Flow Visualization in a Centrifugal Blood Pump with an Eccentric Inlet Port

Artificial Organs

28(6):564–570, Blackwell Publishing, Inc.© 2004 International Center for Artificial Organs and Transplantation

564

Blackwell Science, LtdOxford, UKAORArtificial Organs0160-564X2004 International Society for Artificial Organs286564570Original Article

CENTRIFUGAL PUMP FLOW WITH ECCENTRIC INLETT. YAMANE Et al.

Received May 2003; revised November 2003.Address correspondence and reprint requests to Dr. Takashi

Yamane, Deputy Director, Institute of Human Science and Bio-medical Engineering, National Institute of Advanced IndustrialScience and Technology, Namiki 1-2, Tsukuba, Ibaraki, 305-8564,Japan. E-mail: [email protected]

Flow Visualization in a Centrifugal Blood Pump with an Eccentric Inlet Port

*Takashi Yamane, †Takayuki Kodama, ‡Yoshiro Yamamoto, ‡Toshiyuki Shinohara, and §Yukihiko Nosé

*National Institute of Advanced Industrial Science and Technology, Tsukuba, Ibaraki; †Tokyo University of Science, Noda,

Chiba; ‡Miwatec Co. Ltd, Kawasaki, Kanagawa, Japan; and §Baylor College of Medicine, Houston, TX, U.S.A.

Abstract:

Flow visualization analysis was applied to theBaylor/Miwatec centrifugal artificial heart to evaluate itsfluid dynamic characteristics regarding antithrombogenic-ity. An eccentric vortex was found both in the upper andthe lower gaps of the impeller, which is supposed to becaused by the eccentric inlet port. Therefore, one-way flowtoward the outlet is formed and washes the pivot. Thecombination of an eccentric vortex and a pivot bearingthat is washed is unique to the Baylor/Miwatec pump. For

the male pivots exposed to periodic wash, the minimumshear rate around the bottom pivot was estimated to be650/s, which is higher than the threshold for thrombusformation shown by other studies. The wall shear rate atthe impeller bottom surface was found to be larger in thetop contact mode than in the bottom contact mode.

KeyWords:

Centrifugal pump—Flow visualization—Eccentricinlet port—Shear rate.

Nowadays, flow visualization measurements orcomputational fluid dynamic analyses are widelyapplied to the development of artificial hearts. Tovisualize the wall shear stresses, a paint erosionmethod was used for a diaphragm-type pulsatilepump (1) and an oil film method (2–4) and a quanti-tative oil streak method (5) were applied to centrif-ugal pumps. Although these methods are essentiallyqualitative, they provide useful information regard-ing thrombus formation because both methodsdirectly detect areas of stagnation where blood clot-ting seems to initiate. Only recently have precise flowvelocity measurements been performed for artificialhearts using particle image velocimetry (6–8). Thesedata can be compared with the results of computa-tional fluid dynamic analyses.

We have applied this flow visualization techniqueto a compact centrifugal blood pump developed foran implantable biventricular assist system (Fig. 1) by

Baylor College of Medicine/Miwatec Co., Ltd. undera Non-Energy and Industrial Technology Develop-ment Organization, Japan (NEDO) project (9,10). Aqualitative study using an oil film method to clarifythe wall shear stress on the impeller bottom surfaceand the opposing casing surface was conducted byIchikawa et al. to find the optimum secondary vaneconfiguration (2,3). The objective of the present arti-cle is fluid dynamic evaluation for antithrombogenic-ity of the centrifugal blood pump through flowvisualization analysis.

MATERIALS AND METHODS

The original Baylor/Miwatec centrifugal pump hassix semiopen vanes and the impeller diameter is50 mm. The pump impeller is supported by a doublepivot bearing. Because the top bearing is designed tobe supported by the casing surface, the inlet port islocated apart from the center and is inclined to forman “eccentric inlet port.” The top bearing makes con-tact when the rotational speed is sufficiently high andhydrodynamic force lifts up the impeller: this istermed “top contact mode.” The bottom bearing con-tacts when the rotational speed is not so high as tolift up the impeller against the magnetic couplingforce: this is termed “bottom contact mode.”

CENTRIFUGAL PUMP FLOW WITH ECCENTRIC INLET 565

Artif Organs, Vol. 28, No. 6, 2004

For the flow visualization experiments a 250%scale-up acrylic model was used (Fig. 2). A 64 wt%NaI solution with a matched refractive index (= 1.49)was used as working fluid to avoid image deforma-tion. SiO

2

beads (0.15 mm in diameter) with a specificgravity (= 1.9) to match the working fluid were used.Matching of specific gravity was verified by settingthe solution still on a desk to assure that the particlesdid not rise or fall for more than 10 min.

Strict application of the similarity law requires thatboth Reynolds number and specific speed should becoincident with each other between the acrylicmodel/working fluid and the actual device/blood.Reynolds number is defined as Re =

RU

/

n

(where

R

:impeller radius,

U

: impeller tip velocity,

n

: kinematicviscosity of the liquid), which is determined by theimpeller rotational speed. Specific speed is defined as

n

s

=

N Q

1/2

H

-

3/4

(where

N

: impeller rotational speed

FIG. 1.

Implantable biventricular assistpump developed by Baylor College ofMedicine/Miwatec Co., Ltd.

Double pivot bearing systemInlet port

Outlet port

Impeller magnet

Secondary vane

Alumina ceramic shaft

FIG. 2.

Apparatus and conditions for flowvisualization.

1400~2400 rpm ( 12/125 rotational speed)Rotational speed

4~6 L/min ( 3/2 flow rate) Flow rate

4 time-step particle tracking methodImage analysis

High speed video (4500 frame/s)Imaging

Ar-ion laser light sheet (4W)Illumination

SiO2 beads (f0.15 mm, specific G.: 1.9)Tracer particles

64wt% NaI solution (refractive I.: 1.49, specific G.: 1.9)Working fluid

250% scale-up acrylic model (refractive I.: 1.49)Scale-up model

Similarity law: Coincidence of Reynolds number and specific speed

Laser HS Video

Model

Ø

Ø

566 T. YAMANE ET AL.


[rpm],

Q

: flow rate [m

3

/min],

H

: total head of thepump [m]), which is mainly determined by circuitresistance. This can easily be verified by substitutingapproximate equations

H

=

R

2

w

2

/

g

and

r

gH

=

LQ

2

(

w

: impeller angular velocity,

L

: circuit resistance)into the above definition.

Because the size of the acrylic model is 250% ofthe actual pump and the kinematic viscosity of thefluid is 60% of blood, the rotational speed should bereduced to 9.6% of the actual size pump to match theReynolds number and the flow rate should beincreased to 150% to match the specific speed. Allthe values described in the present paper are for anactual (100%) size pump.

Video images were taken with a high speed videosystem at a recording rate of 4500 frame/s (PhotronFastcam-Ultima-UV, Tokyo, Japan) with the illumi-nation of a 4 W argon-ion laser light sheet (Lexel 95-7, Fremont, CA, U.S.A.) and analyzed with particleimage velocimetry software (“Current,” Kanomax,Osaka, Japan). The video images were converted toblack and white images first. Subsequently, the areaat the center of each particle image was calculated,and identification of each particle between differentframes was conducted by particle tracking velocime-try (PTV). Four-frame PTV (identifying each particle

FIG. 3.

Location of visualized section forside view around the bottom pivot(1800 rpm, 5 L/min).

LLLLaaaasssseeee rrrr

Camera

VisualizedArea

Impeller

PrimaryVane

Secondary Vane

Velocity [m/s]Black : 0~0.09Blue : 0.09~0.18Green : 0.18~0.26Red : 0.26~

Side view Planar view

Bottom casing

Impeller

Secondaryvane

Casing

FIG. 4.

Side view of the flow around the bottom pivot (1800 rpm,5 L/min).

Velocity [m/s] Black: 0~0.08 Blue: 0.08~0.15 Green: 0.15~0.23

Red: 0.23~

(a) Bottom contact mode

(b) Top contact mode

Casing

Casing

FIG. 5.

Location of visualized section for the planar view aroundthe bottom pivot (1800 rpm, 5 L/min).


VisualizedArea

Plane View

Laser

Impeller

Side view

Plane view

Rotatio

nal d

irecti

on

of im

pelle

r

Pivot

Rotation

Secondaryvane

Casing

Pivot

Inlet

0 1 2 3

0.8 (m/s)

4 (mm)

Motor

Inlet

Out

let

Outlet



by smooth trace for four frames) was used for in-plane flow and by spring model PTV (identifyingeach particle by minimum strain energy for deforma-tion between two frames) was used for out-of-planeflow or secondary flow because of short trackingtime. After obtaining velocity vectors, the shear ratewas calculated. The in-plane shear rate was calcu-lated as a velocity gradient (

∂

v

/

∂

x

+

∂

u

/

∂

y

).

RESULTS

The operating condition of the pump model wasset to a rotational speed of 173 rpm and flow rate of7.5 L/min, corresponding to the actual condition of1800 rpm and 5 L/min.

When we observed the flow around the bottompivot along a diameter, from a side as shown in Fig. 3,the spring model method was used because the mainflow is normal to the visualized plane. Then, a specialone-way flow toward the outlet was found in the gapbetween the impeller and the casing (Fig. 3) thoughcentrifugal pumps generally have a symmetric flowpattern such as an inward flow along the stationarysurface due to pressure gradient and an outward flowalong the rotating surface due to centrifugal force.Closer views of the flow around the pivot were furtherinvestigated for the bottom contact mode and for thetop contact mode (Fig. 4). In the bottom contactmode, flow approaches the pivot from the left andthen forms a stagnant area behind the pivot. In con-trast, flow, whose velocity is over 0.2 m/s, washes thetip of the pivot in the case of the top contact mode.

Then we observed the planar flow pattern aroundthe bottom pivot (Figs. 5 and 6). The laser light sheetwas set to illuminate a range from 0 to 0.6 mm (asthe

¥

1 model size) above the bottom surface of thecasing. The four-frame particle tracking method wasapplied for this case. Other than the standard condi-tion of 5 L/min and 1800 rpm, the flow rate was variedfor 4, 5, 6 L/min and the rotational speed for 1600,1800, 2200 rpm, as of actual size. Then, we found inall cases an eccentric vortex, near the casing wall,whose center had an offset of 5–6 mm in the directionalong the center line of the inlet and the outlet. Thisresult shows that the eccentric vortex induces one-

FIG. 6.

Planar view of flow around thebottom pivot (effect of specific speed).

1600 rpm-5 L/min 2200 rpm-5 L/min



Clockwise rotation

1800 rpm-5 L/min

0 1 2 3 4

0.8(m/s)

(mm)

0 1 2 3 4

0.8(m/s)

(mm)

0 1 2 3 4

0.8(m/s)

(mm)

0 1 2 3 4

0.8(m/s)

(mm)0 1 2 3 4

0.8(m/s)

(mm)

FIG. 7.

Location of visualized section for planar view of the uppergap.

Rotation


Rotational direction of impellerLaser

Impeller

Plane view

1800 rpm-5 L/min

(mm)43210

Center

PivotVisualized

Area

2.8 (m/s)

Casing



way flow along the diameter toward the outlet aroundthe pivot, as is shown in Fig. 3. The flow patterns fora higher rotational speed resembled those for a lowerflow rate. This is because the ratios of radial velocityto tangential velocity are close. The ratio can beregarded as the square of the specific speed as wasmentioned in materials and method above.

To clarify the cause of the eccentric vortex, weinvestigated the planar flow patterns in the uppergap, illuminating the plane close to the tip of thevanes. Then, a similar eccentric vortex was foundagain though the velocity is much higher (Figs. 7 and8). The flow patterns for a higher rotational speedresembled those for a lower flow rate similarly to theflow patterns in the lower gap. The vortex patternwas elliptic for 1600 rpm with a flow rate of 5 L/minand at 1800 rpm with a flow rate of 6 L/min while itwas circular for 1800 rpm with a flow rate of 4 L/minand for 2200 rpm with a flow rate of 5 L/min. Theflow patterns were highly influenced by the inflowfrom the eccentric inlet. This suggests that the eccen-tric vortex was caused by the eccentric inflow and thepressure field formed was transmitted behind theimpeller.

The flow around the bottom pivot was examinedintensively. Because the eccentric vortex inducesflow along the diameter toward the outlet, the rela-tive velocity on the rotating pivot surface is a trailingflow on the left side and an opposing flow on the rightside (Fig. 9). Though the impeller back surface isslightly conical, the illuminated range was 0–0.8 mmfrom the impeller bottom plane. A four time-frameparticle tracking method was applied and analyzed

to give relative velocities for 46-degrees of rotationto obtain an instantaneous image so as to capture therotating secondary vanes and a stationary eccentricvortex simultaneously. Only the obtained magnitudeof the relative velocity is shown in Fig. 9. The sharpvelocity boundary at the 6 o’clock direction is causedmainly by the lack of particle images resulting fromshading by the motor. It can be understood that theimpeller surface around the pivot is exposed toperiodically oscillating flow due to the eccentricvortex.

FIG. 9.

Relative velocity distribution on the impeller bottom sur-face (effect of rotational speed).

0.0

1.21 [m/s]

1600 rpm-5 L/min


Eccentric vortex

Rotation

Pivot

Visualized Area

FIG. 8.

Planar view of flow around thetop pivot (effect of specific speed).

1600 rpm-5 L/min

1800 rpm-4 L/min

2200 rpm-5 L/min

1800 rpm-6 L/min


1800 rpm-5 L/min

Clockwise rotation

2.8 (m/s)

2.8 (m/s)

2.8 (m/s)

2.8 (m/s)

2.8 (m/s)

(mm)

(mm)

(mm)

(mm)

(mm)



The wall shear rate for the impeller around thepivot was examined by measuring the velocity gradi-ent obtained from the velocity profile in the impeller/casing gap at 2.6 mm radius. In Fig. 10, the left end(

x

= 0) corresponds to the rotating impeller surfaceand the right end (

x

= 1) to the pivot tip. The velocitygradient was evaluated in the 0.1 mm layer from theimpeller surface. Then, we obtained wall shear ratesof 650/s for the bottom contact mode and 1000/s forthe top contact mode (Fig. 10). The wall shear rateswere obtained in a similar manner for different rota-tional speeds as in the bar graph (Fig. 11). Generally,the wall shear rate for the top contact mode is higherthan that of the bottom contact mode. Though theshear rate decreases linearly as the rotational speeddecreases for the bottom contact mode, it does notgo below a certain value for the top contact mode,which would be advantageous for prevention ofthrombus.

DISCUSSION

The existence of the eccentric vortex was previ-ously observed indirectly by the oil film method (2,3).It is also important that similar flow patterns with aneccentric vortex were obtained both in the upper andthe lower gaps of the impeller. This can be inter-preted as showing that the pressure field, whichforms in the upper gap due to the eccentric inflow,transmits to the lower gap and forms a similar eccen-tric vortex pattern as in the upper gap. That is, thecause of the eccentric vortex would be the pressuredistribution induced by the eccentric inflow.

Regarding the oscillating shear rate around thepivot, Hashimoto’s study (11) is interesting. Heexamined the clotting characteristics of canine bloodwith a cone-cup rheometer whose speed was oscillat-ing periodically. He clarified that a minimum shearrate of 100/s during oscillations corresponds to thethreshold for clotting. If this rule is applied to thepresent result, blood clotting is suggested to occur fora rotational speed less than 1200 rpm only for bottomcontact mode.

The reason why the top contact mode has a highershear rate can be explained as follows: Because themean tangential velocity for the top contact mode isapproximately 75% of that of the bottom contactmode, the velocity gradient near the rotating wallbecomes higher for the top contact mode. The reasonfor the difference of the mean velocity is not clear,but probably the center of the eccentric vortex comescloser in the top contact mode due to a narrow uppergap for the impeller.

Though we analyzed the shear distribution on thesurface of the rotating impeller, we did not examine

FIG. 10.

Evaluation of minimum wall shearrate.

(a) Bottom contact mode

(b) Top contact mode

Minimum wall shear rate:650/s

Minimum wall shear rate:1000/s

LLLLaaaasssseeee rrrr

ObservedArea at4% radius(Minimum ShearRate)

Vortex Center

1800 rpm, 5 L/min

1800 rpm, 5 L/min

FIG. 11.

Effect of rotational speed on minimum shear rate atimpeller bottom surface.



the surface of the stationary lower casing. This isbecause thrombus was reported in animal experi-ments only on the impeller surface and not on thestationary casing surface (10). Although Fig. 10shows that the shear rate on the impeller surface islower than that on the casing bottom surface near thepivot, further investigation would be needed toestablish the stagnation point on the casing bottom.

CONCLUSIONS

Flow visualization analysis was applied to a Bay-lor/Miwatec centrifugal artificial heart to evaluatethe fluid dynamic characteristics with regard to anti-thrombogenicity.

1

An eccentric vortex was found in the upper andthe lower gaps of the impeller due to the eccentricinlet port. Therefore, a one-way flow toward theoutlet is formed and washes the pivot. This isunique to the Baylor/Miwatec pump.

2

The eccentric vortex induces periodic washing ofthe pivot. The minimum shear rate around thepivot was estimated to be 650/s, which is higherthan the threshold for thrombus formation shownby other studies.

3

The wall shear rate at the impeller bottom surfaceis larger in the top contact mode than in the bot-tom contact mode.

REFERENCES

1. Nakata M, Masuzawa T, Tatsumi E, et al. Characterizationand optimization of the flow pattern inside a diaphragm bloodpump based on flow visualization techniques.

ASAIO J

1998;44:M714–8.2. Ichikawan S, Nonaka K, Linneweber J, et al. Flow visualiza-

tion study, investigating the critical point of thrombus forma-tion in a gyro centrifugal pump.

ASAIO J

2001;47:101.3. Ichikawan S, Nonaka K, Linneweber J, et al. Flow visualiza-

tion study to avoid thrombus formation in a gyro centrifugalpump.

Artif Organs

2001;25:796.4. Tsukiya T, Taenaka Y, Lee HS, Tatsumi E, Takano H. Visu-

alization study of the transient flow in the centrifugal bloodpump impeller.

ASAIO J

2001;47:117.5. Lemire PP, McDaniel JC, Wood HG, et al. The application of

quantitative oil streaking to the HeartQuest left ventricularassist device.

Artif Organs

2002;26:971–3.6. Asztalos B, Yamane T, Nishida M, Masuzawa T, Konishi

Y. Evaluation of shear and recirculation in centrifugalartificial heart by flow visualization.

J Visualization

2000;3:79–92.

7. Manning KB, Miller GE. Flow through an outlet cannula ofa rotary ventricular assist device.

Artif Organs

2002;26:714–23.

8. Day SW, McDaniel JC, Wood HG, et al. A prototypeHeartQuest ventricular assist device for particle image veloci-metry measurements.

Artif Organs

2002;26:1002–5.9. Yoshikawa M, Nonaka K, Linneweber J, et al. Development

of the NEDO implantable ventricular assist device with gyrocentrifugal pump.

Artif Organs

2000;24:459–67.10. Nonaka K, Linneweber J, Ichikawa S, et al. Development of

the Baylor gyro permanently implantable centrifugal bloodpump as a biventricular assist device.

Artif Organs

2001;25:675–82.

11. Hashimoto S. Clot growth under periodically fluctuating shearrate.

Biorheology

1994;31:521–32.

Documents

Flow Visualization in a Centrifugal Blood Pump with an Eccentric Inlet Port