Symmetry breaks quickly in simple flow

[mcgratta]

Run this case below. It only takes a few seconds. The resulting flow should be symmetric and isothermal, but it is neither. It is worst when CONSTANT_SPECIFIC_HEAT_RATIO=T, but it still fails if is F. If I remove one of the VENTs injecting the gas, the case remains iso-thermal, but obviously not symmetric.

&HEAD CHID='sym_test' /

&MESH IJK=64,1,16, XB=0.0,3.2,-0.001,0.001,0.0,0.8 /

&TIME T_END=8.0 /

&PRES CHECK_POISSON=T /

&MISC STRATIFICATION=F, NOISE=F, CONSTANT_SPECIFIC_HEAT_RATIO=T /
&RADI RADIATION=F /

&SPEC ID='GAS BG', MW=29, BACKGROUND=T /
&SPEC ID='GAS A', MW=29 /
&SPEC ID='GAS B', MW=29 /

&SURF ID='WALL', FREE_SLIP=T, DEFAULT=.TRUE. /

&SURF ID='BLOW A', VEL=-0.5, MASS_FRACTION(1)=1., SPEC_ID(1)='GAS A' /
&SURF ID='BLOW B', VEL=-0.5, MASS_FRACTION(1)=1., SPEC_ID(1)='GAS B' /

&VENT XB=0.0,0.0,-0.001,0.001,0.0,0.4, SURF_ID='BLOW A' /
&VENT XB=0.0,0.0,-0.001,0.001,0.4,0.8, SURF_ID='BLOW B' /
&VENT PBX=3.2, SURF_ID='OPEN' /

&SLCF PBY=0.,QUANTITY='MASS FRACTION', SPEC_ID='GAS A',  CELL_CENTERED=T, VECTOR=T /
&SLCF PBY=0.,QUANTITY='MASS FRACTION', SPEC_ID='GAS B',  CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='MASS FRACTION', SPEC_ID='GAS BG', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='TEMPERATURE', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='DIVERGENCE',  CELL_CENTERED=T /

&DEVC ID='T_max', XB=0.0,3.2,-0.001,0.001,0.0,0.8, QUANTITY='TEMPERATURE', SPATIAL_STATISTIC='MAX', TEMPORAL_STATISTIC='INSTANT VALUE' /
&DEVC ID='T_min', XB=0.0,3.2,-0.001,0.001,0.0,0.8, QUANTITY='TEMPERATURE', SPATIAL_STATISTIC='MIN', TEMPORAL_STATISTIC='INSTANT VALUE' /

&DUMP DT_DEVC=0.00001 /

&TAIL /

[drjfloyd]

With a quick look. The .out file show these all have the same properties and if I add slices for specific heat, conductivity, etc. Those show identical values until temperature starts to change in which case conductivity increases with temperature. This doesn't seem to be something wrong with the functions for getting properties.

[mcgratta]

I don't think so either. I'm just surprised at how quickly the symmetry and iso-thermal break down. Like in 3 time steps. No time step changes, no changes in geom, no obsts, no mesh boundaries.

[drjfloyd]

It also doesn't happen if one of the VENTs is the background species or both vents are the same species. Only when the two vents are A and B

[mcgratta]

Nice clue.

[mcgratta]

How did you apply the BG species at the VENT. I get

ERROR(330): SURF BLOW B cannot use background species for MASS_FRACTION. (CHID: sym_test)

[mcgratta]

Nevermind. I know how.

[mcgratta]

Setting GVEC=0,0,0 maintains symmetry even though I would think we would not have to set that to do so.

[mcgratta]

When I run the case longer, the initial slop goes away. There seems to be something about the ramp up or initial/boundary conditions.

[drjfloyd]

It is still seen with just a 2 cell wide domain but not as strong

&HEAD CHID='test' /

&MESH IJK=8,1,4, XB=0.0,0.8,-0.001,0.001,0.2,0.6 /

&TIME T_END=2,DT=0.01 /

&PRES CHECK_POISSON=T /

&MISC STRATIFICATION=F, NOISE=F, CONSTANT_SPECIFIC_HEAT_RATIO=T /
&RADI RADIATION=F /

&SPEC ID='GAS BG', MW=29, BACKGROUND=T /
&SPEC ID='GAS A', MW=29 /
&SPEC ID='GAS B', MW=29 /

&SURF ID='WALL', FREE_SLIP=T, DEFAULT=.TRUE. /

&SURF ID='BLOW A', VEL=-0.5, MASS_FRACTION(1)=1., SPEC_ID(1)='GAS A' /
&SURF ID='BLOW B', VEL=-0.5, MASS_FRACTION(1)=1., SPEC_ID(1)='GAS B' /

&VENT XB=0.0,0.0,-0.001,0.001,0.3,0.4, SURF_ID='BLOW A' /
&VENT XB=0.0,0.0,-0.001,0.001,0.4,0.5, SURF_ID='BLOW B' /

&OBST XB=0,0.8,-0.001,0.001,0.2,0.3/
&OBST XB=0,0.8,-0.001,0.001,0.5,0.6/

&VENT PBX=0.8, SURF_ID='OPEN' /

&SLCF PBY=0.,QUANTITY='MASS FRACTION', SPEC_ID='GAS A', CELL_CENTERED=T, VECTOR=T /
&SLCF PBY=0.,QUANTITY='MASS FRACTION', SPEC_ID='GAS B', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='MASS FRACTION', SPEC_ID='GAS BG', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='TEMPERATURE', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='DIVERGENCE', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='CONDUCTIVITY', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='SPECIFIC HEAT', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='DENSITY', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='VISCOSITY', CELL_CENTERED=T /
&SLCF PBY=0.,QUANTITY='SPECIFIC ENTHALPY', CELL_CENTERED=T /

&TAIL/

[mcgratta]

If you set FLUX_LIMITER='GODUNOV' on the MISC line, the temperatures will be exactly 20 C. I worked my case above every which way, and it seems to me that the flux limiters other than CENTRAL and GODUNOV all create a small error. CENTRAL will eventually go "unstable" in the sense that tiny errors will grow and throw off the isothermal flow. I assume that the two simple limiters work because they act linearly when one sums the species equations, but I haven't nailed down a slam dunk reason.

[mcgratta]

On a related note, I just added a case called Pressure_Solver/obst_activation_default_gases.fds which is what this issue is based on. In this case, there are multiple meshes and obstructions that are crated and removed at mesh boundaries. I set the flux limiter to GODUNOV because, for the moment, I am trying to fix a small discrepancy when obstructions are created or removed at mesh boundaries in the presence of mixing gases. I think that this case be dealt with separately form the flux limiter issue.