flow visualization as a complementary tool to hemolysis testing in the development of centrifugal...

6
Flow Visualization as a Complementary Tool to Hemolysis Testing in the Development of Centrifugal Blood Pumps Takashi Yamane, Bala ´ zs Asztalos, Masahiro Nishida, *Toru Masuzawa, *Koki Takiura, *Yoshiyuki Taenaka, ²Yoshiaki Konishi, ²Yuki Miyazoe, and ²Kazuyuki Ito Mechanical Engineering Laboratory, Tsukuba, *National Cardiovascular Center, Osaka, and ²Nikkiso Co., Ltd., Tokyo, Japan Abstract: With a 250% scaled-up pump model, high speed video camera, and argon ion laser light sheet, flow patterns related to hemolysis were visualized and analyzed with 4 frame particle tracking software. Different flow patterns and shear distributions were clarified by flow visualization for pumps modified to have different hemolysis levels. A combination of in vitro hemolysis tests, flow visualization, and CFD analysis suggested a close relationship between hemolysis and high shear caused by small impeller/casing gaps. Because arbitrary cross sections can be illuminated by laser light sheet, flow visualization is a useful tool in finding locations related to hemolysis in the design process of rotary blood pumps. Key Words: Flow visualization— Rotary blood pumps—Hemolysis—mechanical trauma. During the development process of a centrifugal blood pump, large numbers of hemolysis tests are required even if one desires only small design im- provements. If the cause of hemolysis is essentially mechanical, flow visualization and computational fluid dynamics (CFD) analysis are useful in estimat- ing the test results and reducing the number of he- molysis tests. Therefore, a correlation study between hemolysis tests and flow visualization as well as CFD analysis was conducted to establish a design method using extracorporeal blood pumps. In terms of flow visualization, high shear can be regarded as one of the causes of hemolysis, and stag- nation or standing vortexes can be regarded as causes for thrombus formation. Previous studies in- dicate that hemolysis is caused by collision or inter- action of red blood cells (RBCs) with solid surfaces (1), by fatigue of RBC membranes in shear flow (2), or other mechanisms. Therefore, high shear and high speed regions need to be carefully investigated to predict locations of hemolysis. For locations that are inaccessible to flow visualization, CFD is useful. MATERIALS AND METHODS The visualization apparatus consisted of a 250% scaled-up pump model with a 6 vaned semiopen im- peller (Fig. 1, diameter: 125 mm, similar to HPM-15, (Nikkiso Co., Ltd., Tokyo, Japan), high speed video camera (4,500 frame/s, Photron, Tokyo, Japan), Ar- gon ion laser light sheet (7 W, LEXEL, Fremont, CA, USA), and 4 frame particle tracking software (Current, Kanomax, Osaka, Japan) (3). The pump model was made of square transparent acrylic block for flow visualization. The working fluid was 64 wt% NaI solution containing 0.1 wt% Na 2 S 2 O 3 as stabi- lizer, matching the refractive index to that of the acrylic model. The tracer particles were 150 mm SiO 2 beads, the density of which is close to that of the working fluid. Video images taken by the high speed video camera were analyzed off-line. A series of hemolysis tests using commercially available centrifugal blood pumps and their modi- fied pumps (4) revealed that hemolysis increases when the width of the impeller/casing gap is reduced in the radial and/or axial directions. Therefore, a pump with a standard gap width was compared to a Received November 1997. Presented in part at the 5th Congress of the International So- ciety of Rotary Blood Pumps, held September 10–12, 1997, in Marseille, France. Address correspondence and reprint requests to Dr. Takashi Yamane, Biomimetics Division, Mechanical Engineering Labora- tory, 1-2 Namiki, Tsukuba, 305-8564 Japan. Artificial Organs 22(5):375–380, Blackwell Science, Inc. © 1998 International Society for Artificial Organs 375

Upload: takashi-yamane

Post on 06-Jul-2016

213 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Flow Visualization as a Complementary Tool to Hemolysis Testing in the Development of Centrifugal Blood Pumps

Flow Visualization as a Complementary Tool to HemolysisTesting in the Development of Centrifugal Blood Pumps

Takashi Yamane, Balazs Asztalos, Masahiro Nishida, *Toru Masuzawa, *Koki Takiura,*Yoshiyuki Taenaka, †Yoshiaki Konishi, †Yuki Miyazoe, and †Kazuyuki Ito

Mechanical Engineering Laboratory, Tsukuba, *National Cardiovascular Center, Osaka,and †Nikkiso Co., Ltd., Tokyo, Japan

Abstract: With a 250% scaled-up pump model, high speedvideo camera, and argon ion laser light sheet, flow patternsrelated to hemolysis were visualized and analyzed with 4frame particle tracking software. Different flow patternsand shear distributions were clarified by flow visualizationfor pumps modified to have different hemolysis levels. Acombination of in vitro hemolysis tests, flow visualization,

and CFD analysis suggested a close relationship betweenhemolysis and high shear caused by small impeller/casinggaps. Because arbitrary cross sections can be illuminatedby laser light sheet, flow visualization is a useful tool infinding locations related to hemolysis in the design processof rotary blood pumps. Key Words: Flow visualization—Rotary blood pumps—Hemolysis—mechanical trauma.

During the development process of a centrifugalblood pump, large numbers of hemolysis tests arerequired even if one desires only small design im-provements. If the cause of hemolysis is essentiallymechanical, flow visualization and computationalfluid dynamics (CFD) analysis are useful in estimat-ing the test results and reducing the number of he-molysis tests. Therefore, a correlation study betweenhemolysis tests and flow visualization as well as CFDanalysis was conducted to establish a design methodusing extracorporeal blood pumps.

In terms of flow visualization, high shear can beregarded as one of the causes of hemolysis, and stag-nation or standing vortexes can be regarded ascauses for thrombus formation. Previous studies in-dicate that hemolysis is caused by collision or inter-action of red blood cells (RBCs) with solid surfaces(1), by fatigue of RBC membranes in shear flow (2),or other mechanisms. Therefore, high shear and highspeed regions need to be carefully investigated to

predict locations of hemolysis. For locations that areinaccessible to flow visualization, CFD is useful.

MATERIALS AND METHODS

The visualization apparatus consisted of a 250%scaled-up pump model with a 6 vaned semiopen im-peller (Fig. 1, diameter: 125 mm, similar to HPM-15,(Nikkiso Co., Ltd., Tokyo, Japan), high speed videocamera (4,500 frame/s, Photron, Tokyo, Japan), Ar-gon ion laser light sheet (7 W, LEXEL, Fremont,CA, USA), and 4 frame particle tracking software(Current, Kanomax, Osaka, Japan) (3). The pumpmodel was made of square transparent acrylic blockfor flow visualization. The working fluid was 64 wt%NaI solution containing 0.1 wt% Na2S2O3 as stabi-lizer, matching the refractive index to that of theacrylic model. The tracer particles were 150 mm SiO2

beads, the density of which is close to that of theworking fluid. Video images taken by the high speedvideo camera were analyzed off-line.

A series of hemolysis tests using commerciallyavailable centrifugal blood pumps and their modi-fied pumps (4) revealed that hemolysis increaseswhen the width of the impeller/casing gap is reducedin the radial and/or axial directions. Therefore, apump with a standard gap width was compared to a

Received November 1997.Presented in part at the 5th Congress of the International So-

ciety of Rotary Blood Pumps, held September 10–12, 1997, inMarseille, France.

Address correspondence and reprint requests to Dr. TakashiYamane, Biomimetics Division, Mechanical Engineering Labora-tory, 1-2 Namiki, Tsukuba, 305-8564 Japan.

Artificial Organs22(5):375–380, Blackwell Science, Inc.© 1998 International Society for Artificial Organs

375

Page 2: Flow Visualization as a Complementary Tool to Hemolysis Testing in the Development of Centrifugal Blood Pumps

modified pump, the axial and radial gap widths ofwhich had been reduced to 1/3 and 1/6 of the originalwidths, respectively (Fig. 2).

RESULTS AND DISCUSSION

The operating parameters for the scaled-up mod-els (125 mm and 132.5 mm in diameter) were flowrate Q 4 7.5 L/min, rotational speed v 4 259 rpm,and the dynamic viscosity of the NaI solution n 4 1.8× 10−6 m2/s. This condition corresponds to a flowrate of 5 L/min, rotational speed of 2,690 rpm, andpressure of 213 mm Hg for an actual size pump witha standard configuration when the similarity law issatisfied achieving the same Reynolds number vR2/n(R 4 radius), for typical human blood (n 4 3 × 10−6

m2/s). The same rotational speed, v 4 259 rpm, wasused in the small gap size model because similarpressure was attained at the speed. Therefore, theimpeller tip speed was 4.23 m/s for an actual sizepump with a standard gap width and 1.69 m/s for thescaled-up model. Pressure over 200 mm Hg was se-lected to achieve a high hemolysis rate.

Typical differences in flow patterns for the 2 pumpconfigurations were observed at the tongue and atthe impeller tip regions.

The tongue is a kind of bifurcation where the out-flow to the diffuser separates from the circulatingflow in the volute. Flow patterns in the tongue regionare shown as simple accumulations of velocity vectorplots in Fig. 3a and b. This expression is suitable forunderstanding turbulence. A laser light sheet, the

thickness of which is 1.6 mm, illuminated the centerplane of the diffuser and reached into the impellerdisk surface as shown in the bottom left portion ofthe figures. Impeller outflow smoothly entered thestraight diffuser for the standard gap size pump. In

FIG. 1. Shown is the 250% scaled-up model with a 6 vanedsemiopen impeller (125 mm in diameter, similar to the NikkisoHPM-15 blood pump illuminated by a laser light sheet).

FIG. 2. The schemas are of the pump configurations (250%scaled-up models, unit: mm), standard gap size pump (a) andsmall gap size pump with 1/6 size radial gap and 1/3 size axialgap (b).

T. YAMANE ET AL.376

Artif Organs, Vol. 22, No. 5, 1998

Page 3: Flow Visualization as a Complementary Tool to Hemolysis Testing in the Development of Centrifugal Blood Pumps

contrast, outflow for a small gap size pump movedtoward the inner wall of the diffuser forming a highspeed jet and accompanied a low speed turbulentregion in the vicinity of the outer wall. Thoughhigher tangential velocity was expected near the vo-lute walls for a small gap size pump than for a regu-lar gap size pump, the data to date have not shownclear differences.

Because the flow stays mostly in the plane of ob-servation, evaluation of two-dimensional (2-D) sheardistribution is possible. When position and velocity

in a two-dimensional region are denoted by (x,y)and (u,v), respectively, 2-D shearing velocity is de-fined by

­u/­y + ­v/­x (1)

It is interesting to note that the above definition re-sembles that of 2-D vorticity:

­v/­x − ­u/­y (2)

If ­u/­y vanishes, the 2 quantities are equivalent. Inthe analysis, grid point average vectors were calcu-lated first, and then shearing velocities were evalu-ated. The results for the 2 configurations are shownwith the contours of isoshearing velocity in Fig. 4aand b. For both configurations high shear spots canbe found at both sides of the diffuser entrance. Onthe other hand, more spots with high shear can befound in the diffuser for the small gap size pumpthan for the standard gap size pump.

The flow patterns in the impeller tip region aredescribed with grid point average vectors (Fig. 5aand b). This is because the main flow is perpendicu-lar to the illuminated plane and radial componentsare comparatively small and random. Several typicalstreak lines are also drawn for understanding. Re-verse flow along the front casing is observed for thestandard gap size pump, and consequently a streamalong the vane curves and collides with the frontcasing surface. The reverse flow in the front gap iscaused by the pressure difference between the highpressure at the impeller tip and the low pressure atthe inlet. A spiral vortex can also be seen in thevolute region (i.e., in the radial space between thevane tip and the casing). In contrast, in the small gapsize pump, reverse flow is negligible, and conse-quently a stream along the vane goes straight andcollides with the volute surface. Spiral vortexes inthe volute cannot be seen in this configuration. Be-cause higher tangential velocity exists in the voluteregion than in the front gap, high wall shear stress isexpected for the small gap size pump. Moreover, onthe rear side of the impeller, outward flow by cen-trifugal force along the impeller surface and inwardflow by pressure gradient along the casing surfacecan be observed.

The flow through the washout holes was also ex-amined in relation to thrombus formation and he-molysis. Vector data for the images with and withoutwashout holes were accumulated separately to inves-tigate the difference of flow patterns (Fig. 6a and b).When a washout hole passed, a stream extendingfrom the rear to front of the impeller was clearlyobserved. After a washout hole passed over, the

FIG. 3. Shown are flow patterns in the tongue region (vectorsaccumulated over 1,020 frames and impeller surface vectorseliminated) for the standard gap size pump (a) and the small gapsize pump (b).

CENTRIFUGAL BLOOD PUMP FLOW VISUALIZATION 377

Artif Organs, Vol. 22, No. 5, 1998

Page 4: Flow Visualization as a Complementary Tool to Hemolysis Testing in the Development of Centrifugal Blood Pumps

stream stopped and a flow in the opposite directionappeared. The flow in the washout holes alwaystended to be from the rear to front of the impellerbecause the pressure on the rear side was always

higher than that at the inlet region. Because such astream passes through the washout holes, it is sug-gested that the edges and inside wall of holes shouldbe sufficiently smooth.

FIG. 4. The shear distributions inthe tongue region (analysis basedon Fig. 3) shown are for the stan-dard gap size pump (a) and thesmall gap size pump (b).

T. YAMANE ET AL.378

Artif Organs, Vol. 22, No. 5, 1998

Page 5: Flow Visualization as a Complementary Tool to Hemolysis Testing in the Development of Centrifugal Blood Pumps

The present flow visualization has clarified the dif-ference of flow patterns for 2 pump configurations,one with a regular size gap and one with a small sizegap. A series of hemolysis tests indicated that thehemolysis level is high for pumps with small gapwidths between the impellers and the casings (4).Therefore, the visualization results sustain the hemo-lysis test results. Similar flow patterns correspondingto Figs. 5 and 6 were also obtained by CFD analysis

FIG. 5. Shown are the flow patterns in the impeller tip region(expression in grid point average vectors over 1,020 frames) forthe standard gap size pump (a) and the small gap size pump (b).

FIG. 6. The flow patterns around a washout hole that are shownoccur while a washout hole passes the plane of observation (vec-tors accumulated for 22 degree fan angle range) (a) and after awashout hole passes over the plane of observation (vectors ac-cumulated for 38 degree fan angle range) (b).

CENTRIFUGAL BLOOD PUMP FLOW VISUALIZATION 379

Artif Organs, Vol. 22, No. 5, 1998

Page 6: Flow Visualization as a Complementary Tool to Hemolysis Testing in the Development of Centrifugal Blood Pumps

with the k-« model and finite volume method (5).These suggest that flow visualization is capable ofestimating hemolysis locations such as high shear lo-cations and can assist the designer in improving thehemolytic property by changing the geometry at thehigh shear locations.

CONCLUSIONS

With a scaled-up pump model, high speed video,and laser light sheet, flow patterns related to hemo-lysis were visualized and analyzed with particletracking software. Different flow patterns and sheardistributions were clarified by flow visualization forpumps modified to have different hemolysis levels.A combination of in vitro hemolysis tests, flow visu-alization, and CFD analysis suggested a close rela-tionship between hemolysis and high shear causedby small impeller/casing gaps. Because arbitrarycross sections can be illuminated by a laser lightsheet, flow visualization is a useful tool in finding

locations related to hemolysis in the design processof rotary blood pumps.

REFERENCES

1. Blackshear PL Jr, Dorman FD, Steinbach JH, Maybach EJ,Singh A, Collingham RE. Shear, wall interaction and hemo-lysis. ASAIO Trans 1996;12:113–20.

2. Hashimoto S. Erythrocyte destruction under periodically fluc-tuated shear rate. Comparative study with constant shear rate.Artif Organs 1989;13:458–63.

3. Nishida M, Yamane T, Orita T, Asztalos B, Clarke H. Quan-titative visualization of flow through a centrifugal bloodpump: Effect of washout holes. Artif Organs 1997;21:720–9.

4. Takiura K, Masuzawa T, Endo S, Wakisaka Y, Tatsumi E,Taenaka Y, Takano H, Yamane T, Nishida M, Asztalos B,Konishi Y, Miyazoe Y, Ito K. Development of design meth-ods of a centrifugal blood pump with in vitro tests, flow visu-alization and computational fluid dynamics: Results in hemo-lysis tests. Artif Organs 1998;22:393–398.

5. Miyazoe Y, Sawairi T, Ito K, Konishi Y, Yamane T, NishidaM, Masuzawa T, Takiura K, Taenaka Y. Computational fluiddynamic analysis to establish design process of a centrifugalblood pump. Artif Organs 1998;22:381–385.

T. YAMANE ET AL.380

Artif Organs, Vol. 22, No. 5, 1998