Validation

“Remember that all models are wrong; The practical question is how wrong do they have to be to not be useful”

George E.P. Box

Computations are in general prone to various types of errors. One premise of our service is to deliver reliable results with the highest demand on both accuracy and efficiency. For each application it is important to know the acceptable level of error.

Programming Errors. As we are mainly using commercial software, code validation is not necessary. Benchmark tests are constantly performed by the software developers themselves.

Iteration Errors. Root mean square normalized residuals are monitored in each simulation for the estimation of iteration errors. The target criterion is a residual drop of four to five orders of magnitude.

Discretization Errors. To minimize discretization errors, systematical sensitivity analyses are performed for the fluid and solid mesh, the time step size and (when required) the pulse cycle independence.

Modeling Errors. Some modeling approximations (e.g. geometry, boundary conditions, fluid properties, turbulence) have to be made in each study which may affect the solution. To check if the physical model is represented accurately enough, our computational models are constantly and thoroughly validated against experimental data of high accuracy, e.g. Particle Image Velocimetry (PIV) measurements. PIV is a non-invasive, optical flow visualization technique and based on the motion-detection of illuminated particles what allows the calculation of the velocity field. For this purpose, we work in close collaboration with the Department of Cardiovascular Engineering at the RWTH Aachen.

Some examples of our validation process:

Total Artificial Heart

Quantitative validation of a Fluid-Structure-Interaction (FSI) simulation of a pulsatile Total Artificial Heart against stereo Particle Image Velocimetry (Stereo-PIV) measurements.

Cannulation of the subclavia

The numerical model of subclavian artery cannulation during cardiopulmonary bypass was validated experimentally with an in vitro test rig.

Turbulent jet flow

Several turbulence models to predict jet flows were evaluated. The calculated flow fields were validated against Particle Image Velocimetry (PIV) measurements.