It is typical for DNS and LES
It is typical for DNS and LES simulations especially with added effects (density stratification or frame rotation) that one may need to check the flow from time to time to see if it is fully developed (according to certain criterion) so, one can start sampling. This creats a break in making the computation fully automatic which is necessary in running a high resolution computation on a cluster with PBS. A solution to this problem is to do the spatial averaging (collapsing field) in homogeneous directions for fields ever certain time interval (proportional to the integral time scale of the flow) and delay the time averaging of such collapsed fields till the end of the computation. This has two advantages:
1) one may freely choose when to start sampling independent of having to stop the simulation.
2) as a result of 1) one does not need to choose the criterion for fully developed flow a priori. Trying different criteria is possible without repeating the computation.
3) one has a history of temporal evolution of specially averaged fields with negligible desk storage. This is useful for example in environmental applications.
The above is the motivation to integrate the postChannel utility with channelOodles solver, which pays back for flows subject to some homogeneous directions. The problem is how to transform postChannel (collapse fields) to work in parallel. Do you know any part of the code that by studying, I can learn how to do this. Thanks.
I thought to study patchIntern
I thought to study patchInternalField() and try to implement a slicedFvPatchField at every y level to do the averaging. Any suggestion? Does any body know where to start? Thanks.
|All times are GMT -4. The time now is 03:40.|