3D Interpolation around solid regions in volume data

Hello.

I have a large table of 4-D data (X,Y,Z,Pressure) from a CFD simulation which is non-uniform. I have re-sampled using the meshgrid and griddata to produce a linear interpolated uniform dataset. However, there are fixed solid (no flow data) regions within the workspace that are now extending beyond their bounds due to the interpolation process. I have tried flagging the solid regions using a 0 or a NaN but neither has helped to produce accurate boundaries even at a high resolution grid. Is there any a way to avoid the interpolation process using any of the points from the solid region?

[Matt J]

Since the data is discrete, how are you able to judge where the true bounds (in continuous space) are supposed to be?

Answer 1

[Ahmet Cecen]

If you have geometric knowledge of the invalid regions, one workaround is to eliminate them from the interpolation process. Use griddata or scatteredInterpolant on the data without including the invalid regions. Then you can create a separate binary map of the invalid regions using your geometric knowledge and eliminate them from the result.