Secondary Logo

Journal Logo


Visualization Study of the Transient Flow in the Centrifugal Blood Pump Impeller

Tsukiya, Tomonori; Taenaka, Yoshiyuki; Tatsumi, Eisuke; Takano, Hisateru

Author Information
  • Free


The use of rotary blood pumps, including centrifugal and axial types, in a left ventricular assist system (LVAS) has several advantages over the use of existing pulsatile devices used for this purpose. Many researchers are now developing LVASs using rotary blood pumps. 1–4 Advantages to their use include: a rotary pump does not occupy a large amount of space because it pumps out the blood continuously, unlike the intermittent ejection by displacement type devices. The structure of the device is extremely simple, with a single moving part, known as the impeller, rotating at a constant speed to develop a pressure rise on the working fluid. The total size of the whole system, including the controller unit and the rechargeable battery, can be made sufficiently small for use as an implantable system.

Biocompatibility of a rotary blood pump in terms of antithrombogenicity is largely dependent on sufficient washout conditions in the pump, as well as the sealless mechanism. Stagnation of blood flow in the vicinity of the blood contacting surfaces can lead to platelet adhesion and finally to thrombus formation. Adverse flow phenomena, including flow separation, reattachment, and recirculation due to the secondary flows developed in the flow passages, not only deteriorate the hydrodynamic performance of the pump but also give rise to a stagnant area in the passage. Locations with insufficient shear stresses imparted by the blood flow, or with separation, are known to coincide with the spots where thrombus formation or platelet adhesion is observed on blood contacting materials. 5

Unlike pumps for industrial use, which are designed to work at the single best efficiency point, 6,7 the blood pumps for an LVAS are required to cover a wide range of flow rates according to the condition of the patient supported with the device. Moreover, the flow rate discharged by the pump is not always constant, primarily due to the beating of the natural heart. Therefore, complete elimination of such adverse flow phenomena cannot be expected over all the flow rates. It is an important issue when designing the configuration that the washout flow near the blood contacting surfaces be maintained, especially at low flow rates. Flow visualization studies are helpful in meeting this need. Many researchers have applied flow visualization techniques in the development of blood pumps. 8–11 Assessment of washout conditions is made possible by measuring the relative velocity in the vicinity of the blood-contacting surface, but it requires additional technical equipment to apply it to moving parts, such as the impeller.

The present study highlighted flow in the impeller. The relative velocity distribution to the rotating impeller was observed by the high-speed videography and particle image velocimetry (PIV) with the purpose of illustrating the unsteady fluid motion observed from the relative field.


Experimental Setup

The experimental set-up used in this study is illustrated in Figure 1. The simplified pump model was designed after the centrifugal pump that our research group has been developing as an implantable ventricular assist device. 12 The double scale model with an 80 mm diameter impeller was manufactured of transparent acrylic resin for better visual resolution. The impeller of the model was directly connected with the shaft of a DC brushless motor (FBL2-AL, Oriental Motor Co., Tokyo) by means of the sealing part. The impeller contained six two-dimensional circular vanes with an exit angle of 67 degrees. The casing had a circular inner wall with a straight outlet port. The configuration of the impeller and casing are shown in Figure 2. This pump unit was connected to the closed loop, consisting of an ultrasonic flowmeter, a resistor, and a reservoir, and was installed on the translation stage.

Figure 1
Figure 1:
Experimental setup of flow visualization using particle image velocimetry.
Figure 2
Figure 2:
Geometric configuration of the pump unit.

To acquire the image of the velocity distribution relative to the impeller blade that rotates at constant speed, we used a optical system by which the image of the rotating object remained stationary. We used a system called an “image derotator” using a beamsplitter and a right-angle prism attached to the shaft of a DC motor. 13 The principle of the system is briefly illustrated in Figure 3. The image of the rotating object at the angular velocity of ω is always symmetrical to the roof edge of the prism. When the prism is rotating at ω/2, the reflected image of the object stays in the same location. The reflected image can be observed by means of the beam splitter located between the object and the prism. The position of the pump was adjusted so that the rotating axis of the impeller coincided with the center of the roof edge of the prism by observing the reflected image captured by a CCD camera on the monitor.

Figure 3
Figure 3:
Principle of the image derotator. The right angle prism rotates around the common axis with the impeller at half speed. The reflected image seen through the image derotator generates the image observed from the rotating frame. Theta is the angular position of the blade tip inside the casing.

Image Recording

We used two different recording techniques.

High-speed videography.

The pump was illuminated by a continuous laser light sheet (GLG3680, NEC, Tokyo, Japan). The light was reflected by polystyrene particles with an average diameter of 200 microns and recorded by a high-speed camera (HSV-500, Nac, Tokyo, Japan) at 125 frames per second. The laser sheet was shed perpendicular to the axis of the impeller and produced a two dimensional velocity profile in the passages between the blades.

Particle image velocimetry.

The particle image velocimetry (PIV) system consisted of a double pulsed 25 mJ Nd-YAG laser (MINILASE-II, New Wave Research, Sunnyvale, CA), a frame-straddling CCD camera (PIVCAM10-30, TSI Co., St. Paul, MN), and a synchronizing unit (Laser Pulse Synchronizer, TSI Co., St. Paul, MN) controlled by a PC. The interval of the two laser pulses was 500 ns. Polyamide 12 particles with a mean diameter of 50 microns (Vestosint, Degussa-Huls, Dusseldorf, Germany) were dispersed in water. The laser displacement meter generated a pulse voltage per rotation, and the delay from the trigger was adjusted to obtain images with different relative positions between the blade and the outlet port.

Experimental Conditions

Experimental conditions were determined in accordance with the affinity law because the double scale model was used in this study for reasons of visual resolution. A rotary blood pump is usually compact; therefore, the effect of viscous force has to be considered, usually represented by the pump Rey-nolds number. 5,6 Consequently, we set the pump speed to operate the model pump under a similar pump Reynolds number, whereas the flow rate was adjusted in the form of nondimensional flow rate to average flow rate as in an LVAD.

The hydrodynamic characteristics of the pump in nondimensional form with a head coefficient ψ and a flow coefficient φ are shown in Figure 4: Ψ = Δ P/ρ (R ω) and φ = Q/2 π B ω, where ΔP and [φ] are the pressure rise (mm Hg) and the flow rate (L/min), R the outer radius of the impeller (80 mm), B the impeller height at the outer edge (10 mm), and ω the angular velocity of the impeller (rad/s). The measurements were conducted at the flow coefficients of 0.005, 0.01, and 0.015, which corresponded to 1.7, 3.4, and 5.6 L/min at 500 rpm. The pump Reynolds number, defined as ρ R2 ω/μ, was 8.37 × 104, Specific speed of the model pump at 3.4 L/min was 84. The laser sheet illuminated three different axial positions: (a) near the tip, (b) in the middle plane, and (c) near the hub, as shown in Figure 5.

Figure 4
Figure 4:
Nondimensional hydrodynamic characteristics of the pump. The arrows denote the flow coefficients of 0.005, 0.01, and 0.015, where visualization was conducted.
Figure 5
Figure 5:
Location of the planes illuminated by the laser light sheet: (A) top position, the upper edge of the impeller; (B) middle position, the center of the blade; (C) bottom position, in the vicinity of the shroud plane.


The captured high-speed videography image illustrating velocity distribution over the whole impeller in the middle plane at [φ] = 0.01 is shown in Figure 6. The bright area in the casing corresponds to the trace of the outlet port. The trajectories of the particles revealed that the fluid in the impeller passage was mainly discharged in passing the outlet port, while complicated vortices occupied the passage at the other positions. The recorded video of the relative fluid motion exhibited characteristics of unsteady motion of flow in terms of shear stresses acting on the blood-contacting surfaces: (1) Recirculation in the passage between the blades induced constant outward flow in the vicinity of the pressure surface, even when the passage was not near the outlet port; (2) Part of the flow near the pressure surface entered the adjacent passage over the blade tip, inducing vortices perpendicular to the main stream. This secondary flow was believed to contribute to enhancement of the washout conditions on the suction surface; nevertheless, it causes deterioration in hydrodynamic efficiency.

Figure 6
Figure 6:
Trajectories of the particle by videography.

Figure 7 shows the transient fluid motion passing the outlet port in the form of instantaneous velocity vectors determined by PIV under the same conditions. The following could be identified from the data obtained: (a) Before the passage reached the outlet port, there was only minimal net discharge. A large scale vortex occupied the center of the passage. (b) At the moment when the preceding blade passed the tongue, a relatively uniform distribution was observed, except for the large scale vortex remaining at the outer edge of the pressure surface. (c) When the tongue was in the midst of the two blades, the position of the main stream shifted toward the pressure side, where the large vortex had diminished. The secondary flows, which were generated as the preceding passage moved over the blade, developed near the suction surface. (d) After the passage had passed the outlet port, net discharge from the passage came to a halt. Outward flows still existed near the pressure surface.

Figure 7
Figure 7:
Transient fluid motion on passing the outlet port: (A) before the passage reaches the outlet port, (B) the moment when the preceding blade passes the tongue, (C) the moment when the tongue was in between the two blades, (D) after the entire passage passes the outlet port.

The flow structure, dominated by the secondary flows, was explained by measuring three different axial positions, as shown in Figure 8. Measurements were made at the moment when the preceding blade passed the tongue. The velocity distribution on the plane near the blade tip, shown in Figure 8a, clearly indicated the existence of the secondary flow entering from the adjacent passage over the blade. This flow caused separation at the blade tip and vortices perpendicular to the main flow. Presence of a large scale vortex was identified in the three different positions. This vortex blocked part of the passage, and a low velocity area was found at its center. This vortex diminished, as shown in Figure 7, after passing the outlet port, which was considered to contribute to washout of the bottom hub surface. The suction surface of the impeller blade on the bottom plane, seen in Figure 8c, showed outward flow near the surface. The corner between the blade and the bottom hub surface was sufficiently maintained.

Figure 8
Figure 8:
Effect of axial position: (A) top, (B) middle, (C) bottom.

Figure 9 illustrates the effect of flow rate on the flow profile. The velocity vectors were obtained for the moment when the preceding blade passed the tongue. A large scale vortex developed near the outer edge of the pressure surface, and the position and scale remained almost the same regardless of flow rate. The magnitude of the main flow velocity corresponded to the change in flow rate, whereas velocity distribution at the inlet position changed dynamically according to the change in the angle of attack. Flow separation from the leading edge on the suction side was observed at [φ] = 0.005. As is shown in Figure 7, the vortex diminished after passing the outlet port at [φ] = 0.01, but we found the vortex tended to remain longer and was more stable at other positions.

Figure 9
Figure 9:
Effect of flow rate: φ = 0.005, φ = 0.01, φ = 0.015.


The reliability of a centrifugal blood pump used long term as an LVAD is largely dependent upon antithrombogenicity and on the low hemolytic properties of the device. Design optimization in terms of fluid dynamics is one of the most important issues in development of rotary blood pumps, because those properties have close relationships with flow conditions. The local intense shear stresses and the period for each blood cell to be exposed to those stresses are considered to be the major causes of hemolysis. 14 However, a blood contacting surface with low shear stresses can cause platelet adhesion, which leads to thrombus formation. The impeller design of the pump is sure to influence flow conditions, and its effect on washout and adverse flow phenomena occurring in the impeller must be investigated. Flow visualization provides us with much useful information for detecting the locations with low shear stress, or the spots of intense stresses. However, interpretation of the flow inside fast moving parts, such as the impeller, is not technically easy. We have visualized the relative velocity profiles in the impeller and have clearly depicted the unsteady fluid phenomena in the impeller passage.

Flow in the passage between the blades showed a periodic fluctuating velocity distribution, primarily due to the sudden change of boundary conditions when the passage passed the outlet port of the casing. The net discharge from the passage was observed only when it passed the outlet port. One of the reasons for this would be the circular configuration of the casing, which is unlike the volute casing used in most industrial pumps, having a diverging flow duct outside the impeller to achieve circumferentially uniform discharge from the impeller. The nonuniform discharge, moreover, results in a nonuniform fluid force acting on the impeller. Flows in the impeller have been assumed to have adverse flow phenomena, such as separation, but the images and PIV results obtained in this study have demonstrated that the secondary vortical flows shed by the blade tip developed in the suction surface of the blade contribute to enhancement of washout by imparting shear stress perpendicular to the main stream on the surface. The change in flow rate directly changed the angle of attack at the leading edge of the blade accompanying flow separation. We have to consider, therefore, the trade-off of these flows between the attainable efficiency and washout over a wide range of flow rates in designing the pump configuration.

The unsteady fluid velocity distribution presented in this study also has the potential to be used as verification of computational fluid dynamics (CFD) study of the pump. The use of CFD in the design process seems to have become remarkably popular, because it can save time and cost before starting manufacture of a prototype. There are also many studies using CFD to predict pump biocompatibility, such as hemolytic properties and antithrombogenicity. 15–17 As far as the impeller flow is concerned, however, the fluid motion is very complicated and transient, and it requires careful selection of the options that the solving scheme offers us, including a computation grid, discretization scheme, and turbulence modeling, so that the results ensure the transient motion and accuracy of the quantitative prediction of biocompatibility of the device being developed.


The relative velocity distribution of flow in the impeller of a centrifugal pump designed to work as an LVAD was studied. The transient fluid dynamic phenomena, important in estimating biocompatibility of the device, were visualized by the use of high speed videography and PIV. It is also shown that the secondary flows developed in the passage, which are adverse in terms of hydrodynamic efficiency, contributed to the washout conditions on the blood contacting surface.


The authors are very grateful for the financial support from Research of Advanced Medical Technology provided by the ministry of Welfare, Labour, and Health of Japanese Government.


1. Tayama E, Olsen DB, Ohashi Y, et al: The DeBakey ventricular assist device: Current status in 1997. Artif Organs 23: 1113–1116, 1999.
2. Butler K, Thomas D, Antaki J, et al: Development of the Nimbus/Pittsburgh axial flow ventricular assist system. Artif Organs 21: 602–610, 1997.
3. Yamazaki K, Litwak P, Tagusari O, et al: An implantable centrifugal blood pump with a recirculating purge system (Cool-Seal system). Artif Organs 22: 466–474, 1998.
4. Nojiri C, Kijima T, Maekawa J, et al: Terumo implantable left ventricular assist system: Results of long-term animal study. ASAIO J 46: 117–122, 2000.
5. Schoephoerster RT, Oynes F, Nunez G, Kapadvanjwala M, Dewanjee MK: Effects of local geometry and fluid dynamics on regional platelet deposition on artificial surfaces. Arterioscler Thromb 13: 1806–1813, 1993.
6. Brennen CE: Hydrodynamics of Pumps. Oxford England, Oxford University Press, 1994.
7. Stepanof J: Centrifugal and Axial Flow Pumps: Theory, Design, and Application, 2nd ed. Maladar Florida: Krieger Publishing Company, 1957.
8. Wernicke JT, Meier D, Mizuguchi K, et al: A fluid dynamic analysis using flow visualization of the Baylor/NASA implantable axial flow blood pump for design improvement. Artif Organs 19: 161–177, 1995.
9. Ikeda T, Yamane T, Orita T, Tateishi T: A quantitative visualization study flow in a scaled-up model of a centrifugal blood pump. Artif Organs 20: 132–138, 1996.
10. Mussivand T, Day KD, Naber BC: Fluid dynamic optimization of a ventricular assist device using particle image velocimetry. ASAIO J 45: 25–31, 1999.
11. Wu ZJ, Antaki JF, Burgreen GW, Butler KC, Thomas DC, Griffith BP: Fluid dynamic characterization of operating conditions for continuous flow blood pumps. ASAIO J 45: 442–449, 1999.
12. Tsukiya T, Taenaka Y, Tatsumi E, Takano H: Performance of a newly developed implantable centrifugal blood pump. ASAIO J 47: 559–562, 2001.
13. Nakayama Y, Aoki K, Yamamoto T, Ohta H: Measurement of relative velocity distribution in centrifugal blower diffuser (in Japanese). Trans JSME 51: 325–331, 1985.
14. Nevaril CG, Lynch EC, Alfrey CP, Hellums JD: Erythrocyte damage and destruction induced by shearing stress. J Lab Clin Med 71: 784–790, 1968.
15. Bludszuweit C: Model for a general mechanical blood damage prediction. Artif Organs 19: 583–589, 1995.
16. Yeleswarapu K, Antaki J, Kameneva M, Rajagopal K: A mathematical model for shear-induced hemolysis. Artif Organs 19: 576–582, 1995.
17. Tamagawa M, Akamatsu T, Saitoh K: Prediction of hemolysis in turbulent shear orifice flow. Artif Organs 20: 553–559, 1996.
Copyright © 2002 by the American Society for Artificial Internal Organs