Hi Bill,
thanks for the suggestion for parallel, I'll test in the next days the
multigrid parallel solver.
I go a moment back to an old question on userDefinedForcing. Following the
examples given, I added an option to add gravity and an additional viscosity
term proportional to the velocity.
The Forcing I give is
fg(I1,I2,I3,uc) = - myParameter(1)*u(I1,I2,I3,uc);
fg(I1,I2,I3,vc) = myParameter(0) - myParameter(1)*u(I1,I2,I3,vc);
i.e. a friction in both x and y direction and a volume force like gravity in y
direction.
My flow goes in the up-down direction in a "free channel": assigning noSlipWall
conditions on the left-right boundaries, it exists a limit velocity v_lim that
can also be found theoretically. I choose the parameters so that v_lim = -2.0
and I setup an uniform incoming flow with v = -1.0. Note that I actually had to
insert accelerations in fg(all,all,all,n) and
not forces (e.g., for the gravity I insert g and NOT \rho*g )
The velocity field behaves perfectly, I see a transient form the initial
velocity to the terminal velocity which fits perfectly with the theoretical
results.
But the pressure behavior seems really bad to me. It starts from values like
(pmin=1, pmax=2) and goes down to (pmin= - 80, pmax = - 45) after 4000
iterations, and still has the tendency to decrease, and, more important, there
are discontinuities in its evolution, (I've attached a graph).
Have I missed something? My only doubt is, does Overture automatically evaluate
the divergence of the Forcing that has to be added to the Pressure Poisson
equation? (In my case anyway it should be zero since I'm dealing with
incompressible flows and \/u =0 ...)
Thanks for any clue
Regards
Alessandro
Date: Fri, 1 Jun 2012 07:42:39 -0700
From: henshaw@xxxxxxxx
To: overture@xxxxxxxxxxxxx
Subject: [overture] Re: modify ins equations
Hi Alessandro,
userDefinedForcing assigns the forcing 1 grid at a time.
If ucg is a realCompositeGridFunction, then ucg[grid]
is a realMappedGridFunction.
Note that u(I1,I2,I3,uc) is an array (not a scalar).
You should use something like
fg(I1,I2,I3,uc) = myParameters(3)*u(I1,I2,I3,uc);
fg(I1,I2,I3,vc) = myParameters(4)*u(I1,I2,I3,vc);
etc.
You will want to check that the forcing really works. As I indicated
we have been making many changes to the forcings. These
will appear in the next version.
...Bill
On 06/01/2012 02:45 AM, Alessandro Orchini wrote:
Hi Bill,
thanks for the suggestions, I'm working on the code.
I'm just not sure how to behave with the velocities: in
userDefinedForcing the velocity field is already read from the
db by
const int & uc = parameters.dbase.get<int >("uc");
const int & vc = parameters.dbase.get<int >("vc");
const int & wc = parameters.dbase.get<int >("wc");
If I define a
real TheVelocity
and assign it component by component by
for(n==0; n<numberOfDimensions; n++)
{
if(n==0)
TheVelocity=u(I1,I2,I3,uc);
else if(n==1)
TheVelocity=u(I1,I2,I3,vc);
else if(n==2)
TheVelocity=u(I1,I2,I3,wc);
fg(I1,I2,I3,n)=myParameters(3)*TheVelocity;
}
does this routine apply the forcing on all the grids? Usually
when I read a .show file I specify the grids with square
brackets, e.g.
u[grid](i,j,k,n)
and I wonder what if the [grid] option is omitted.
Thanks!
Regards,
Alessandro
Date: Mon, 21 May 2012 09:55:03 -0700
From: henshaw@xxxxxxxx
To: overture@xxxxxxxxxxxxx
Subject: [overture] Re: modify ins equations
Hi Alessandro,
Forcings can be specified in
cg/common/src/userDefinedForcing.C (we have been
making many changes to the forcing's which will appear in the
next version).
Gravity can already be turned on if you solve the INS plus
Boussinesq equations (i.e.
add a Temperature equation) See the examples in the
cg/ins/cmd.
Boundary values are set in
cg/common/src/userDefinedBoundaryValues.C
...Bill
On 05/21/2012 08:17 AM, Alessandro Orchini wrote:
Hi all,
I would like to add forces at the INS equation (a volume
force as gravity, and also a force proportional to the
velocity filed),
and define new boundary conditions. In particular, I need
to impose at the inflow and outflow boundaries (top and
bottom respectively) the velocity profiles to be
u(x,t)=0;
v(x,t) = a*(cosh(b*x)/cosh(b*c) - 1);
being a,b,c constants.
I've found in ins/src the whole source code, but what and
where I have to modify exactly? (lineSolveINS.h? Cgins.C?
ins(bC).C? ...)
Thanks,
Regards
Alessandro
Attachment:
Pressure.eps
Description: image/eps
