[manuc]

Hello All

I am interested to monitor the value of my field at center of my geometry and create a field at each time step, which is equal to the product of the field at that location with the value of the field at the center of the domain (calculation of correlation).

 C(x,y,z,t) =T(x,y,z,t)* T(x0,y0,z0,t)

I access a specific point (x0,y0,z0) and use it in expressionfield with the help of the thread below
groovyBC and probe measurement


Now I would like to calculate the product of the velocity (y component) at the sensor1 (see below) with the Y-velocity in a particular plane

What I tried
Create a cloud point.

functions
   	{    

		createMeasurement 
		{ 
			type createSampledSet; 
			outputControl outputTime; 
			outputInterval 1; 
			setName sensor1; 
			set  { 
					type cloud; 
					axis xyz; 
					points ( 
					(0.01 0.05 0.05) 
					); 
          		 }      
		 }
......

Try1: Create a 3Dfield with the calculation that I need. i.e. multiplying the value from cloudpoint with U.y in the whole domain

UpCorr
            {
             functionObjectLibs ("libswakFunctionObjects.so");
			 type              expressionField;
 			 outputControl     timeStep;
                 outputInterval    1;
                 variables         ("UpL{set'sensor1}=(U-            							UMean);");   
                 expression        "U.y*UpL.y";
                 fieldName         UpCorr;
                 verbose           true;
                 autowrite         true;

            }

The issue is I only need to do the above calculation at a single plane. But now I am forced to save/calculate for a 3D field.
What I tried and failed
Create a plane as in swakexpression tutorials calculation but try with type expressionfield .

functions
{
....
UpCorr2
		 {
    		 functionObjectLibs ("libswakFunctionObjects.so");
                 type        expressionField;
    			 valueType   surface;
    			 surfaceName aperturePlane;
   			 surface
    				 {
     				     type             plane;
       			     basePoint        (0.01 0.05 0.05);
        			     normalVector     (0 1 0);
                          interpolate       true;
                      }   
                 variables         ("UpL{set'sensor1}=(U-            							UMean);");   
                 expression         "U.y*UpL.y";
                fieldName           UpCorr2;
                outputControlMode   timeStep;
                outputInterval      1;
                verbose             true;
                autowrite         true;

            }
....
}

But this creates a 3D scalarfield. What I expect is just a 2D field (just for the plane). Can someone help me to do it just for the plane

Thanks in advance

[krikre]

If I'm getting this right, you're trying to post-process during runtime. Why not sample the fileld in the points you're interested in and do all the post-processing in python afterwards?

[manuc]

Yes.What you understood is right. But I am interested in timeaveraged correlation.
<C(x+r,y,z,t)>=<C(x0,y0,z0,t).C(x+r,y,z,t)>

For this I need to calculate runtime timeaveraged of the correlation field. If I understand correctly , if I need to do the same in python I need to save a lot of time Instances

[krikre]

Sorry for not going into your swak dictionaries but it seems to me you didn't try the sampling thing which is the simples, I think. Here's what I'd do first and later look for better ways to do it.


1. Sample the the field you're interested in.
https://www.openfoam.com/documentati...e/userse21.php
Google other sources.
grep through $FOAM_TUTORIALS


After your simulation is finished, you should have a directory named "postProcessing" in your case directory. Somewhere in there there will be a TXT file with time sequences of your field for all the xyz points you specified.


2. Write the python script. import numpy. Point numpy to the TXT file with genfromtxt
https://www.numpy.org/devdocs/refere...enfromtxt.html


You should now have a matrix with the time in your first column and field values in other columns. Might get a bit tricky with the vector components. You might either remove all of the parenthesis "()" or read the field with some other utility. See PyFoam. But you should be able to reach a point where you have a matrix with i going through the time intances and j going through your space locations. You'll also want to read a txt file with the xyz coodinates where i is the label of your point and j in [0, 1, 2] is the xyz of that point. You should now be able to calculate all the averages, derivatives, intergrals etc. much more easily than coding with swak. This should work for 100 or 1000 spatial points and let's say 10k time steps. Start with fewer points, fewer time steps, then improve the procedure.

[manuc]

Thanks for the message. I didn't find any tutorial yet which could do expression calculation with sampling along a plane. I don't want the calculation to happen in the whole domain which is costly. I want the correlation to be calculated in the plane interested. Use the temporal average of correlation along the plane or a line in that plane as convergence criteria. So I need to monitor in runtime and can't calculate it after simulation. What i need is to ensure the second order statistics - two point correlation, is statistically steady, which demands the calculation along the plane in runtime. I tried to use functionobject " calculateglobalvariables" . It does calculate the field at every time step only at the plane that I am interested in. But I couldn't manage to calculate it's time average during runtime. Of course , I can calculate once the simulation is complete using MATLAB or python,but then I can't use it as a convergence criteria.

What you suggest is about plotting the data I have at the end as a graph. What I am interested is monitoring this graph during runtime to ensure the statistics is converged. Hope I have made clear on what I am interested. Any suggestion in this direction would be really nice.
Regards
Manu

[gschaider]

Originally Posted by manuc

Thanks for the message. I didn't find any tutorial yet which could do expression calculation with sampling along a plane. I don't want the calculation to happen in the whole domain which is costly. I want the correlation to be calculated in the plane interested. Use the temporal average of correlation along the plane or a line in that plane as convergence criteria. So I need to monitor in runtime and can't calculate it after simulation. What i need is to ensure the second order statistics - two point correlation, is statistically steady, which demands the calculation along the plane in runtime. I tried to use functionobject " calculateglobalvariables" . It does calculate the field at every time step only at the plane that I am interested in. But I couldn't manage to calculate it's time average during runtime. Of course , I can calculate once the simulation is complete using MATLAB or python,but then I can't use it as a convergence criteria.

What you suggest is about plotting the data I have at the end as a graph. What I am interested is monitoring this graph during runtime to ensure the statistics is converged. Hope I have made clear on what I am interested. Any suggestion in this direction would be really nice.
Regards
Manu

Calculations on a sampledSurface can be done with swakExpression with "valueType surface;". It is not possible to write out the whole surface values. Only accumulations (min, max, average, median ...). There is a functionObject dumpSwakExpression that would write out all the values (but without the spatial information9



If you want to use an expression as a convergence criterion then the functionObject writeAndEndIfSwakExpression is your friend. Time averaging can be done with storedVariables



An elaborate example for convergence checking with swak4Foam can be found in https://openfoamwiki.net/images/7/7b...onTraining.pdf


If you are really hardcore there is also the possiblity to use the python-integration



Sorry for being a bit cryptic but technically I'm on holiday

[manuc]

Dear sir
Thanks for the reply.
Regards
Manu