Researchers at the Cleveland Clinic use computational fluid dynamics (CFD) software to peek inside designs of their catheter and rotary pumps. The devices keep blood flowing through patients while they wait for procedures such as heart transplants. Heart pumps have been trending toward smaller and safer devices every since doctors there put one in a dog in 1957.

Two types are prevalent. Catheter pumps are intended for relatively short periods. These insert into a patient's heart through the femoral artery. Rotary pumps are intended for periods that may stretch into months. Each design has a set of challenges that CFD software has helped meet.

Catheter pumps

The R&D effort for catheter pumps was led by Principal Research Engineer Markus Lorenz. They use an impeller only 4-mm diameter spinning at 60,000 rpm along with a stator and magnets that act as bearings at the pump's front and rear. Blood enters the pump and is pushed out sideways at about 45°. Lorenz says the design's low fluid-shear stresses move blood without damaging red blood cells. Too much damage could cause patients to suffer anemia and a range of other problems. Lorenz also avoided designs that show local areas of slow or recirculating flows inside the pump. These areas let blood stagnate, potentially leading to life-threatening blood clots.

The first step in the design was to get an inflow stator with a minimal pressure drop and with the best flow field for the rotating impeller. The team ran several simulations using CFD-BladeGen and CFX-5 software to find the right design for the stator. “As we refined designs, the software helped identify potential areas of stasis and high-shear stress,” says Mark Goodin, a former director of the Clinic's Medical Device Innovations Group.

Data from the stator simulation let engineers refine the impeller vanes so blood flows smoothly into the pump at the same vane angle and minimizes harmful turbulence. “Without the software, it would have been difficult to precisely set vane angles because engineers would have had to rely on simple equations to describe the flow field,” says Lorenz.

To refine the pump design, Clinic engineers combined prototype testing with CFD analysis. Before running an analysis, the team built a solid model of the pump. The 3D CAD model was used to produce a prototype by stereolithography. This let the team test performance of the initial pump design and correlate results from CFD simulations.

“The software, however, let us visualize the flow field inside the pump,” says Lorenz. “Initially we found vortices downstream of the impeller vanes. Results also helped us improve the impeller so it generates minimal internal shear stresses, and reduces the exposure time of blood to the pump,” says Lorenz.

Lei Gu, the researcher responsible for the CFD analysis, built a CAD model of the 170-mm long, 4-mm ID housing for the impeller. A final hybrid mesh used about 900,000 nodes and more than 3 million elements. CFD requires meshing only the fluid regions of the pump. Prismatic or brick elements are small against walls and increasingly larger into the fluid domain to accurately resolve velocity gradients in the wall boundary layer. After several layers against the wall, the remaining interior is filled with pyramidal and tetrahedral elements. The final mesh has 2.4 million tetrahedra, 850,000 wedges and 26,000 pyramids. The model solved with time-averaged Navier-Stokes equations and double-precision computation. The software also simulates turbulence. Solving in parallel across four, 2-GHz CPUs, and 2-Gbytes RAM, the analysis converged (solved) in 20 hr. after 400 iterations.

Rotary pumps

New designs for the rotary, or longer-term ventricular assist, pump also require CFD analysis. In a nutshell, what resulted was a simpler, less expensive and more reliable design. Unlike other rotary pumps, the Clinic's design needs no position sensors or active-control feedback. The pump consists of an impeller, with all of the rotating parts, a motor-stator, a magnetic-bearing stator, and a spiral-shaped (volute) housing. A brushless dc motor drives the pump. To reduce size, inflow is perpendicular to the rotating pump axis. Hence, the inflow bend forms an elbow.

As in the catheter pump, flow analysis allows testing design tweaks that point to a minimum inflow-pressure drop and minimum turbulence inside the inlet. The model was made of 345,000 elements with 91,000 nodes. Prismatic elements were inflated (small to larger) from the wall boundaries, with tetrahedral elements filling the interior regions. The mesh used about 261,000 tetrahedra and 84,000 pyramids.

Simulations took about two hours and 100 iterations to reach convergence. “In the future, we'd like to analyze a model of the blood flowing inside the pump. This would help determine how much hemolysis (separation of the red blood cells) actually takes place as the pump operates,” says Lorenz. To do so, engineers would introduce imaginary particles representing the constituents of blood, and track what happens to each as it moves through the pump.

Clinic officials say commercial partners are interested in both projects. Prototype testing of both pumps is helping determine how much hemolysis is taking place, and both are being readied for animal testing, a first step toward commercialization.

Make contact

Ansys Inc., ansys.com
Cleveland Clinic, clevelandclinic.org

CFD is not just for researchers anymore

The capabilities of computational fluid dynamics grew up in the aerospace industry and probably still carries a reputation for complexity that only PhDs understand. CFD-software developers, however, have made great strides bringing the capability of the technology to work-a-day engineers.

“The advantage of the software is that it lets user see inside devices in a way not possible by other means,” says Mark Burrows, application engineer with Ohio CAE, Hudson, Ohio, (ohiocae.com). “In addition, the Workbench interface is making it easier for full-time analysts and everyday engineers to get their jobs done.”

For full-time analysts, says Burrows, Workbench makes it possible to set up problems that might involve fluids, structures, and heat transfer. Such problems are often called multiphysics. “Not long ago such problems could involve three separate programs solved in a sequence possibly by three separate people. The idea has been to design an independent interface so many programs can work well together. Now, a dedicated analyst can set up and run all three simulations at one time.” The software also simplifies single simulations for individual engineers.

Burrows says improvements have also come by way of new capability in the meshers. “For instance, stress analysis might work well with equal sized elements. But in fluid studies, flow conditions at the wall are different than just a few millimeters away. So a mesher must know to place small elements at the wall where the flow might be laminar and increasingly larger ones away from it where the flow could turn turbulent,” says Burrows.

Another useful meshing feature lets users “grab” the mesh and pull more elements into areas of interest. “So if you suspect turbulence or stagnation after a first run, users can quickly adjust the mesh in those areas and rerun the simulation. It's more intuitive than other meshing methods.”