[overture] Re: low massive rigid bodies
- From: Bill Henshaw <henshaw@xxxxxxxx>
- To: "overture@xxxxxxxxxxxxx" <overture@xxxxxxxxxxxxx>
- Date: Fri, 02 Dec 2011 15:01:53 -0800
Hi Alessandro,
This is a known problem that the standard algorithm fails with light
rigid bodies.
In cg.v24 there is an initial attempt to fix this problem. You can look
at the
cgins script cg/ins/cmd/oneDrop.cmd and use for example
cgins oneDrop -g=oneDrop1.hdf -tf=2. -tp=.01 -bodyDensity=.1
-relaxRigidBody=1 -alpha=.1 -nc=10
where the grid "oneDrop1.hdf" can be made with the attached ogen script.
If you run this
example you will see that the number of correction steps is large (7-10)
at the start but then
things settle down to 2-3 corrections.
...Bill
Alessandro Orchini wrote:
Hi,
I'm having a problem with 2D simulations involving a low massive thin
flat plate in a uniform flow. I think that since the density of the
rigid body is not much higher than the one of the fluid, there's a
singularity in the equation of motion: with a low inertia there are
huge forces on the body, which tries to assume the flow velocity in a
small time.
I assume a possible solution would be to choose time-steps and t_max
lower and lower, but it would take an amount of time to run the entire
simulation.
Furthermore, i've seen that if I constrain the motion in 1-D along and
do not modify any other parameter, the simulation works fine.
I wonder if there is another possible solution for this kind of
problems. I attach both cgins and ogen .cmd files in which are
contained all the parameters I'm using.
Thanks for the help
Alessandro.
*
* one drop in a channel
*
***************************************************************************
* scale number of grid points in each direction by the following factor
* ***NOTE: restore file to factor=1 for regression tests *****
* $factor=.5; $name = "oneDrop0.hdf"; $mgLevels=2;
$factor=1; $name = "oneDrop1.hdf"; $mgLevels=2;
* $factor=2; $name = "oneDrop2.hdf"; $mgLevels=3;
* $factor=4; $name = "oneDrop4.hdf"; $mgLevels=3;
* $factor=8; $name = "oneDrop8.hdf"; $mgLevels=4;
printf(" factor=$factor, mgLevels=$mgLevels\n");
*
*-----
$mgFactor=2**$mgLevels;
*-----
*
* Define a subroutine to convert the number of grid points
sub getGridPoints\
{ local($n1,$n2)=@_; \
$nx=int(($n1-1)*$factor+1.5); $ny=int(($n2-1)*$factor+1.5); \
$nx=int( int(($nx-1)/$mgFactor)*$mgFactor+1.5); if( $nx==1 ){
$nx=int($mgFactor+1.5); } \
$ny=int( int(($ny-1)/$mgFactor)*$mgFactor+1.5); if( $ny==1 ){
$ny=int($mgFactor+1.5); }\
}
*
***************************************************************************
create mappings
*
rectangle
# $xa =-1.; $xb=1.; $ya=-3.; $yb=1.;
$xa =-1.; $xb=1.; $ya=-1.; $yb=1.;
set corners
$xa $xb $ya $yb
lines
$nx = int( 25*($xb-$xa)+1.5); $ny=int( 25*($yb-$ya)+1.5);
getGridPoints($nx,$ny);
$nx $ny
boundary conditions
1 1 1 1
mappingName
channel
exit
*
Annulus
lines
$nya=9; # fix radial lines since outer radius is reduced
$nymg=int( int(($nya+$mgFactor-1)/$mgFactor)*$mgFactor+1.5);
getGridPoints(57,$nymg);
$nx $nymg
inner and outer radii
$innerRadius=.25;
$outerRadius=$innerRadius+(.25/$factor)*($nymg/$nya);
$innerRadius $outerRadius
centre for annulus
0. 0.
boundary conditions
-1 -1 1 0
mappingName
drop-unstretched
exit
*
stretch coordinates
Stretch r2:itanh
STP:stretch r2 itanh: layer 0 1 4 0 (id>=0,weight,exponent,position)
stretch grid
mappingName
drop
exit
*
*
exit
generate an overlapping grid
channel
drop
done
change parameters
ghost points
all
2 2 2 2 2 2
exit
* display intermediate results
compute overlap
* pause
exit
*
save an overlapping grid
$name
oneDrop
exit
Other related posts:
