Non-symmetric solution in WMLES of square duct

[soaringxmc]

Hi @federicoVerolini

Please post your issue here!
Please provide an input.nml for a minimal example to reproduce your issue. Also, attach necessary description to reproduce the issue.

[federicoVerolini]

input.nml file:

&dns
ng(1:3) = 512, 128, 128
l(1:3) = 12.8, 2., 2.
gtype = 1, gr = 0.
cfl = 0.95, dtmin = 1.e5
visci = 42000.
inivel = 'duc'
is_wallturb = T
nstep = 100000, time_max = 4970., tw_max = 0.1
stop_type(1:3) = F, T, F
restart = T, is_overwrite_save = T, nsaves_max = 0
icheck = 10, iout0d = 10, iout1d = 100, iout2d = 100000, iout3d = 50000, isave = 1000
cbcvel(0:1,1:3,1) = 'P','P', 'D','D', 'D','D'
cbcvel(0:1,1:3,2) = 'P','P', 'D','D', 'D','D'
cbcvel(0:1,1:3,3) = 'P','P', 'D','D', 'D','D'
cbcpre(0:1,1:3) = 'P','P', 'N','N', 'N','N'
cbcsgs(0:1,1:3) = 'P','P', 'D','D', 'D','D'
bcvel(0:1,1:3,1) = 0.,0., 0.,0., 0.,0.
bcvel(0:1,1:3,2) = 0.,0., 0.,0., 0.,0.
bcvel(0:1,1:3,3) = 0.,0., 0.,0., 0.,0.
bcpre(0:1,1:3) = 0.,0., 0.,0., 0.,0.
bcsgs(0:1,1:3) = 0.,0., 0.,0., 0.,0.
bforce(1:3) = 0., 0., 0.
is_forced(1:3) = T, F, F
velf(1:3) = 1., 0., 0.
dims(1:2) = 0, 0
\

&les
sgstype = 'dsmag'
lwm(0:1,1:3) = 0,0, 1,1, 1,1
hwm = 0.1
\

&cudecomp
cudecomp_t_comm_backend = 0, cudecomp_is_t_enable_nccl = T, cudecomp_is_t_enable_nvshmem = T
cudecomp_h_comm_backend = 0, cudecomp_is_h_enable_nccl = T, cudecomp_is_h_enable_nvshmem = T
\

output of the code: streamwise (x) averaged section of the square duct computed by the subroutine out2d_duct(fname,ng,lo,hi,idir,l,dl,z_g,u,v,w).

time averaging interval: 200 -> 4970 time units, saving output files every 100 time-steps.

analyzed fields: mean streamwise velocity contours and secondary motions streamfunction contours.

issue: the average flow over the square duct section is symmetric over each of the four quadrants, but is not symmetric on every octant of the square duct, i.e. taking a corner bisector, the flow is not symmetric with respect to it.

[soaringxmc]

Hi Federico

Try current setting but with static smag model.

Then, you could try these step by step to find out the reason. We should use a steady laminar case for debugging.
laminar (say, reb = 1000?), dns
laminar wrles
laminar wmles, but no sgs
laminar wmles

Just use fixed matching location should be enough.

[soaringxmc]

It seems that the non-symmetry issue is caused by the dynamic Smagorinsky model. I would suggest further checking the quantities used in the dsmag routine (in sgs.f90) to identify which of them might be non-symmetric—particularly those obtained through test filtering.

You may refer to the CaLES theory manual on overleaf for the specific formulation of the dynamic Smagorinsky model.

[soaringxmc]

Any update on the debugging?

[federicoVerolini]

The issue was related to the sgs.f90 module, in particular to the subroutine cmpt_sgs(), case('dsmag').
The code automatically assumes channel flow case if not manually switched to duct flow case (#define _CHANNEL instead of manually switching to #define _DUCT). Consequently, when it dynamically computes the Cs coefficient, performs averages in both streamwise and spanwise directions, while in the duct case only homogeneous streamwise direction must be considered. This resulted in an asymmetry of the velocity field with respect to corners bisectors.
I have modified my sgs.f90 file in order to automatically detect the kind of flow simulated, depending on the wall boundaries imposed. I have performed some test with duct flows and now the velocity field results to be symmetric.
Thank you for your suggestions, hope this answer may be useful.

[soaringxmc]

Ah...I have assumed you were using #define _DUCT for both sgs and output, so no bug is found in dsmag

[federicoVerolini]

Happily, not. In order to avoid any forgetfulness of manually switching between #define _CHANNEL and #define _DUCT, I have implemented in my fork of the code in sgs.f90 an automatic procedure in order to detect the flow case depending on the imposed walls.