5◦C (water vs.acrylic glass→ ∆ni = 12.5%) and is almost transparent. Bythis incidence angles of an optical path to the refracting edgeup to ε = 87◦ can be achieved (at a maximum distorsion of3◦). For most applications it should be possible to adapt thegeometry, to avoid the critical incidence angles of εc > 87◦.One drawback is higher viscosity, hence Reynolds numbercompared to water (factor 30).

With this method the flow phenomena and their relationto the intake profiles are shown and compared in differentplanes, parallel to the hub and meridian, inside the impeller.Local partial and over load regions depending on the elbowgeometry and its orientation are visualized and quantified.Furthermore appearing secondary flows close to hub, trying tocompensate the local partial and over load, are discussed. Inparticular the incidence angles ∆β1 between distorted inflowand blade design at the leading edges are regarded in detail.

Even though the more complex geometry of the elbow Bcauses as expected more losses (→ ζElbow, A > ζElbow, B ,see eq. 5) and slightly decreases the pressure head of thesystem B, the efficiency of the system is increased since thepower consumption λ and efficiency η of the impeller areimproved by the more uniform intake profile.

It is known from literature that increased power consump-tions λ, like for system A, could be connected to the increasedshock losses. This is visualized by the extracted incidenceangle ∆β1.

By comparing both system A and B to a defined idealsetup with straight intake to the impeller, system B shown animprovement in the impeller efficiency of more than 50% tothe overall improvment potential (A vs. C).

For practical applications it is shown that for pump intakeelbow systems, not only the pressure drop of elbow needsto be considered, but the entire system including the compo-nents interaction, in this case the distorted intake profile to theimpeller, should be taken into account.

Experimental investigation on pump-intake-elbow systems using refraction index matching and TR-SPIV — 9/9

REFERENCES[1] Dean, W. R., and Hurst, J. M., Note on the motion of fluid

in a curved pipe, The London, Edinburgh, and DublinPhilosophical Magazine and Journal of Science, Series 4,1927.

[2] Nippert, H., Über den Strömungsverlust in gekrümmtenKanälen, VDI-Verlag, Heft 320,1929.

[3] Weske, J. R., Experimental investigation of velocity dis-tributions downstream of single duct bends, NACA, Nr.1471, 1948.

[4] Siekmann, H. E., and Schröter, R., Auswirkung un-gleichmäßiger Zuströmung auf spezi-fisch schnelläufigeKreiselpumpen der Wasserversorgung und Entsorgung,VDMA Pumpentagung Karlsruhe, Germany, 1984.

[5] Thamsen, P. U., Experimentelle Untersuchung über dieAuswirkungen von Zuströmstörungen auf das Betriebsver-halten von Tiefbrunnenpumpen, PhD thesis, TU Braun-schweig, Germany, 1992.

[6] Jäger, R., Experimentelle Untersuchungen zum Ein-fluß von Zulaufstörungen auf das Betriebsverhalten vonKreiselpumpen einfacher Geometrie, PhD thesis, TUDarmstadt, Germany, 1998.

[7] Roth-Kliem, M., Experimentelle und nu-merische Unter-suchung zum Einfluss von Einbaubedingungen auf dasBetriebs- und Kavitationsverhalten von Kreiselpumpen,PhD thesis, TU Darmstadt, Germany, 2006.

[8] Gülich, J. F., Kreiselpumpen - Handbuch für Entwicklung,Anlagenplanung und Betrieb. Springer, 2010.

[9] Wiederseiner, S., Andreini, N., Epely-Chauvin, G., andAncey, C., Refractive-index and density matching in con-centrated particle sus-pensions: a review, Experiments inFluids, Vol. 50 pp. 1183–1206, 2011.

[10] Uzol, O., Chow, Y.-C., Katz, J., and Meneveau, C., Unob-structed particle image velocimetry measurements withinan axial turbo-pump using liquid and blades with matchedrefractive indices, Experiments in Fluids, Vol. 33 pp.909–919, 2000.

[11] Raffel, M., Willert, C., Wereley, S., and Kompenhans. J.,Particle image velocimetry: a practical guide. Experimen-tal fluid mechanics, Springer, 2007.

[12] Sieber, S., Schänzle, C., Çaglar, S., and Gabi, M., Evolu-tionary optimization of a pump inlet elbow, InternationalRotating Equipment Conference, Düsseldorf, Germany,2015.

[13] Sieber, S., and Gabi, M., Optimization of pump intakeelbow systems, 15.th International Symposium on Trans-port Phenomena and Dynamics of Rotating Machinery,Honolulu, USA, 2014.

[14] Mittag, S. Gabi, M., Experimental and numerical inves-tigation on centrifugal pumps under non-rotational sym-metric inflow conditions12th International Symposium on

Experimental, 12.th Computational Aerothermodynamicsof Internal Flows, Lerici, Italy, 2015.

[15] KSB AG, Auslegung von Kreiselpumpen – Zähigkeitund Pumpenkennlinie – Korrekturfaktoren, p.51 KSB AG,1999.