NUMERICAL DESIGN IMPROVEMENT AND EXPERIMENTAL TESTING OF A SMALL VAD

Background: Computational Fluid Dynamics is a useful tool for developing Ventricular Assist Devices (VADs). However, the results are not necessarily
trusted, and validation studies are essential to increase confidence.
Validation studies usually require expensive, time consuming, for example Particle Image Velocimetry (PIV). Simpler validation methods, which could
be incorporated more naturally into the design process, are therefore
desirable.
Aim: The aim of this work was to investigate the extent to which design
changes in the computational domain produced measureable effects on the
experimental pressure-flow characteristics, with a view to using rapid prototyping
of early design iterations to increase confidence in CFD.
Methods: A small pump, similar to a VAD, was designed using CAD. The geometry
was meshed and CFD calculated using ANSYS CFX. Mesh studies were
conducted, and several turbulence methods were investigated, to assess errors.
Transient simulations were performed to estimate the steady flow pressure-
flow curves for a range of speeds. Based on examining the results a
series of manual design changes were made and the simulation results were
updated for each design iteration. A physical prototype of the pump was created
from 3D printed parts; these fitted together allowing replacement of
individual components.
The pump was driven with an external motor and shaft. The pump is currently
being tested in a custom designed rig.
Results: For the original design the operating speed to reach the design point
(100 mmHg at 5 l/min) was 10,500 rpm. At this speed the design iterations
resulted in changes to the pressure head of between 10 and 200 mmHg;
alternatively speed changes of 600 to 5000 rpm were required to produce
the design point.
Conclusions: These pressure differences are greater than both CFD and
transducer measurement errors, meaning the design changes should produce
measurable effects. However, rapid prototyping also has inherent errors.
Good agreement between CFD and experimental pressure-flow curves
in early design iterations could be extrapolated to assume good agreement
at the later design stages.