Current outputs for CONVECT and GASP
These are for the purposes of showing current results to my supervisor.
S.McMahon 1 May 1999.
GASP run
-
This was run on orac, my Pentium II 450.
-
113x113 grid points
-
lower boundary is tangency (slip), upper is no-slip, adiabatic
-
turbulent viscosity but no cross-terms
-
the CPU time was 73 hours.
Velocity
contours
Mass fraction of driver gas contours
Discussion
-
Seems to be no test gas slug - ie. shock and contact surface are coincident
-
definitely no driver gas jetting seen
-
many features similar to CONVECT output
Small CONVECT run
-
This was done on four Sun Ultra 1 workstations in the School of Chemistry
-
100x50 cells
-
after shock reflection
-
the run time was about 54 hours.
Velocity contours
Discussion
The reflected shock results look about what is expected except for feature
appearing at the symmetric boundary. This says that there is some
problem with the symmetric boundary code.
Large CONVECT run
-
This was done on zetes on four processors
-
150x250 cells
-
no accurate timings available - about two weeks wall time.
Velocity contours
Velocity
surface plot
Discussion
The bigger problem went for much longer than would be expected from simply
computing over more cells. I suspect that somehow smaller time steps
were causing further instabilities which in turn kept the time step small.
Problems with the symmetric boundary are even more evident. The 'build-up'
of velocity at the shock seems similar at both upper and lower boundaries.
With artificial viscosity fix
-
Standard 10 x 0.038 meter shock tube problem with two species and "correct"
Reynolds numbers
-
Added a more restrictive CFL condition which takes into account instabilities
from high cell aspect ratios.
-
Added artificial viscosity correction based on the form of an artificial
viscosity term which arises in the Lax-Wendroff method.
Discussion
The boundary layer is much more evident and about the right size (exact
matching not done at the moment). These results are after 18 hours
calculation on four processors of zetes.
Symmetric boundary fix