[Nek5000-users] Code output

Wed Mar 3 07:03:34 CST 2010

Mani,

                                         final         inital      target
                                      residual        residual    residual

>     499270    Hmholtz VELX:      9   2.9703E-11   1.3708E+07   3.0000E-10

which implies that you are driving the residual in the viscous substep
down by 1.e18 in the iterative solver.

Normally, I would control Reynolds number by reducing viscosity and
keeping the velocity to be 1.0 --- In that way, your time scale is
~1.0 (assuming the domain size to also be roughly unit size).

In Nek5k, we support this by allowing the use to specify Reynolds
number directly by setting param2 in the .rea file to be a negative
number.   For example, assume your box size is one and your lid
speed is one and you would like Re=60,000.  You would set param 2
to be -60000.   The result is that nek5000 immediately sets param 2
to 1./60000 when it reads the value.   Same goes for conductivity
(param 8).

However, if you prefer to work in other unit systems (e.g., by
changing the velocity), that is also fine.   You might in that
case find it useful to set params 21 (DIVERGENCE) and 22 (HELMHOLTZ)
to 0.  Then set TOLREL (param 24) to, say, .01 or .001.

Zeroing p21 and p21 turns off the absolute tolerances, and lets
Nek5000 compute the proper tolerances through tolrel (p24).
In principle, the values of tolrel I specify above shoud be
plenty adequate -- nek5000 generates its own tolerances inside
for each solver according to problem specifications and eigenvalue
estimates.  I find it tends to be a bit too convservative, resulting
in tighter tolerances than necessary (but not 1.e-18), so generally
favor the approach above based on convective time scales.

Hope this helps.

Paul

On Wed, 3 Mar 2010, nek5000-users at lists.mcs.anl.gov wrote:

> Hi,
>
>    I'm trying to solve for a lid driven cavity flow (in 2D for now) using 
> NEK5000. The viscosity has been set to one and the lid is moving with a 
> velocity of 10000. I don't understand the meaning of the numbers given in the 
> code output. For ex:
>
> Step 499268, t= 6.8040858E-02, DT= 1.2825985E-07, C=  0.409 5.9000E+04 
> 9.7954E-02
>             Solving for fluid
>     499268    Hmholtz VELX:      9   2.9714E-11   1.3717E+07   3.0000E-10
>     499268    Hmholtz VELY:      9   1.3165E-11   1.1815E+07   3.0000E-10
>    499268 U-Press std. :     91   2.9185E-08   3.0000E-08   2.6864E-03 
> 8.4799E-02
>      499268  DNORM, DIVEX  2.918545176353094E-008  2.918487500510926E-008
>     499268   6.8041E-02  9.2034E-02 Fluid done
> Step 499269, t= 6.8040986E-02, DT= 1.2825985E-07, C=  0.409 5.9000E+04 
> 9.9300E-02
>             Solving for fluid
>     499269    Hmholtz VELX:      9   2.9708E-11   1.3713E+07   3.0000E-10
>     499269    Hmholtz VELY:      9   1.3158E-11   1.1809E+07   3.0000E-10
>    499269 U-Press std. :     91   2.9192E-08   3.0000E-08   2.6781E-03 
> 8.6898E-02
>      499269  DNORM, DIVEX  2.919249585636570E-008  2.919235717748844E-008
>     499269   6.8041E-02  9.4083E-02 Fluid done
> Step 499270, t= 6.8041114E-02, DT= 1.2825985E-07, C=  0.409 5.9000E+04 
> 1.0086E-01
>             Solving for fluid
>     499270    Hmholtz VELX:      9   2.9703E-11   1.3708E+07   3.0000E-10
>     499270    Hmholtz VELY:      9   1.3151E-11   1.1802E+07   3.0000E-10
>    499270 U-Press std. :     91   2.9200E-08   3.0000E-08   2.6698E-03 
> 8.3875E-02
>      499270  DNORM, DIVEX  2.919940124964188E-008  2.920027044061877E-008
>     499270   6.8041E-02  9.1076E-02 Fluid done
>
> It would be great if someone could explain what the numbers mean. I'm worried 
> about the numbers in the second column being of the order of 1E+07. Is that a 
> problem? Also while specifying the boundary conditions, how can one specify 
> user set boundary conditions for more than one boundary using userbc? Is 
> there a parameter which one could check to see if the grid resolution if 
> enough for the case there is being run or should we resort to checking for 
> grid independence?
>
> The rea, map and usr files are attached.
>
> Thanking you,
> Mani chandra
>
> P.S I'd really like to thank the developers for releasing this software for 
> free and for documenting it so as to spread it's usage.
>