www.ansys.comANSYS Advantage • Volume I, Issue 2, 2007s12

BIOMEDICAL: ARTIFICIAL ORGANS

An important challenge facing thedesign of turbodynamic ventricularassist devices (VADs) intended forlong-term cardiac support is the opti-mization of the flow geometry tomaximize hydraulic efficiency whileminimizing the peak shear stress in theblood flow. High efficiency reduces therequired battery size while low shearreduces the number of red blood cellsthat are ruptured by the pump. A pedi-atric heart-assist pump is particularlychallenging. Due to its small size(about 28 mm diameter by 51 mmlength), the design laws for adult-sizedpumps do not apply, and they cannotbe scaled. Therefore, the design ofpediatric blood pumps must rely onmodern design approaches to opti-mize the flow path. Computational fluiddynamics (CFD) has been widely usedin the field of artificial heart pumps forthe analysis of internal flow because itoffers an inexpensive and rapid meansof acquiring detailed flow field informa-tion that is expensive and painstakingthrough in vitro testing. LaunchPointTechnologies, Inc., in the UnitedStates, which developed the first mag-netically levitated (maglev) heart pump(the Streamliner ventricular assistdevice that reached animal trials in1998), finds that CFD is a powerful toolin the performance assessment andoptimization of artificial heart pumps.

LaunchPoint has developed a CFD-based design optimization approach

Designingwith HeartCFD-based design optimization for a miniature ventricular assistimplant can shave years off themedical device development cycle.

By Jingchun Wu, LaunchPoint Technologies, Inc., California, U.S.A.and Harvey Borovetz, McGowan Institute for Regenerative MedicinePennsylvania, U.S.A.

that integrates internally developed 3-D inverse blade design methods,parameterized geometry models, automatic mesh generators and math-ematical models of blood damage withthe commercial ANSYS CFX solver.The system provides rapid optimiza-tion for various types of centrifugal,mixed-flow and axial-flow bloodpumps. The ANSYS CFX solver waschosen because of its robustness forcomputations with multiple frames ofreference (MFR) (the coupling betweenrotating and stationary components).

A new LaunchPoint VAD, Pedia-Flow™ is intended to deliver a flow rateof 0.3 to 1.5 l/min against 100 mmHgpressure rise to neonates and infantsweighing 3 to 15 kg. The PediaFlowwas designed with a magnetically sus-pended, mixed-flow style impeller witha single annular flow gap between therotor and housing to avoid unfavorableretrograde flow and separation. Theshear stress transport (SST) model, alow Reynolds number turbulencemodel, was selected for the turbulentflow simulation, which was justified by the representative Reynolds number of ~30,000 based on theimpeller outlet diameter and the pumptip speed. Although blood exhibits non-Newtonian behavior at very lowshear rates, many studies have shownthat blood can be modeled as a Newtonian flow at a shear rate largerthan the threshold of a 100 s -1. The

shear rate in the computational modelof the PediaFlow is much larger thanthis threshold, so Newtonian blood witha constant viscosity of 0.0035 Pa-s anda density of 1040 kg/s3 was assumedfor the simulations.

The CFD-predicted velocity vectorsat both the mid-span blade-to-bladeregion of the impeller and the vane-to-vane region of the stay-vanes show avery smooth distribution without anyvortices at the nominal flow conditionfor the optimized PediaFlow model. Asliterature is replete with anecdotal evi-dence that recirculating flows lead toattachment of platelets to biomaterialsurfaces — which in the clinical VADsetting can promote blood clot forma-tion — reverse flows and vortices areundesirable. The CFD results foundthat a smooth and gradual transition inthe secondary flow velocity was present at the curvature of one inflowand outflow cannula geometry. Thisgraduation helps to prevent separationand reversal flow for the primary flowvelocity. In addition, the predictedpathlines of representative particlesthrough the entire flow region did notexhibit any vortices.

The exposure of blood elements toshear stress above a certain thresholdas a function of exposure time cancause hemolysis, which actively breaksopen the red blood cells; activateplatelets, which can cause clottingproblems; and denature proteins, which

The PediaFlow ventricularassist device provideslong-term cardiac supportfor infants.

ANSYS Advantage • Volume I, Issue 2, 2007www.ansys.com s13

BIOMEDICAL: ARTIFICIAL ORGANS

alters the proteins so they can no longer carryout their cellular functions. Thus, it is desirableto minimize the shear stress that blood passing through the pump may experience.Using the results of the CFD simulation, a plotof shear stress versus exposure time for particles passing through the pump demon-strates relative uniformity within the annularflow gap region, but it is less uniform withinboth the impeller and stay-vane regions. The overall mean blood damage through the entire domain of the model is dividedaccording to the three main regions of theflow path: impeller, annular gap and the stay-vane. The analysis reveals that the hemolysislevel in the annular gap region is highest,accounting for more than 50 percent of thetotal, while the level of hemolysis in theimpeller region and stay-vane region is almostthe same, each causing approximately 20 to25 percent of the total blood damage.

CFD-based design optimization with theintegration of the ANSYS CFX solver can significantly reduce the design optimizationcycle from years, compared to the traditionaltrial-and-error methods, to just severalmonths. It provides detailed and useful flowfield information from which blood damagemay be computed, and it also predicts thehydrodynamic characteristics such as therelationship of developed pressure and efficiency to flow rate. ■

This research was supported in part by NIH ContractNo. HHSN268200448192C (N01-HV-48192).

PediaFlow is a trademark of WorldHeart, Inc.

Shear stress history from impeller inlet to stay-vane outlet Proportion of total blood damage at different pump components undernominal flow condition

Predicted smooth velocity vectors at mid-span blade-to-blade region of the impeller (left) andmid-span vane-to-vane region of stay-vanes (right)

Secondary flow streamlines at sections of inflow cannula (left) and sections of outflowcannula (right)

Pathlines of particles at inflow cannula and impeller side (left) and stay-vanes side andoutflow cannula (right)

0.0025

0.0020

0.0015

0.0010

0.0005

0.0000Impeller Annular Gap Stay Vane Total

Giersiepen

Heuser

0.002337 0.002341

21.20% 19.5%

52.5%57.7%

26.27%22.8%

dHb

300

250

200

150

100

50

00.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18

Shea

r Stre

ss (P

a)