Channel impeller design for centrifugal blood pump in hybrid pediatric total artificial heart: Modeling, magnet integration, and hydraulic experiments

Abstract

Background:

The purpose of this research is to address ongoing device shortfalls for pediatric patients by developing a novel pediatric hybrid total artificial heart (TAH). The valveless magnetically-levitated MCS device (Dragon Heart) has only two moving parts, integrates an axial and centrifugal blood pump into a single device, and will occupy a compact footprint within the chest for the pediatric patient population.

Methods:

Prior work on the Dragon Heart focused on the development of pump designs to achieve hemodynamic requirements. The impeller of these pumps was shaft-driven and thus could not be integrated for testing. The presented research leverages an existing magnetically levitated axial flow pump and focuses on centrifugal pump development. Using the axial pump diameter as a geometric constraint, a shaftless, magnetically supported centrifugal pump was designed for placement circumferentially around the axial pump domain. The new design process included the computational analysis of more than 50 potential centrifugal impeller geometries. The resulting centrifugal pump designs were prototyped and tested for levitation and no-load rotation, followed by in vitro testing using a blood analog. To meet physiologic demands, target performance goals were pressure rises exceeding 90 mm Hg for flow rates of 1–5 L/min with operating speeds of less than 5000 RPM.

Results:

Three puck-shaped, channel impellers for the centrifugal blood pump were selected based on achieving performance and space requirements for magnetic integration. A quasi-steady flow analysis revealed that the impeller rotational position led to a pulsatile component in the pressure generation. After prototyping, the centrifugal prototypes (3, 4, and 5 channeled designs) demonstrated levitation and no-load rotation. Hydraulic experiments established pressure generation capabilities beyond target requirements. The pressure-flow performance of the prototypes followed expected trends with a dependence on rotational speed. Pulsatile blood flow was observed without pump-speed modulation due to rotating channel passage frequency.

Conclusion:

The results are promising in the advancement of this pediatric TAH. The channeled impeller design creates pressure-flow curves that are decoupled from the flow rate, a benefit that could reduce the required controller inputs and improve treatment of hypertensive patients.

1 ∣. MOTIVATION AND SIGNIFICANCE

Thousands of pediatric patients suffer from heart failure (HF) secondary to acquired or congenital heart disease. Approximately 14 000 children are hospitalized each year with HF, and their mortality rate is 7%–15%.^1-3 The standard treatment for end-stage HF is heart transplantation. However, there is a shortage of donor organs and an added complication of donor-recipient size matching for pediatric patients. Those pediatric patients who are on the donor waiting list have a mortality rate of 5%–39%,^1,4,5 despite 74% of patients undergoing transplantation within 90 days.¹⁰ These confounding factors motivate the use of mechanical circulatory support (MCS) devices or blood pumps as a bridge-to-transplant. The use of MCS devices in pediatric patients has increased from 508 in 2019 ⁶ to 1031 in 2020 (survival of 74% at 6 months).⁷ Despite the increase in MCS usage, clinically-approved pediatric devices lag well behind those for adults, and the use of adult MCS devices at off-design operations in pediatric patients is recognized to produce irregular blood flow, contributing to blood damage (hemolysis) and clotting (thrombosis).^8,9

To address the unmet clinical need for pediatric patients with HF, we are advancing the development of an innovative hybrid-design, valveless, magnetically levitated, MCS device that integrates two blood pumps for pediatric patients (Figure 1). This new device (Dragon Heart) has only two moving parts: an axial pump impeller to support the right side of the heart and the pulmonary circulation, and a centrifugal pump impeller to support the left side of the heart and the systemic circulation. The two blood pumps share a common rotational axis and have distinct, separate fluid domains. Here, the rotating centrifugal impeller traverses around a separate axial flow pump domain. In this rotary pump configuration, we utilize the available central hub region of the centrifugal impeller to establish vertical space for the axial flow blood pump.^9,10

FIGURE 1.

A schematic of the Dragon Heart design concept, a pediatric total artificial heart that uniquely integrates an axial and centrifugal pump into a single device to support both systemic and pulmonary circulations with only two moving parts, the magnetically levitated impellers.

Rotary pumps generally impart energy to the blood using bladed impeller designs. These blood pumps have narrow clearances between the rotating impeller surfaces and the stationary housing surfaces that are recognized to produce regions of high fluid stresses, thus risking blood cell trauma and clotting. In third-generation MCS technology, magnetic bearings are utilized to levitate the rotating impeller in the pump housing and facilitate much wider clearances between the rotating and stationary domains, thereby reducing fluid shear stress levels. Prior work in the development of the Dragon Heart focused on advancing the impeller and volute designs to maximize efficiency and reduce blood damage. The axial and centrifugal pumps were tested individually using shaft driven configurations.^9,10 A previously developed magnetically levitated axial flow blood pump^11,12 is being leveraged as the axial flow blood pump for the Dragon Heart without modification. The two blood pumps are designed to support pediatric systemic and pulmonary circulations (1–6 L/min: 60-140 mm Hg and 15–30 mm Hg at rotational speeds of less than 5000 RPM and less than 15 000 RPM, respectively). This research reflects the next phase of design and maintains the same pressure-flow and rotational speed requirements. Here, we report our concentrated design effort on a newly conceived bladeless, channeled centrifugal pump impeller design, with integrated magnetic components, through modeling and prototype testing.

2 ∣. MATERIALS AND METHODS

2.1 ∣. Design requirements and concept

Prior testing of centrifugal prototypes was accomplished by using a shaft-driven configuration with rotation driven by an external brushless DC (BLDC) motor.^9,10 In this study, we sought to remove the co-axial drive shaft and integrate magnets into the impeller to levitate and spin the centrifugal impeller around the axial pump domain. A fully validated axial flow pump is being leveraged with no modifications to the pump housing, impeller, or magnetic drive system; thus, we concentrated our design focus on the centrifugal pump. To this end, the maximum diameter of the outer casing of the axial-flow pump was used to establish a minimum internal diameter constraint for the centrifugal pump design; this dimension was 52 mm.^13,14

Considering the fluid forces and magnetic suspension designs, it has been well established that the axial fluid force is larger than the radial fluid force on the contact surface of an impeller in a rotary blood pump,^11,15 and the ability to adjust axial stiffness in order to counteract these forces was a design requirement. Thus, to maximize the potential space for magnets within the impeller for axial support stiffness, we pursued a new bladeless, channeled geometry for the centrifugal impeller. This bladeless channeled design, Figure 2A, enabled our team to utilize commercially available magnets to establish integrated magnetic bearings in the impeller and pump housing. We were able to tune axial stabilization against fluid forces on the impeller. An extended hub was also created on the underside of the impeller to house radially oriented alternating polarity magnets and phase lock the impeller with a rotating magnet ring, driven by an external BLDC motor (Figure 2B). These magnets supported the impeller radially and facilitated impeller rotation when the outer magnet ring rotated. A balance of the radial forces that were created by the interaction of the magnet rings facilitated the maintenance of an “air” gap between the two impeller rings. They also added axial stiffness as the inner and outer magnet rings were designed to passively align axially. The axial positioning of the outer magnet ring was determined by the location of the support structures.

FIGURE 2.

(A) A schematic of centrifugal pump design and magnetic drive system for conceptual testing. The hollow center conduit of the pump contains a validated axial flow pump. The impeller has an extended hub containing alternating polarity radially oriented permanent magnets. The impeller and magnet ring are locked in phase with the outer motor magnet ring, which is driven by a belt and pulley system and a BLDC motor. (B) Axial cross-section of the coupled hub and motor magnet rings demonstrating the alternating polarity radially oriented magnet coupling concept. Spacing holes for titanium rods can also be seen.

2.2 ∣. Magnetic suspension design and magnetic placement

We estimated stiffness capability and magnet sizing using the manufacturing information and magnetic properties for the N52 neodymium magnets (K& J Magnetics Inc., Pipersville, PA). The magnetic force between two permanent magnets was approximated using the magnet geometry, type/grade, and the distance and material between the magnets. There were two locations where magnet pair strength, based on gap distance, needed to be considered. The expected physical gap between the inner/outer magnet rings was 6 mm, allowing a prototype wall thickness of 1.5 mm, a secondary flow path width of 0.5 mm, and a gap width of 2 mm. The magnets were recessed 1 mm into each of the magnet rings (Figure 2B).¹⁶ Similarly, the expected physical gap between magnets in the impeller and housing was 1.5 mm, allowing a 0.5 mm tip clearance fluid gap and the magnets to be recessed 0.5 mm. The theoretical attractive force between two aligned permanent magnets separated by a small gap was calculated using Equation (1):

$$F(d)=\frac{3{\mu }_0}{2\pi }\frac{m_1{m}_2}{d^4}$$ (1)

where “$d$” represents the gap width, “${\mu }_0$” represents the permeability of the material in between the magnets, and “$m_1$” and “$m_2$” represent the magnetic moments of the two magnets, which are based on magnet geometry and material properties.¹⁶ By Equation (1) and the manufacturer's experimentally derived pull-strength data, magnets were selected that produced 30 N of radial attractive force per magnet pair in the hub and outer magnet rings (12 pairs) and 5 N of repulsive axial force per magnet pair in the impeller and pump housing.

2.3 ∣. External brushless DC motor driving rotation

Figure 2A illustrates the centrifugal pump design that includes the magnetic component locations in the impeller and the pump housing used for testing the integrated design concept. In the impeller, we mounted a ring of 12 50.8 × 12.7 × 6.35 mm (2′ × 0.5′ × 0.25′) radially oriented alternating polarity magnets. This inner ring was magnetically coupled to the outer magnet ring, which contained 12 25.4 mm × 12.7 mm × 6.35 mm (1′ × 0.5′ × 0.25′) radially oriented alternating polarity magnets. Figure 2B shows an axial cross-section of the coupled hub and motor magnet rings. The two magnet rings remained coupled through their magnetic attraction, locking them in phase with each other due to the alternating polarity magnets. The outer magnet ring was then coupled to a BLDC motor through a pulley and cable system. A CAD rendering of the pump and drive system can be found in Figure 3. The short spindle that coupled the outer motor magnet ring to the belt-bearing pulley was a thin-walled (1 mm) tube whose inner diameter (10 mm) was wide enough to allow fluid to exit the axial pump. The impeller and hub magnet ring was supported radially with both a polished titanium hydrodynamic bearing and the balanced attractive forces between the hub and motor magnet rings. To facilitate radial alignment of the pump and outer magnet ring during assembly, the outer magnet ring had 1 mm holes (seen in Figure 2A) oriented such that non-ferrous titanium rods could be inserted for alignment during assembly. These rods were used to maintain the centered position of the outer magnet ring around the pump housing while all components were being secured. The BLDC motor was designed for a higher rotational speed with lower torque than was required for these experiments. Thus, the pulley and belt drive mechanism allowed us to step down the rotational speed and step up the torque. We utilized a 57 mm OD pulley (32 teeth) on the hollow drive shaft and a 25 mm (12 teeth) OD pulley on the BLDC motor to yield about a 3:1 reduction in rotational speed and an increase in torque.

FIGURE 3.

Schematic of the benchtop testing drive system. The outer motor magnet ring (magnetically coupled to the hub magnet ring and impeller inside pumps) is mounted to an extended hollow shaft with room for the axial pump outlet. Mounted on the other end of the shaft is a belt-driven attached to brushless direct current (BLDC) motor. The rotational speed is stepped down 3:1 by the pullies from the BLDC motor.

2.4 ∣. Bladeless channel design for impeller

Per the design, the desired 5 N of repulsive force could be achieved with a pair of 6 mm diameter, 12.7 mm (0.5 in) long cylindrical magnets. This introduced the constraint of having sufficient space between channels to house magnets. To maximize the potential space for magnets, we pursued a bladeless, channeled geometry for the centrifugal impeller design. Designed impeller channels were created between raised triangular regions with adequate space to encapsulate the axial stabilization magnets. This impeller design concept allowed for the housing of magnets in places where the magnetic “air” gap distance will be narrow enough to produce the required force (gap less than 1 mm). By parametric analysis of key dimensions (i.e., channel number [4, 5, 6, and 7], inclusion of small channels in the raised triangular regions, overall impeller height [12.5 mm, 15 mm], small channel width [7.5 mm, 10 mm], and leading-edge small channel height [2 mm, 4 mm]), approximately 50 channeled designs were created and evaluated against pressure rise, capacity generation, and scalar stress requirements using computational modeling.

Additionally, a series of channeled designs were evaluated without raised triangular regions in favor of curved channels that are countermolds of conventional bladed impeller designs (i.e., an inverse molding where the channel is now located where the blade is present). A representative sampling of the highest-performing impeller designs can be found in Figure 4. The performance of these impellers was compared against design requirements of pressure rise, flow capacity, fluid stress levels, and velocity streamlines. After careful assessment, we identified the 3 optimal bladeless channeled designs for the impeller. These designs were evaluated using both CFD and benchtop hydraulic testing. Details about the computational methods are described in the next section.

FIGURE 4.

Representative selection of channeled impeller geometries with space for axial stabilization magnets. (A) 4 blade, No channel; (B) 5 blade, No channel; (C) 6 blade, No channel; (D) 7 blade, No channel; (E) 8 blade, No channel; (F) 4 blade, channel; (G) 5 blade, channel; (H) 6 blades, channel; (I) 7 blade, channel; (J) 8 blade, channel; (K) 3 channel blade countermold; (L) 4 channel blade countermold; (M) 5 channel blade countermold. Arrows indicate the location where magnets could be added.

2.5 ∣. Computational modeling of bladeless channel designs

We employed ANSYS-CFX (ANSYS Inc., Canonsburg, PA) to simulate both steady and quasi-steady blood flow through the centrifugal pump and channeled impeller design. The centrifugal pump model (Figure 5) consisted of several fluid domains: (1) inlet volute; (2) bladeless channel impeller design; (3) outlet volute; (4) the secondary flow path. Each domain was connected via fluid–fluid interfaces. We determined the global Reynolds number to be greater than 4000, therefore, we implemented a standard turbulence model. Mesh quality was confirmed using standard metrics, including aspect ratio, Jacobian ratio, skewness, and an ANSYS-based element quality metric. A grid independence study was completed to ensure that the modeling physics was independent of mesh density. Simulations were considered converged when the maximum Naiver-Stokes equation residuals fell below 1 × 10⁻³. For each model, we evaluated the pressure generation, flow capacity, axial and radial fluid forces on the impeller, fluid stress levels, and fluid streamlines.

FIGURE 5.

Computational fluid dynamics (CFD) model. Contains a rotating impeller domain and stationary inlet volute, outlet volute, and secondary flow path domains. The model uses a turbulent velocity inlet profile, static outlet pressure, and a non-Newtonian pediatric blood fluid model.

2.5.1 ∣. Fluid physics

Blood is known to be a shear-thinning solution, thus we adopted a non-Newtonian fluid model, as developed by Good et al.¹⁷ This fluid model is based on a Generalized Oldroyd-B (GOB) model and uses established coefficients from the literature. Coefficients were selected for blood with a standard hematocrit of 40%. These were derived from experiments on pediatric blood. A fluid density of 1050 kg/m³ was specified.

2.5.2 ∣. Frozen rotor steady state simulations

Flow through the bladeless impeller was initially modeled as steady state. The channeled impeller was specified as rotating in the counterclockwise direction, and the impeller hub, in the secondary flow path domain, was specified as a spinning surface. A frozen rotor interface linked regions of differing reference frames. A nonuniform mass inflow rate and operational rotational speed were specified for each steady-flow simulation. Rotational speeds of 300, 400, 600, 750, and 1000 RPM were modeled. At the outlet, an opening with a static and physiologic pressure condition was specified as the afterload. To simulate fully developed flow entering the pump, a turbulent velocity profile was defined at the inlet by Equation (2). A maximum inflow velocity was specified which produced a volumetric flow rate between 1 and 12 L/min.

$$u(r)=u_{\max }{\left(1-\frac{r}{R_{\max }}\right)}^{\beta }$$ (2)

where $r$ signifies the radial dimension, $R_{\max }$ is the maximum radius of the inlet, $u_{\max }$ refers to the expected maximal velocity at the center of the inlet, and $\beta$ corresponds to the turbulence power value.

2.5.3 ∣. Quasi-steady state simulations

To assess the effects of impeller rotational position on the performance of the centrifugal pump, a set of quasi-steady state simulations was conducted. We incrementally rotated the impeller blades by 3° and created a new, separate model for each rotational position (i.e., 120 models at 3° impeller rotations to achieve a 360° full rotation). Quasi-steady state studies were completed at a flow rate of 2 L/min and rotational speeds of 750 and 1000 RPM.

2.6 ∣. Prototype fabrication & levitation test

The centrifugal pump body channeled impellers, and magnet rings were printed using a stereolithography resin printer (Applied Rapid Technology, Fredericksburg, VA) and Somos 11 122 watershed resin. The titanium hydrodynamic bearing was custom machined and polished in the Drexel University Machine Shop. Support structures for the device were also fabricated at Drexel, including the pump mount, outer magnet ring mount, and motor mount. Rubber O-rings were used to seal adjoining pump components. A Teflon touchdown ring was inset into the radial edge of the impeller bottom to minimize friction if a touchdown occurred. Parts and components were mounted and attached to an optical board for experiments.

Prior to benchtop hydraulic testing, the prototype was evaluated to assess impeller levitation and no-load rotation. This was done by centering the bottom of the pump (containing the inner hub magnet ring) within the outer motor magnet ring. The outer ring contained removable thin titanium rods that kept the pump centered within the outer magnet ring during assembly. Once assembled and centered, the levitation and no-load rotation were assessed by manually rotating the outer magnet ring and visually inspecting what happened inside the pump and the level of resistance by assessing resistance friction and touchdown.

2.7 ∣. In vitro hydraulic testing

After levitation and no-load rotation abilities were evaluated, the hydraulic flow loop was filled with a water-glycerol (60%/40% v/v) blood-analog solution, adjusted to have a dynamic viscosity of 3.5 ± 0.1 cP, measured via an Ostwald viscometer, and a density of 1.05 ± 0.01 g/cm³, measured using a hydrometer. We assessed the free rotation of the impeller in the fluid-filled hydraulic loop, and the motor was activated. To confirm adequate axial stiffness, the impeller was incrementally ramped to 1000 RPM, the maximum speed of the experiment, and we observed the pump and how it functioned. The watershed resin material of the pump body and magnet rings was transparent for ease of observation.

Pump performance using each of the channel impeller designs was measured using a hydraulic flow loop with the water-glycerin blood analog solution. The flow loop consisted of the pump, the motor/drive system (Faulhaber Micromo LLC, Clearwater FL.), a differential pressure transducers (Validyne Engineering, Northridge, CA), two custom fluid reservoirs, a flow probe (Transonic Systems Inc, Ithaca NY), and a flow restriction clamp. The differential pressure transducer measured the difference in pressure between the outlet and inlet fluid reservoirs. Data were collected using Labview. At least 1000 data points were sampled for both pressures and flow for each experimental operating condition. These values were time-averaged to determine the pressure rise and flow capacity performance. Each of the printed impeller channel designs was tested in the hydraulic flow loop at rotational speeds of 200, 300, 400, 600, 750, and 1000 RPM. The resistance clamp was used to achieve the full range of flow capacity in approximately 0.2 L/min increments. At each flow rate, a differential pressure reading was sampled between the two reservoir tanks to determine pressure generation.

3 ∣. RESULTS

3.1 ∣. Computational results

3.1.1 ∣. Impeller design

After evaluating more than 50 initial impeller designs, the series of impellers that used a countermold of a traditional blade shape, seen in Figure 6A, were selected. The CFD predicted streamlines without considerable flow recirculation or stagnation, and the ability to produce pressure rises beyond what was required at rotational speeds below the required threshold. Figure 6B illustrates the fluid streamlines of the 4-channel impeller at a rotational speed of 750 RPM and flow rate of 2.5 L/min. Scalar stresses were low in magnitude and limited to the cutwater region (i.e., beginning of outlet volute where the radial gap between impeller and housing is narrowest) and maximum scalar stresses remained within acceptable limits. Figure 6C shows a point cloud of scalar stresses within the pump fluid domain. This series of impeller designs also provided the largest internal space to include magnets for rotor support.

FIGURE 6.

(A) Impellers selected for further study after initial computational modeling of impeller concepts. (B) Fluid streamlines showing predicted internal blood flow dynamics in the 4-channel impeller at 750 RPM and 2.5 L/min. (C) A point cloud showing the predicted scalar stress experienced by the blood in the 4-channel impeller at 750 RPM and 2.5 L/min.

3.2 ∣. Steady state computational fluid dynamics results

The steady state computational results for the three impellers can be seen in Figure 7 and are summarized in Table 1. The pressure-flow results for all three evaluated impellers are shown in Figure 7. At 750 and 1000 RPM, the pressure generated by the three evaluated impellers exceeded the required pressure generation of the pump. Generated pressures were between 90 and 100 mm Hg at 750 RPM and 160 and 185 mm Hg at 1000 RPM. The pump generated less than 20 mm Hg for all flow rates and impellers at 300 RPM. Generally, the 3-channeled impeller produced less pressure at a given flow rate and rotational speed than the 4- or 5-channeled designs, which were similar to each other. This effect was more evident with increasing rotational speed. In all cases, the pressure generation varied little with the flow rate and showed flat pressure-flow curves. This is a unique characteristic of this impeller design, and it is not typically seen in bladed centrifugal pumps.^18,19

FIGURE 7.

Estimation of pressure-flow performance of the channel impeller designs using CFD. Capacity range of 1–12 L/min and rotational speeds of 300–1000 RPM were evaluated. 3C = three channel design; 4C = four channel impeller design; 5C = five channel impeller design. Higher speeds produced higher pressure generation; pressure rise was consistent across the capacity range at a given rotational speed.

TABLE 1.

Summary of modeling data: Collected for the 3-, 4-, and 5-channeled impeller designs

	Number of channels	Average pressure generation (mm Hg)	Average axial force (N)	Average radial force (N)
300 RPM	3 Channel	14.4 ± 0.4	3.5 ± 0.3	2.8 ± 0.1
	4 Channel	15.6 ± 0.8	5.0 ± 0.2	1.9 ± 0.2
	5 Channel	15.6 ± 0.6	4.2 ± 0.4	2.2 ± 0.3
400 RPM	3 Channel	25.9 ± 0.3	7.7 ± 0.4	5.1 ± 0.2
	4 Channel	28.2 ± 0.9	10.4 ± 0.2	3.3 ± 0.2
	5 Channel	28.4 ± 0.7	9.1 ± 0.2	4.0 ± 0.3
600 RPM	3 Channel	58.5 ± 1.3	19.2 ± 0.6	11.6 ± 0.6
	4 Channel	64.5 ±0.8	25.5 ± 0.3	8.2 ± 0.2
	5 Channel	65.0 ±1.0	22.8 ± 0.4	8.9 ± 0.3
750 RPM	3 Channel	92.1 ± 0.9	30.9 ± 0.6	18.9 ± 0.3
	4 Channel	100.9 ± 1.2	40.7 ± 0.3	13.0 ± 0.3
	5 Channel	101.8 ± 1.1	36.3 ± 0.6	14.0 ± 0.5
1000 RPM	3 Channel	162.4 ± 1.7	66.7 ± 0.7	24.5 ± 0.6
	4 Channel	180.4 ± 1.7	73.9 ± 1.4	23.5 ± 0.6
	5 Channel	182.1 ± 1.8	66.7 ± 1.6	24.5 ± 1.0

Note: Averages (±standard deviation) are across flow rates from 1–12 L/min for each design and rotational speed. Pressure generation and axial and radial fluid forces on the impeller were dependent on rotational speed, and minimal variation can be seen across modeled flow rates.

As expected, higher rotational speeds yielded higher radial and axial fluid forces, per predictions (Table 1). Radial and axial fluid forces were highest at 1000 RPM and reached up to 35 N and 77 N of force, respectively. Generally, radial forces were greatest in the 3-channel models, followed by the 4- and then 5-channel designs and axial forces were largest in the 4-channel models, followed by the 5- and then 3-channel models.

3.3 ∣. Quasi-steady state computational fluid dynamics results

The quasi-steady state computational results for the 4-channeled impeller can be seen in Figure 8. The pressure generation results as a function of rotational position can be seen in Figure 8A for rotational speeds of both 750 and 1000 RPM at 2 L/min of flow. Both rotational speeds showed that pressure generation was affected by the relative rotational position of the impeller, with the pressure generation varying by up to 19 mm Hg at 750 RPM and 28 mm Hg at 1000 RPM. Fluid forces on the impeller were also evaluated with respect to the impeller rotational position, and the axial and radial results can be seen in Figure 8B,C, respectively. The axial fluid force varied with rotational position by up to 12 N at 750 RPM and 22 N at 1000 RPM. The radial fluid force varied with rotational position by up to 7 N at 750 RPM and 12 N at 1000 RPM.

FIGURE 8.

Prediction of pressure generation across the 4-channeled impeller and fluid forces on the channel impeller rotor surface using CFD. (A) Pressure generation for the channeled impeller, found to be dependent on rotational position, (B) axial fluid forces exerted on the channeled impeller, (C) radial fluid forces exerted on the channeled impeller. The function of rotational position for the 4-channeled impeller design at 2 L/min flow and either 750 RPM or 1000 RPM.

3.4 ∣. Levitation and no-load rotation

Centrifugal pump prototypes were assessed for no-load levitation and rotation abilities. It was found that if the outer magnet ring was properly aligned axially, the impeller would spin with low resistance. This indicated that the impeller was not experiencing significant touchdown on the impeller hub or the impeller top or bottom. If the impeller was not adequately axially positioned and stabilized, the underside of the impeller make contact with the housing. It was found the difference in the axial length of the magnets in the two coupled rings allowed the axial position of the impeller to be properly positioned to facilitate rotational operation.

3.5 ∣. Hydraulic testing

After no-load levitation and rotation were assessed, the pump was filled with blood analog solution and slowly ramped to the maximum test speed of 1000 RPM. It was found that the pump could reach the maximum required rotational speed, as desired. After draining the flow loop, the impeller and other internal walls of the device were examined for signs of wear resulting from the impeller touchdown. No areas of wear were observed on the top or bottom of the impeller, indicating the impeller remained in an axial position without touchdown. The narrow surface area showed evidence of wear on the interior of the inner magnet ring where it interacts with the hydrodynamic bearing.

Finally, hydraulic testing was completed. Pressure-flow performance results are seen for the three impellers at rotational speeds of 600, 750, and 1000 RPM in Figure 9. As expected, pressure generation increased with the increasing rotational speed at a given flow rate, and pressure generation decreased slightly with increasing flow rate at a given rotational speed. It was observed that the pump would produce 130 mm Hg at 5 L/min flow and 1000 RPM. It was also observed that increasing the number of channels would enhance the pressure-flow performance. Figure 10 compares the computational and experimental hydraulic results and demonstrates the alignment of results between sets of data with respect to trends describing the effects of rotational speed, flow rate, and the number of channels on pressure generation. The pressure-flow performance curves were consistent across the flow capacity range. At a given speed, the pump produced virtually the same pressure regardless of flow rate. The dominant factor contributing to the pressure generation was the rotational speed. The impellers were also able to produce a wide range of flow conditions at a given rotational speed.

FIGURE 9.

Experimental hydraulic pressure-flow data. (A–C) pressure-flow results for the 3-channel (3C), 4-channel (4C), and 5-channel (5C) designs at 600 RPM (a), 750 RPM (B), and 1000 RPM (C). (D) Pressure-flow results for the 4-channel design at 200, 300, 400, 600, 750, and 1000 rpm.

FIGURE 10.

Comparison of computational and experimental pressure-flow results. Computational (CFD) and experimental (Exp) results for the 3-channel (3C), 4-channel (4C), and 5-channel (5C) designs at (A) 750 RPM and (B) 1000 RPM. In both figures, the computational data for the 4C and 5C designs have significant overlap and is difficult to differentiate. This is also seen in Figure 7.

4 ∣. DISCUSSION

There are limited available options for pediatric heart failure patients, particularly those that require TAH or VAD support on both sides of the heart. The implementation of adult MCS in the pediatric population has demonstrated some success.^20,21 The pulsatile paracorporeal EXCOR (Berlin Heart, Berlin, Germany) offers a size for infants and small children, but these pumps have known clotting risks.¹⁴ In patients requiring MCS for both ventricles, the only durable biventricular support option available for approved clinical use is the SynCardia TAH, a pulsatile pump with two percutaneous drive lines and moving membranes and valves. The smallest size (50 ml) could be used in older pediatric patients, however, the device has known shortcomings including a propensity for driveline infection and thrombus formation.²² To address the substantial unmet clinical need for pediatric MCS devices, the Dragon Heart (Figure 1) has undergone iterative design development in the pursuit of a TAH having both an axial and centrifugal blood pump.

Here, this work served several purposes and advanced the design of the Dragon Heart toward a magnetically levitated centrifugal prototype that can be tested and characterized in vivo. An existing validated magnetically levitated axial pump can be inserted “off the shelf” into the center conduit of the presented centrifugal pump design, demonstrating the ability of complete integration for the first time. All previous iterations of the Dragon Heart were shaft driven, and the existence and location of the shafts precluded the integration of both pumps under one housing. We focused on advancing the centrifugal pump design while leveraging an existing magnetically levitated axial flow blood pump.

When designing the centrifugal pump, we made the decision to develop a new impeller design that allowed the incorporation of magnets into the top of the impeller. To this end, we evaluated more than 50 impeller designs that would leave space for magnets. After reviewing fluid streamlines and shear stress predictions using CFD, we chose to further study three impeller designs that had either 3, 4, or 5 channels in the shape of a blade countermold. All three of these designs achieved performance targets. To numerically solve the Reynolds-Averaged Navier Stokes (RANS) equations, a turbulence model needed to be implemented based on the global Reynolds number for flow in the pump. The k-epsilon turbulence model was selected based on its robustness and ability to accurately predict bulk flow pressure conditions. After completing steady-state and quasi-steady state computational modeling, we evaluated pressure generation and fluid forces. We then fabricated a prototype version of our design and completed benchtop hydraulic testing in order to verify predictions and performance.

Mesh quality was considered; all mesh quality metrics were achieved, and no issues were found upon closer inspection of the mesh in these channel regions. The CFD and experimental results showed that the pressure-flow performance curves are less dependent on the flow rate and more dependent upon rotational speed. These flatter performance curves could be an attractive feature when considering control approaches and algorithm development in the future. However, we acknowledge that this property could limit blood outflow from the pump in suddenly hypertensive patients. It is important that blood volume flow is maintained if the patient's blood pressure rapidly rises.

We observed several sudden increases (i.e., spikes) in the fluid force data, as can be seen in Figure 8. For the 4-channel 1000 RPM at 8 L/min where an axial force spike was predicted, we confirmed that the pressure generation did not deviate from the trend. The rapid and small magnitude shifts in fluid forces could be due to resonance between channel number and alignment with the cutwater region and perhaps inflow volumetric flow conditions. These sudden increases in forces correlate with the impeller channel numbers and alignment with the cutwater, as expected.^12,13 Pressure generation and fluid forces are known to be interdependent, and thus geometric differences among channel designs contributed to the fluid force shifts. The magnitudes of fluid forces from the CFD studies suggest that fluid physics between the inlet and outlet are dynamic.

When comparing the computational and experimental hydraulic performance results, the overall trends align well. For the impeller designs, we observed results indicating the designs exceed the required pressure-flow requirements at both 750 and 1000 RPM. It was also observed for all three impeller designs that pressure generation is decoupled from flow rate within physiologic flows and pressure ranges. All three channeled designs per CFD performed similarly, with the 4- and 5-channeled designs outperforming the 3-channel design by approximately 10% on average. This is thought to be due to the decrease in fluid volume within the pump domain. Computational and experimental results for the 4-channeled impeller aligned closely with each other, particularly at physiologic flows and pressures. When rotating at 750 RPM, the 4-channeled design produced approximately 100 mm Hg at flow rates of 2–5 L/min, achieving the expected performance. The CFD and experimental data were within 5% of each other at these conditions. The deviation between the two data sets increased at higher flows and higher rotational speeds. CFD model overpredicted pressure generation (<33%) at 1000 RPM and 7 L/min flow. The CFD modeling also overpredicted the pressure generation for the 3-channel design by 8.4%–25.2% at 600–750 RPM and flows of 2–5 L/min. The CFD underpredicted the pressure generation for the 5-channel design by 8.1%–36.2% at 600–750 RPM and flows of 2–5 L/min. The CFD also generally predicted performance bands between impeller designs than were measured experimentally. The computationally predicted magnitudes of axial and radial fluid forces on the impeller and hub were higher than expected, as compared to other clinically used blood pumps. For instance, the impeller in the HeartMate 3 LVAD (Abbott, Chicago, IL) experiences about 1 N of radial force²³ and the axial pump used in this study experiences less than 5 N of axial force.²⁴ This design must be improved to lower these axial fluid forces.

In general, these magnetic bearing and motor coupling configurations produced a successful and operational centrifugal pump design. The selected and commercially available magnets suspended and rotated the impeller within the fluid domain. Educated estimations were made about the forces required between magnet pairs based on the expected required torque and magnitude of fluid forces. Force values were selected to “overengineer” the design and provide stabilizing forces in far excess of what was expected. In the end, it was determined that the axial offset in the magnet rings added sufficient axial stiffness to overcome the axial thrust force.

Minor wear was noticed on the impeller hub during experimental testing; this suggests that there was an impeller touchdown during operation. Although there was no obvious impact to pump performance for any of the prototypes, we determined that the hydrodynamic bearing and the balance of the radial forces between the magnet rings are insufficient for radial stabilization. This motivates the use of active magnetic bearings in our future prototypes. In this iteration, no interactions between the axial and centrifugal pump magnetic components were observed. The centrifugal pump, the focus of this study, did not have an active magnetic bearing with position sensors that could be affected by a second, independent, magnetic flux field.

This research study also demonstrated one way to provide axial stability and stiffness to a magnetically levitated pump using passive permanent magnets. Fluid forces will be highest when axially thrust upward on the underside of the impeller, and the magnitudes of the forces are positively correlated with rotational speed. In this design, the axial stiffness provided by the offset magnet rings needed to be sufficient to overcome this thrust force and preserve the impeller's centered alignment such that there was no touchdown or interlock with the pump housing. Using sets of magnets with differing axial lengths, the axial position of the impeller can be determined. If the magnet pair is of sufficient strength, then the stiffness of the magnetic coupling is able to overcome the upward axial thrust force that is observed in centrifugal pump designs. The predicted axial fluid forces from the modeling were found to be significant; but despite this, magnets did not need to be added to the impeller in order to provide additional axial stiffness. Instead, the axial length difference between the hub and motor magnets created a restoring force when the magnets were offset axially. This restoring force was sufficient to overcome the significant axial thrust forces. The primary function, initially, of the coupled magnet rings was to provide radial stiffness and transfer rotational force. However, it was found that an axial offset between these rings can be leveraged to provide axial stiffness.

Another interesting finding was that the rotating passage frequency of the channeled impeller design created a pulsatility component in the pressure-flow performance operational curves. The CFD results showed a significant pressure differential depending on the rotational position of the impeller. The frequency and magnitude of the pressure pulse were a function of the operating speed and impeller channel number and design. Thus, this showed that the channel impeller design facilitated a pulse wave that is independent of controller modulation.

4.1 ∣. Study limitations

There are several study limitations that we would like to acknowledge. Size reductions and redesign are required for the centrifugal pump such that this device could be employed to support pediatric patients. We gleaned valuable information about the performance potential using these commercially available magnets. Downstream and upstream fluid dynamic conditions should be more fully considered using multi-scale modeling approaches. Patient-specific diseased states should also be investigated to inform the device design. A non-Newtonian fluid model was used that used empirically derived constants for pediatric blood with a hematocrit of 40%; pediatric blood hematocrit is known to vary from 30%–65% across the age range, thus the impact of fluctuations in blood properties on pump performance and hemocompatibility should be investigated. Additionally, virtual anatomical fit studies were not completed to ensure device configuration would facilitate implantation in patients. This will be performed in future work. Prototype testing of the two blood pumps being integrated under the same device housing is an important next step. Finally, the use of the hydrodynamic bearing and the long secondary fluid path likely yielded high shear stresses in the areas with narrow fluid gaps, and therefore the implementation of fully active magnetic suspension and motor configuration for the centrifugal pump would be a significant next step in this design effort.

5 ∣. CONCLUSIONS

For the pediatric Dragon Heart MCS technology, this research focuses on the design and development of the centrifugal pump component. We realize a centrifugal blood pump design that directly integrates with an existing magnetically levitated axial flow blood pump. Here, we performed steady-state and quasi-steady-state computational modeling and benchtop experimental hydraulic testing to evaluate design performance. The centrifugal pump designs achieved pump performance requirements with the 5-channel impeller outperforming other configurations. Leveraging this design phase, future work will focus on downsizing the centrifugal and axial blood pumps to reduce the size to a level that will meet the cardiovascular demands of younger and small pediatric patients. The next phase will also focus on the integration of the motor into the centrifugal pump and the addition of an active magnetic bearing system. Future work will include anatomical fit studies to elucidate size constraints for the device. Testing of future prototypes will include tandem operation and in vitro testing to assess how the magnetic drive and bearing systems in the two pumps interact. We seek to develop and translate the Dragon Heart to address the unmet medical need for a mechanical circulatory assist device to treat pediatric heart failure, and this work advances our novel design concept closer to pre-clinical animal testing.