patchAverage
Summary
patchAverage uses the "average" operation (which computes an ensemble average) rather than the "areaAverage". Its results are therefore not equal to patchIntegrate/area. This seems to me an incorrect behavior
Steps to reproduce
postProcess -func 'patchAverage(name='anyPatch', anyField)'
Example case
Compare result from postProcess -func 'patchAverage(name='anyPatch', anyField)' with postProcess -func 'patchIntegrate(name='anyPatch', anyField)'
What is the current bug behaviour?
Integral and average are not exactly related by the area.
What is the expected correct behavior?
Integral and average should be related by the area.
Relevant logs and/or images
Environment information
OpenFOAM version : v1712 Operating system : Ubuntu 18.04 Compiler : gcc 7.3
Possible fixes
In /etc/caseDicts/postProcessing/surfaceFieldValue/patchAverage change: operation average; to operation areaAverage;
Activity
@Prashant - do we need a more flexible spec? I can quite imagine the next reasonable requirement might be a mass-flux weighted patch average.
Mark
To get a flux I calculate the whole surface field (in this case fluxC)
if(writeFluxC) fluxC == phi/mag(mesh.Sf())*fvc::interpolate(C) - fvc::interpolate(Deff)*fvc::snGrad(C);
and then integrate with patchIntegrate or patchAverage
In the general case you would need to include up to three fields (phi, C, D) in the argument list (+ mag(Sf) perhaps) to calculate the diffusive and convective fluxes. This would be nice because right now I need to define a surfaceScalarField (fluxC) just to calculate fluxes across one or two boundary patches.
But I think patchAverage is a bug and different from earlier OF versions (before function objects). Consider a field defined on a unit square with the upper half = 2, the lower left quarter = 1 and lower right = 0. The average is 5/4 (meaning integral divided by area). PatchAverage (if I understand it correctly) will give a mesh-dependent value - for example if the top half was one cell (area 1/2) and the bottom was two cells (area 1/4) it will give an average of 1 rather than 5/4.
By Tony Ladd on 2019-05-01T12:13:27 (imported from GitLab project)
-
Haven't checked with doing this as a postProcess, but in the surfaceFieldValue you should be able to use a weightedAreaAverage and (for a patch) have
phias the weight field.BTW: I agree with your assessment that a patchAverage should be area-weighted.
Edited by Mark OLESEN -
@tonyladd - ok to close?
