ArtIficia( Organs 21(7):680-685, Blackwcll Science. Inc. 0 1997 Internalional Society Cur Artificial Organs

Flow Visualization Studies to Improve the Spiral Pump Design

Aron Andrade, Jos6 Biscegli, Jose Eduardo Sousa, "Yukio Ohashi, and "Yukihiko Nose

Department of Bioengineering, Institute Dante Parzanese of Cardiology, Sao Paulo, Brazil, and *Department of Surgery, Baylor College of Medicine. Houston, Texas, U.S.A.

Abstract: The spiral pump (SP) uses centrifugal and axial pumping principles simultaneously, through a conical shaped impeller with threads in its surface. Flow visualiza- tion studies were performed in critical areas of the SP. A closed circuit loop was filled with glycerin-water solution (40%). Amberlite particles (80 mesh) were illuminated by a planar helium-neon laser light (7 mW). The particle ve- locities were recorded with Kodak (TMAX-400) black and white film, and the flow behavior was studied with a micro video camera and color video printer. The flow visualiza- tion studies showed no turbulence or stagnant areas in the

inlet and outlet ports of the SP. When using the impeller with one lead, at the top of the threads some recirculation appeared when the total pressure head increased. Two new impellers were made. One of them had the same coni- cal shape with a thread having 2 leads. The second had a thread with 2 leads, but it also had a bigger cone angle. These modifications improved the pump hydrodynamic performance, decreasing the recirculation in pumping con- ditions that require pressures over 200 mm Hg. Key Words: Centrifugal pump-Blood pump-Ventricular as- sist device-Flow visualization.

There are many different available methods used to investigate flow behavior and velocity distribution in blood pumps. Computer simulation is one tool often used to predict the flow characteristics for spe- cific blood pump regions, and the predictions lead to improvement of the pump design. Many research groups from different laboratories have utilized mathematical simulations to study and to improve their devices (1-3). Several assumptions are neces- sary for each model to simplify and to solve the full Navier-Stokes equation integrated for the control volume. The mathematical simulation has excellent accuracy, which depends upon the proper assump- tions and definition of the boundary conditions for each particular model.

Flow visualization is an important experimental tool to obtain a visible flow pattern and study the velocity field in blood pumps. This noninvasive method has minimum influence on the flow charac- teristics because there is no probe or transducer in the region being investigated. The usage of plastic

Received December 1996. Address correspondence and reprint requests to Dr. Yukihiko

Nose, Department of Surgery, Baylor College of Medicine, One Baylor Plaza, Houston, Texas 77030, U.S.A.

particles as a foreign material in suspension in a blood analog fluid is a common process in a flow visualization study. These particles must have a mini- mum concentration in the fluid to allow for visibility when properly illuminated. The low concentration of particles is important to distinguish the movement of the individual particle. Moreover, the particles must be small enough and have a density similar to that of the fluid. These characteristics minimize any distur- bance in the flow and any differences between the particle movement (direction and velocity) and the flowing fluid motion.

The selection criterion for the maximum diameter of the particles (D,) can be obtained by

(1)

where kf is the kinematic viscosity of the fluid, g is the gravitational acceleration, At is the observation time, pp is the particle density, and pr is the fluid density (4). The maximum diameter of the particles is inversely proportional to the observation time. If the difference between the fluid and particle densi- ties is approximately null, by this criterion. there is no limitation for the particle diameter. However, a

18 x Pr

g x At x (P, - Pf) Dp =

680

FLOW VISUALIZATION I N CENTRIFUGAL BLOOD PUMP 681

general rule for selection of the particle size is that the particles should be at least one order of magni- tude smaller than the smallest region under study.

Minimizing buoyancy effects is necessary to de- crease the difference between the fluid density and the particle density. To alter the fluid density, glyc- erin can be used in different concentrations, varying from 3540Y0, simulating the average density and viscosity of the blood, 1.08 g/cm3 and 3.5 cP, respec- tively. The other possibility is to reduce the density of polystyrene particles with an acetone treatment

The particle velocity is an important variable to be analyzed in different areas of a blood pump. The velocity distribution in the device can show areas of high blood cell damage. The platelets, leucocytes, and erythrocytes are sensitive to high shear stress and exposure time. A correlation between damage of these cells and shear stress and exposure time has been experimentally demonstrated (5).

Centrifugal blood pumps, in general, have a very complex flow behavior that is generated by a sophis- ticated design, established to meet the requirements for a successful device. The basic hydrodynamic re- quirements for a blood pump are as follows: satis- factory pump performance, stable flow and pressure control, low hemolysis, and antithrombogenicity. For centrifugal pumps, several research groups have been reporting detailed flow visualization studies, helping to analyze and improve each specific part of the device (6-8).

In our laboratories a blood pump, with the advan- tages of the centrifugal pumping principle but also using the axial pumping principle simultaneously, is being tested. This pump, the spiral pump (SP), uses a spiral impeller to generate both pumping actions. The impeller is conically shaped and has threads on the surface. It has a pitch of 1 inch and a cone angle of 60 degrees. The rotation of the central cone gen- erates the centrifugal effect, and the threads provide the axial motion of the blood column. The total priming volume of the SP is 75 ml. Figure 1 shows an assembled SP.

The results of hydrodynamic performance and he- molysis tests were previously reported (9). Different results were observed when the distance (gap) be- tween the housing and the top of the thread was modified from 0.5 to 2.0 mm. The best normalized index of hemolysis (NIH) was obtained with a gap of 1.5 mm, being 0.0063 k 0.0020 g/100 L in a left ven- tricular assist device (LVAD) condition (flow of 5 Wmin with a total pressure head of 100 mm Hg) and 0.0251 -t 0.0124 g/lOO L in a cardiopulmonary bypass (CPB) condition (flow of 5 L/min under a total pres-

(4)-

OUTLET

FIG. 1. The photograph shows an assembled spiral pump

sure head of 350 mm Hg). To understand the differ- ent hemolysis results obtained with different gap sizes, a flow visualization study was performed. In this study, 3 critical areas of the spiral pump were analyzed.

MATERIALS AND METHODS

The flow visualization study was performed using a planar helium-neon laser light (7 mW) illuminating Amberlite nonionic polymeric absorbent particles (80 mesh), suspended in a water-glycerin solution (40%) at 25°C. During the tests the solution was pumped in a closed loop circuit at a flow varying between 1 L/min and 5 L/min and a total pressure head between 100 mm Hg and 350 mm Hg. The closed circuit was composed of 1/2 inch tubes, con- nectors, and a flexible polyvinyl chloride reservoir bag (modified clinical blood bag) having a 500 ml volume capacity. The laser light plane was projected in the area to be studied. The following 3 critical areas were analyzed: the inlet port, the outlet port, and the top of the threads. Figure 2 shows an ex- ample of the laser light plane projected in the outlet port of the SP. Different pumps with the same shape but different gaps between the top of the threads and the housing were tested. The gaps were 0.5, 1.0, 1.5, and 2.0 mm.

To analyze the flow pattern, a continuous laser light was used, and the results were recorded using a micro camera charge coupled device (CCD, Elmo

Artif Organs, Vol. 21, N o 7, 1997

682 A. ANDRADE ET AL.

position at each pulse of light to create the illusion that the impeller was stationary.

RESULTS

CYLINDRICAL LENS

BE

FIG. 2. The schematic drawing shows the laser light plane pro- jected in the outlet port of the spiral pump and the rotating paper disc.

Co., Tokyo, Japan) and a tape recorder (Betacam- Sp, Sony, Tokyo, Japan). From the tape, some im- ages were printed using a color video printer (Mavi- graph, Sony, Tokyo, Japan). To determine the particle velocities for a specific direction (in the di- rection of the illuminated plane), a stroboscopic la- ser was used. The images were printed on photo- graphic paper using Kodak TMAX-ASA 400 black and white film.

Because the precision of the shutter speeds of photographic cameras is low, a pulsatile laser light was indicated for a quantitative study to reduce the inaccuracy of the measurements. The shutter speed had to be adjusted to obtain an exposure time long enough to capture 10 pulses of the laser light. The pulse effect of the laser light was obtained using a rotating disc made of paper with a hole that allowed the light to pass through. Each rotation (360 de- grees) allowed a dark period of 335 degrees and a light period of 25 degrees. Figure 2 shows a sche- matic drawing of the rotating disc that produced the stroboscopic effect of the laser light.

The particle velocities in the inlet and outlet ports of the SP were determined by measuring the length of the impressions of the illuminated particles on the photographic paper. When the region of the thread was studied, the length of the particle impression determined the relationship between the particle ve- locities in the radial, axial, and tangential directions. The pulse frequency of the laser light had to be simi- lar to the rotational speed of the SP impeller. The same speed for the laser light pulse and the impeller was important to provide the same image at each light pulse, i.e., the impeller had to be in the same

The images obtained with the micro camera using a continuous laser light show no stagnant area or turbulent flow in the inlet and outlet ports of the SP. Figure 3 shows the image from the inlet port of the SP with a rotational speed of 2,000 rpm, pumping 5 L/min of solution under 140 mm Hg. Figure 4 is the outlet port of the SP at 2,400 rpm, pumping 5 L/min under 220 mm Hg. Varying the gap from 0.5 to 2.0 mm and the total pressure head from 100 to 350 mm Hg for a specific flow, no difference in the flow be- havior in the inlet and outlet ports of the SP was observed.

At the top of the thread, the flow visualization demonstrated that by increasing the total pressure head, some recirculation appears. Figure 5 shows the images from the top of the thread of the SP at 2,300 rpm, pumping 5 Wmin under 200 mm Hg. When the total pressure head increases, the particle velocity at the top of the thread increases. When the size of the gap decreases, the particle velocity also increases, increasing the length of illuminated particle streaks. Figure 6 shows the relationship between the length of the particle streak in the direction of the illumi- nated plane and the gap size.

DISCUSSION

During analysis of the images obtained using the video camera and study of the photos from the inlet and outlet ports of the Spiral Pump, no turbulence or stagnant area was observed. This means that the dif-

FIG. 3. This flow visualization is in the inlet port.

Artif Orguns, Vol. 21, No. 7, 1997

FLOW VISUALIZATION IN CENTRIFUGAL BLOOD PUMP 683

FIG. 4. This flow visualization is in the outlet port

ferent results of hemolysis testing were not caused by any change in flow behavior in these regions. Ob- servation of the flow behavior and the particle ve- locity at the top of the thread provided important information about the different hemolysis test re- sults obtained with different gap sizes. The different gap sizes showed changes in the particle velocities, and this explained the different hemolysis test re- sults. The velocity has a direct relationship with the turbulence and the shear stress, which increase the blood cell damage. Also, differences in velocity dis- tribution modify the exposure time of the blood cells to stress.

When the total pressure head was increased, the recirculation at the top of the thread was also in- creased by the increasing particle velocities. This in-

creased the shear stress and made the blood circulate for a longer period inside the pump. When the gap was increased from 0.5 to 2.0 mm, the particle veloc- ities in the axial and radial directions decreased (Fig. 6) . Decreasing the particle velocity caused the he- molysis to decrease until the gap was 1.5 mm. The hemolysis results with a gap of 1.5 mm were consid- ered satisfactory, and no significant difference was observed when comparing these results with the lit- erature for similar pumps (9). However, a gap of 2.0 mm generated larger hemolysis than a gap of 1.5 mm. This occurred because a larger volume of blood recirculated through the gap, decreasing the pumped flow. A higher rotation speed (rpm) of the impeller was necessary to maintain the same flow and pres- sure.

At a higher pressure or at a rotational speed higher than 2,000 rpm, recirculation was important. Under either condition the axial pumping principle had a greater effect than the centrifugal principle, increasing the recirculation at the top of the thread. Two modifications were proposed to minimize this recirculation and improve the SP design.

The first modification was to make a spiral impel- ler with 2 leads instead of only 1 to decrease the differences in the pressure in the regions above and below each thread. The region below each thread has a higher pressure than the region above it, and as this difference is made smaller, the returning flow decreases. Therefore, maintaining a gap of 1.5 mm and using a thread with 2 leads having the same pitch of 1 inch should decrease recirculation. Moreover, 2 leads make the spiral impeller design more symmet- ric, improving the stability of the impeller by de-

FIG. 5. This flow visualization is at the top of the thread.

Arrif Organr, Vol 21. N o 7, 1997

684 A. ANDRADE ET AL.

1.1 h

E E - 1.0

3 Y

L

0.9 u - f

0.8 c 0

0.7

8 L;I

0 0.5 1 1.5 2 2.5 GAP (mm)

FIG. 6. The graph shows the relationship between the length of the particle streak and the size of the gap.

creasing the asymmetric forces inside the pump that can cause impeller vibration.

The second modification was to increase the angle of the cone for both the impeller and the housing. The angle was changed from 60 degrees to 70 de- grees. This modification increased the centrifugal pumping effect. By increasing the centrifugal forces in the pump, the effects of the axial and centrifugal pumping principles became more equal even in a rotational speed above 2,000 rpm. The recirculation at the top of the thread decreased for this operating condition. On the other hand, this modification pro- vided another advantage; the height of the pump was decreased, and the total priming volume was reduced from 75 ml to 65 ml.

Figure 7 shows the spiral impeller with 1 lead (left), with 2 leads (center), and with 2 leads and a larger cone angle (right).

FIG. 7. The photograph shows 3 designs of the spiral impeller: 1 lead (left), 2 leads (center), and 2 leads and a larger cone angle (right).

Using the same closed circuit as previously de- scribed, the hydrodynamic performance test was re- peated with the pump design modifications. Figure 8 shows the relationship between the total pressure head and flow, comparing the hydrodynamic perfor- mances of an SP with 1 lead (SP I), another with 2 leads (SP II), and a third design with 2 leads and larger cone angle (70 degrees) (SP 111).

CONCLUSION

As shown in previous studies, the SP with a gap of 1.5 mm has a satisfactory pumping performance and causes low NIH when its data are compared to the data for similar devices (9-12).

Flow visualization studies were performed to verify the causes of the NIH increase when the gap differed from 1.5 mm, especially in pumping condi- tions that required a rotational speed above 2,000 rpm. In that condition the axial pumping effect be- came greater than the centrifugal effect, and recir- culation commenced at the top of the threads.

By fabricating the spiral impeller with 2 leads and increasing the angle of the cone, pump hydrody- namic performance improved as a result of the de- crease of recirculation at the top of the threads. These modifications improved the hydrodynamic pump performance. More tests are being performed to verify the effect of the new pump design on the normalized index of hemolysis for CPB pump con- -

0 1 2 3 4 5 6 7 a 9 10 Flow (Urnin)

FIG. 8. The graph shows the relationship between the total pres- sure head and flow rate for 3 different impeller designs and dif- ferent pump rotation speeds (1,000 to 3,000 rpm). The solution was water and glycerin (60%/40%) at 25°C

Artif OrganA, Vol. 21, No. 7, 1997

FLOW VISUALIZATION IN CENTRIFUGAL BLOOD PUMP 685

ditions, i.e., 5 L/min of flow and 350 mm Hg of pres- sure.

Acknowledgments: This study was partially funded by CNPq BrasiliaiBrazil. The authors acknowledge the tech- nical assistance of the employees and the valuable help and suggestions of all colleagues at Baylor College of Medicine (BCM), Institute Dante Pazzanese of Cardiol- ogy (IDPC), Sorocaba College of Technology (FATEC). and University of Sao Paulo (USP). Also, the authors would like to thank the Image Sciences Division at the National Aeronautics and Space Administration (NASA) for their essential collaboration in this study.

REFERENCES 1. Schima H. Trubel W. Mueller MR. Papantomis D , Salat A ,

Schlusche C , Prodinger A, Spitaler F, Krausler S. Loscrt U. Thoma J . Wolner E. Development of a centrifugal blood pump with minimal hemolysis. In: Schima H , Thoma H. Wie- selthaler G. Wolner E, eds. Proceedings of the Internationrrl Workshop on Rotary Blood Pumps. Vienna, Austria. 1991: 141-7.

2. Papantonis D. Numerical prediction of the shear stresses and the mean stay time for radial flow impellers. In: Schinia H. Thoma H, Wieselthaler G, Wolner E, eds. Proceedings ofthr Internatioritrl Workshop on Rotary Blood Pumps, Vienna, Austria, 1991:63-9.

3. Pinotti M, Rosa ES. Computational prediction of hemolysis in a centrifugal ventricular assist device. Artif Organs 1995:19(3): 267-73.

4. Merzkirch W. Flow visualiztzrion. New York: Academic Preys. 1974.

5. Wurzinger LJ, Opitz R. Hematological principles of hcnioly- sis and thrombosis with special reference to rotary blood pumps. In: Schima H , Thoma H, Wieselthaler G, Wolner E. eds. Proceedings of rhe liiterntitional Workshop on Rotary Blood Pumps, Vienna, Austria. 1991: 19-25.

6. Treichler J , Rosenow SE, Damm G, Naito K. Ohara Y. Mi- zuguchi K, Makinouchi K, Takatani S. Nose Y. A fluid dy- namic analysis of a rotary blood pump for design improvc- ment. Artif Organs 1993;17(9):797-808.

7. Schima H, Huber L, Melvin D. Trubel W, Prodinger A. Lo- sert u, Thomas H, Wolner E. Effect of stationary guiding vanes on improvement of the washout behind the rotor in centrifugal blood pumps. ASAlO J 1992;38:M220-4.

8. Ikeda T, Yamane T, Orita T, Tateishi T. A quantitative visu- alization study of flow in a scaled-up model of a centrifugal blood pump. Arrif Organs 1996:20(2):132-8.

9. Andrade A, Biscegli J , Dinkhuyscn J. Sousa JE, Ohashi Y. Hemmings S, Glueck J. Kawahito K, Nose Y. Characteristics of a blood pump combining the centrifugal and axial pumping principles: The spiral pump. Artif Orgcms 1996;20(6):605-12.

10. Ohara Y, Sakuma I, Makinouchi K. Damm G , Glueck J. Mi- zuguchi K, Naito K. Tasai K. Orime Y, Takatani S, Noon GP, Nose Y. Baylor Gyro pump: A completely seal-less centrifu- gal pump aiming for long-term circulatory support. Artif Or- gans 1993;17(7):599-604.

11. Damm G, Mizuguchi K. Bozeman R. Akkerman J, Aber G. Svejkovsky P, Takatani S, Nose Y, Noon GP, DeBakey ME. In vitro performance of the BayloriNASA axial flow pump. Artif Orguns 1993;17(7):609-13.

12. Naito K. Mizuguchi K, Nose Y. The need for standardizing the index of hemolysis. Arrif Organs 1994;18(1):7-10.