Bridging Between Experimental and Computation Fluid Dynamics with Physics Informed Interpolators

Dantec has introduced Physics Informed processing methods to it’s DynamicStudio software package. An approach where the two worlds of Experimental and Computational Fluid Dynamics are combined to offer cutting edge tools for flow field investigations.

  • Convert tracks to grids
  • Reduce noise
  • Enhance spatial resolution
  • Automatically compute derivatives
    • Accelerations
    • Vortex Criterion
    • Pressures

How does it work?

The current implementation is primarily used on particle track data when interpolating from the Langrangian track information onto a regular Eularian grid. This step is usually required for any detailed analysis or derivative calculation from a particle track data set. When using basic interpolation techniques to plot tracking data on a regular grid, a variety of issues can arise. Dantec’s team proposed the use of Radial Basis Functions (RBF) as an initial means of improving the interpolation step.

Figure 1 – Curve represented as a series of Gaussians

To understand the RBF approach, consider that any 2D curve of data can be represented as the sum of a number of Gaussian curves. (Figure 1) Extend that representation to incorporate a shape factor and weighting factors which can further help to describe the true shape of the curve, and you can begin to build up a much more sophisticated interpolation approach (Figure 2). This approach can then actually reduce noise in the measured data. The RBF interpolator scheme is very powerful in that it can be applied to time resolved and double frame data sets.

Figure 2 – RBF representation

In addition to the use of RBFs we can now combine these points with numerical models that describe the physics of the flow field such as acceleration models, Continuity, Vorticity models or Navier Stokes Equations. This is a ‘Physics Informed’ approach to help describe the flow field and it applies a Data Assimilation approach to combine the Physics with the measured particle track data. (https://scholarworks.calstate.edu/concern/publications/6682xb836)

As a result we can enhance the grid resolution, reduce noise and automatically populate derivatives such as pressures and Q-criterion etc.

Why not test this out with your measurement and see what extra information you may discover!

More about DynamicStudio

More about 3D PTV methods