View Issue Details

IDProjectCategoryView StatusLast Update
0001223OpenFOAM[All Projects] Bugpublic2015-01-01 14:06
Reporteruser887Assigned Tohenry 
PriorityurgentSeveritymajorReproducibilityalways
Status resolvedResolutionfixed 
PlatformLinuxOSOpenSuseOS Version11.2
Product Version 
Fixed in Version 
Summary0001223: Case couldn't run continuously after reconstruction and decomposition again, pressure field went wrong directly.
DescriptionI am using OF 2.1.0 version and I use cyclic boundary condition for my pipe flow simulation. I modified icoFoam solver by adding a source term in the equation to maintain the flow while using cyclic boundary conditions. I ran a laminar pipe flow in parallel mode, the thing is once I reconstruct my case and decompose it again in the same manner, I find that the U, p and phi fields under each processor folder are different from the original ones and when I run the re-decomposed case the simulation couldn't continue and the pressure filed went wrong straight away (in particular yielding large values near the two cyclic boundaries).

I tried the same case with different boundary conditions. With normal boundary conditions (e.g., velocity inlet and pressure outlet), the U, p and phi fields after re-decomposition are the same (almost identical with some differences in the last digit) as original ones and simulation can continue smoothly without any errors. However, whenever I use cyclic boundary conditions, the reconstruction and decomposition will result in different initial fields comparing with original ones and simulation went wrong immediately with funny pressure results.

Will the cyclic boundary conditions affect reconstruction and decomposition process? Why I can't reconstruct and re-decompose my case to continue the simulation?
Tagsoutlet, parallel, Pre-processing

Activities

user4

2014-03-17 16:49

  ~0002953

Can you repeat this in 2.3.x? If so can you post a small testcase?

user887

2014-03-17 16:59

  ~0002954

Sorry that we don't have the latest version on our system currently and it may take quite a while to compile it cause we don't have the updated gcc version.

Can you give me a hint on what may went wrong during the reconstruction and re-decomposition? I just found that for the same case, if I reconstruct the case and decompose it again, the fields under the same processor folder are different from the original one (only when I use cyclic boundary condition).

Thank you very much!

user887

2014-03-19 22:48

  ~0002964

I have fixed this by using 'preservePatches' in decomposePardict to avoid splitting the cyclic boundary pairs.

user4

2014-03-21 10:09

  ~0002968

There have been fixes in 23x to decomposePar in combination with cyclics being split across multiple processors.

Issue History

Date Modified Username Field Change
2014-03-14 14:04 user887 New Issue
2014-03-17 10:01 user887 Tag Attached: outlet
2014-03-17 10:01 user887 Tag Attached: parallel
2014-03-17 10:01 user887 Tag Attached: Pre-processing
2014-03-17 16:49 user4 Note Added: 0002953
2014-03-17 16:59 user887 Note Added: 0002954
2014-03-19 22:48 user887 Note Added: 0002964
2014-03-21 10:09 user4 Note Added: 0002968
2015-01-01 14:06 henry Status new => resolved
2015-01-01 14:06 henry Resolution open => fixed
2015-01-01 14:06 henry Assigned To => henry