Converting FEA data as table into n-D gridded-data Array for Use with 'griddedInterpolant'

I have a dataset represented by table obtained from FE Analyses for flux linkage and i want to process this raw data into proper format through MATLAB scripting (as Look up table), where this raw data conaines parametric sweep

I would like to convert this 4-Colomn table into gridded data set in form of 3D Aarry v=(x,y,z), where x,y,z are the first three colomns and the last colomn is v. then i want to process the data with 'griddedInterpolant' to get finer gidded data. The complete dataset provides all of the information needed to create a gridded data set.

Answer 1

[Matt J]

x=unique(t{:,1}); nx=numel(x);
y=unique(t{:,2}); ny=numel(y);
z=unique(t{:,3}); nz=numel(z);

v=permute( reshape(t{:,4},[ny,nx,nz])   ,[2,1,3]);

F=griddedInterpolant({x,y,z}, v)

[Abdullah]

Thank you for your comment. that is good soluation. i think the usage of griddedInterpolant is unusful sense we interpolate with the same grid data (x,y,z), right? we might define a new grid and then use this function.

i am trying to plot the results but seems not working, do you have an idea?

f1=figure(1);
surf(x, y, squeeze(v(:,:,end)));hold on;

[Matt J]

Thank you for your comment. that is good soluation.

You are quite welcome, but please Accept-click the answer with the green button to indicate so.

i think the usage of griddedInterpolant is unusful sense we interpolate with the same grid data (x,y,z), right?

I think you misunderstand. The code I showed you hasn't done any interpolation yet. If you want to upsample v to a finer, grid (for example 2x in all dimensions), you would do,

x=unique(t{:,1}); nx=numel(x);
y=unique(t{:,2}); ny=numel(y);
z=unique(t{:,3}); nz=numel(z);

v=permute( reshape(t{:,4},[ny,nx,nz])   ,[2,1,3]);

F=griddedInterpolant({x,y,z}, v);


X=linspace(min(x), max(x), 2*nx);
Y=linspace(min(y), max(y), 2*ny);
Z=linspace(min(z), max(z), 2*nz);

V=F({X,Y,Z})  %interpolated v

To plot the the k-th slice of this result as a surface, you would do

surf(X,Y,V(:,:,k)')

[Abdullah]

Thank you Matt. it worked

but i am wondering why you put ny and then nx in the permute function?

i tryied to exchange between x with y and y with z , but did not work, any idea?

i mean something like this:

v=permute( reshape(t{:,4},[nx,ny,nz]) ,[2,1,3]);

or

v=permute( reshape(t{:,4},[nx,ny,nz]) ,[3,2,1]);

[Abdullah]

any comment Matt?

[Matt J]

Your table shows that y is the coordinate that changes most quickly as you move through the data, then x, then z

[Abdullah]

This is because i applied a Parametric sweep run in FEA, but i tryied not to sort the data according to the first colomon, so that the first colomn 'x coordinate' changes most quickly, rather than y coordinate.

i used this code, and the results are in the figured attached

[~, index] = sort(If);
If = If(index);
IPeak = IPeak(index);
Beta = Beta(index);
Landa_d = Landa_d(index);

t{:,1}=If'; 
t{:,3}=Beta';
t{:,2}=IPeak'; 
t{:,4}=Landa_d'; 

do you think this will still not work?

[Matt J]

It will work, but you will need,

v=permute( reshape(t{:,4},[nx,nz,ny])   ,[1,3,2]);

since now the y-coordinate varies the slowest.

[Matt J]

You could also just presort with,

t=sortrows(t,[3,2,1]);

v=reshape(t{:,4},[nx,ny,nz]);