Surface mesh extent extrapolate beyond model boundary.

KeivinNikolli

Hello everyone.

I am having some uncertainties about surface generation in LeapfrogWorks. 
In model boundary, I enclosed the object on the interval table and modified the model boundary extents for the purpose of my research. Even though I fixed up the model max and min boundary, Leapfrog is still extrapolating beyond these boundaries.

I would appreciate it if someone can have an explanation for this.

DavidWatkins

Hi @KeivinNikolli,

Are you seeing that non-domained RBF interpolants, when evaluated on other objects, are evaluated beyond their model boundaries?

What is happening is that the evaluations are accessing the underlying continuous RBF function and not the bounded RBF model. This behaviour is definitely inconsistent to the evaluation of other model types in Leapfrog and we have it on the Leapfrog roadmap to change this behaviour, or even give the user the option on whether they want the bound model or underlying continuous mathematical function.

In the meantime to get the results that you want, there are a couple of possible work arounds that I can think of:
- Creating a calculated column off the evaluated column using an IF function that doesn't assign values where outside of the north, east, elevation range of the model.
- Extract the mesh boundary from your current model, incorporate it into a GM as an intrusion, create a multi-domained RBF interpolant off the GM and your input data, and then evaluate that constrained / domained RBF model onto whatever other object you need the model information on.

I hope that helps!

Cheers,
David 

KeivinNikolli

Hey @DavidWatkins ,

Thank you very much for your detailed answer. 

It looks kind of complicated when reading your suggestion. However, I will try to solve it. 

Extra question!!!

I was trying to investigate interpolation methods in Leapfrog. I couldn't find a detailed explanation of what type of function does leapfrog uses when interpolating between borehole data. What I noticed in my research is that when the points lie on the same plane, the surface seems to be created by linear interpolation function.
As soon as I changed this co-planarity between points, the surface shows curved shapes (i think that might be due to cubic function). I am really interested to know how the algorithm selects its appropriate interpolation function.

Cheers to you too,

Keivin

DavidWatkins

Hello again @KeivinNikolli,

Shout out if you get stuck following my suggested work arounds and I can elaborate further.

In regards to the interpolation algorithms, outside of the NN, IDW and Kriging specific tools, all interpolation will be various flavours of 2.5D and 3D RBF interpolants. The parameters used to define the different RBF types are predefined, but vary according to the specific use cases and input data types.

Regards,
David