What would be the best starting point to start studying an example of
folded script looking at the problem of the interaction of a
detonation wave and/or an inert shock wave interaction with a ramp?
JJQ's vki second set of notes deal with such an example at some
length, but I wonder if the procedure and scripts have been improved
since then, and/or ported to the reactive case?

In a related way, the generation of html help files and generation of
mpeg movies detailed in the vki notes do not seem to work with my
installation of amritav3.00 on a 32bit fc7 installation.  is that due
to some incompatibility with the older scripts, or simply an
installation issue on my part?  When i installed it, all the
components seemed to work well, except some serious issues with the
GIMP; gimp-perl is not available for my system.

Matei,

AMRITA+AMR_SOL's operation is essentially independent of the EquationSet
and so if you have a ramp-script that works for the plain EulerEquations
it will automatically work with the ReactiveEulerEquations i.e there is
no porting in the classical sense. Now if you take a look at AMRITA's
big picture:

    http://www.amrita-cfd.org/html/screenshots

you will see on the notice board (at about 11 o'clock) a single mach
reflection for a detonation wave interacting with a ramp. That particular
calculation, done back in 1993, used a body-fitted grid which is
set up in the fashion decribed in the VKI notes.

There is, however, an alternative approach you can follow which
uses a level-set technique on a Cartesian mesh. The examples:

    amrcp latex/Chp2/method.3
    amrcp latex/Chp2/method.4

show it working in the inert case. But you will need to fix
the following bug in AMRITAv3.00 before you can run them.
Hand-edit the file:

     $AMRITA/Amrita/keywords/def/EquationSet/amr_sol.pl

After line 54:

      $amrita->set_thistoken('amr_sol::NEQN',            0);

insert:

      my $frame = "$STACK{$PROCnamebase}_";

This bug crept in when I refactored amr_sol so as to make it
easier to maintain.

The -html option described in the VKI noted has been deprecated.

I would need to check up on the mpeg problem, but I know Hans
has been generating movies and so I guess it could be a local problem
at your end.

The gimp-perl interface is used to craft buttons and the like.
It's not needed to be able to run any of the examples in the VKI
notes, but I can always put together a tarfile containing
the necessary Perl modules.

James

James and I looked at a related problem of inert shocks and
detonations
driven by constant velocity wedges using this technique a couple of
years ago. One note of caution is that we found some solver dependent
issues - in particular the roe schemes didn't tend to the correct
shock/detonation angles in the far field. The solver that worked best
was the lax-friedrichs (W)ENO scheme, for which I think the level
set stuff was originally designed. As I recall James found there
was some "leakage" of info through the level set surface for the Roe
case.
This is all anecdotal, my only point in mentioning it is that if you
go down this
route this is a case where you definetly need to do solver studies
which amr_sol is designed for. I believe I still have the scripts
we used for this, which I can send to Matei if required, but
I can't vouch that they are well polished.
Cheers
Gary.

Gary,

I'm guessing that the entropy-fix which is applied to the
contact wave to suppress the carbuncle phenomenon is the
culprit here.

James

James, Gary,
Thanks for the information.  Are the problems mentioned by Gary solely
associated with the level set method?

In the body fitted grid examples of the vki notes, would there be any
problems with the skewed grids?  I assume that the more skewed the
grids become, the less accurate becomes the estimate of the fluxes at
the cells interface, am I right?  For example, if one extends the
technique to simulate flow over a sphere, the cells near the front and
back of the cyclinder would be highly skewed and large errors would
ensue.  In that case, I guess the level set method would be the better
choice.  Right? The reason why I am asking is that I would also like
to look at shock/detonation problems interaction with cylinders.

matei

Matei,

I would not worry too much about Gary's observations for now,
as setting up a script for a physical flow problem is largely
independent of the choice of patch-integrator. Thus you can
always try out multiple schemes on the one problem so as to
gauge the sensitivity of the observed results to the details
of the numerical method.

Yes, one could say that.

Yes and no. The level set method is inferior to the body-fitted
grid right at the cylinder's surface, but the Cartesian grid would
win out away from the body. So it's a case of "swings and roundabouts"
as we say in the UK. Anyhow, if you take a look in my thesis,
you'll see that a polar grid can give quite good results for
a shock wave interacting with a cyclinder. But I expect it would
not work so well for a detonation. But the beauty of AMRITA is
that once you're up to speed, you'll easily be able to try out
both approaches and so there is no need to feel compelled
to stick with one or the other.