Trying to get a contour plot from CFD ASCII data

Hello everyone. I am trying to use MATLAB as a post processor to visualize my CFD simulation results. I use ANSYS Fluent as the solver and I exported ASCII data (on the node) of certain quantities (velocity, vorticity etc). Now, I am interested in getting a 2D filled contour so I am using the contourf function.

The problem is that I make use of the command griddata for interpolation and meshgrid to generate a rectangular mesh (b/w min and max x and y). But my problem domain inherently is not rectangular but something like this . So of course, griddata interpolates values into regions where there should be none at all. Does anyone know of any workaround for this? Ideally, I would want the meshgrid itself to be oriented exactly to my domain and not be rectangular so that whatever interpolation griddata does, it is only inside the actual domain.

I have attached the code below.

clc
clear
close all

files = dir('*dat');

for i = 1 : 10
    fname = files(i+1).name;
    A{i} = dlmread(fname,'',1,1);
    A{i}(:,3) = 0;  % set all z coordinates to absolute zero for consistency
end

x = A{1}(:,1);
y = A{1}(:,2);

n = 500;   % number of grid points
xx = linspace(min(x),max(x),n);
yy = linspace(min(y),max(y),n);
[xi,yi] = meshgrid(xx,yy);

for i = 1:10

      quantity = A{i}(:,5);
      zi = griddata(x,y,quantity,xi,yi);
      zi(zi == 0) = NaN;

      contourf(xi,yi,zi,50,'Linestyle','none')
      set(gcf,'units','normalized','outerposition',[0 0 1 1]);
      axis equal
      xlabel('X');
      ylabel('Y');
      colormap('jet');
      colorbar
      pause(0.1);

end

Answer 1

[Prashant Arora]

Hi Digvijay,

You can include the boundaries of your domain as a vector (let's say function of x) and replace the values of meshgrid by NaNs. I am including a simple example below.

x = linspace(0,1,1000);
y = linspace(0,10,1000);
boundaryUp = 5 + sin(10*pi*x);
boundaryDown = 1 + sin(5*pi*x);

[X,Y] = meshgrid(x,y);

Y(Y > boundaryUp) = NaN;
Y(Y < boundaryDown) = NaN;
Z = 1 + X.*0 + Y.*0;
mesh(X,Y,Z);

[Digvijay Rawat]

Thanks yaar. I did have this in mind but was too lazy to implement it so was hoping for an easy way out. Cheers!

[Abdallah Ghazal]

Hi, I do the same kind of postprocessing of my CFD results in MATLAB. The problem I have is the long time it takes MATLAB to postprocess that data. I bieleve this is because the 'griddata' function is inside the for loop and I do have more than 100 iterations. I am wondering if there is any way to speed up the postprocessing. For your information, a simulation of 100 iterations takes about an hour of clock time. I would appreciate any assistance. Thanks.