FTLE Computation from Experimental Data

[abhisingh0701]

Hi Christian, is it possible to compute FLTEs directly from experimental data instead of simulating the flow.

[ChristianLagares]

Hi @abhisingh0701, short answer: yes (with caveats).

Now the longer answer, the source of high performance and efficiency behind the codebase is also the largest limitation (to some extent). It was designed to exploit structured grids. The cell owning mechanism traverses the mesh similar to how an octree is traversed. Any type of data is supported but it must be mapped to a structured grid apriori (or at runtime through the read function).

One possible solution is to define a structured grid of the desired resolution and interpolate/map the experimental values into de structured grid.

Does this help?

[abhisingh0701]

Hi Christian,

Thanks for the response. Yes it helped. Does the package have its own interpolants? One of the major issues with experimental data is the noise. During interpolation, that has to be a right balance between noise reduction and smoothing. Your thoughts on this?

[ChristianLagares]

@abhisingh0701, it does implement trilinear interpolation but as part of the particle velocity estimation and cell search scheme.

That said, I wouldn’t recommend it for your use case if I am reading in between the lines correctly. I completely agree with the required balance. There is also the technical issue of how your experimental data is distributed which would likely affect your selection of interpolation scheme.

To be honest, the right balance and approach is likely to be problem-dependent.