Mapping 3D velocity data onto a uniform grid

I am trying to map 3D velocity data from a non-uniform computational fluid dynamics (CFD) mesh grid onto a uniform Cartesian grid. The CFD data is generated using adaptive meshing, and therefore local grid density changes with every time step. However, I need results from each time step on the same Cartesian grid for input into another program.

I have created a uniform grid using meshgrid and then tried interp3 for mapping the data to the new grid. However, I haven't been successful in this approach because of difficulties interpolating from a non-uniform grid. Is there a more appropriate function that I could try?

[ANKUR KUMAR]

Provide us the code because random example of interp3 might confuse you.

[dr23]

%Input full x,y,z position and u,v,z velocity files %All x,y,z cells are not uniform in size

%%Small Sample of x data x = [0.0075246003;0.0059246002;0.0099246003;0.0051246001;0.0083246004;0.0095246003;0.011524601;0.0087246004;0.0095246003;0.0111246]; %Small Sample of y data y = [-0.018952699;-0.0061526997;-0.00055269962;-0.00095269958;-0.0109527;0.00064730041;-0.0029526997;-0.0081526994;-0.0077526996;-0.0041526995]; %Small Sample of z data z = [-0.0329438;-0.027343801;-0.040143799;-0.0465438;-0.0281438;-0.0473438;-0.0489437990;-0.0449438;-0.0473438;-0.0481438]; %Small Sample of velocity u data u = [0.097406045;-0.014577235;0.016985994;0.019886859;-0.044792913;0.0152714;0.0097842654;0.011417145;0.012931908;0.011697576]; %Small Sample of velocity v data v = [-0.043842282;-0.1308506;-0.042359531;-0.042068601;-0.18527155;-0.045896135;-0.042315964;-0.033556957;-0.032933142;-0.039122283]; %Small Sample of velocity w data w = [-0.08908686;0.091919765;0.16070224;0.18066588;0.16235189;0.18855327;0.19154422;0.15817995;0.1560359;0.18092647];

figure(1) quiver3(x,y,z,u,v,w)

V = [u v w];

[Xq,Yq,Zq] = meshgrid(0:0.001:0.025,0:0.001:0.025,0:0.001:0.025);

Vq = interp3(x,y,z,V,Xq,Yq,Zq);

I assume that I need something to deal with the non-uniform grid size of the original mesh before I interpolate.

Answer 1

[jonas]

Try this

[Xq,Yq,Zq] = meshgrid(min(x):0.001:max(x),min(y):0.001:max(y),min(z):0.001:max(z));
Uq = griddata(x,y,z,u,Xq,Yq,Zq);
Vq = griddata(x,y,z,v,Xq,Yq,Zq);
Wq = griddata(x,y,z,w,Xq,Yq,Zq);
quiver3(Xq(:),Yq(:),Zq(:),Uq(:),Vq(:),Wq(:))

[dr23]

Thanks, that helped. However, the data spreads out quite a bit when mapped to the new grid as shown in the attached images. I will need to find a way to keep the velocity data more accurate to its original location

Original:

New Grid:

[jonas]

I dont know what spread out refers to in this context.

[dr23]

When mapped to the new grid, data are dispersed to points outside of the original fluid domain. Locations outside of the geometry now have non-zero velocities.

[jonas]

Oh I understand. The data is being interpolated. It's not too difficult to remove those points again. Let x,y,z be your original scattered data pts and X,Y,Z your new grid

%%3d boundary
tri = delaunayn([x y z]);

%%find points inside of 3d boundary 
tn = tsearchn([x y z], tri, [X(:) Y(:) Z(:)]); 
IsInside = ~isnan(tn)

You can then set the points outside of the original boundary to NaN, and you are golden. You can also have a look at this Q ( link )