The objective of this portion
of the research is to evolve the
computational aspects of CFD to the point where the codes can
be used as a numerical test bed (i.e., in place of an
experiment) during preliminary control testing and validation.
In order to meet this objective, current code(s) must be
modified to provide time accurate results with realistic
boundary conditions and reasonable execution times. The NPARC
code (both two- and three-dimensional versions) is being used
as a vehicle for developing this technology. To date, NPARC has
been modified to provide time accurate results and an improved
compressor face boundary condition. Capabilities have also been
added that allow boundary conditions to be specified as a
function of time.
Current work is focused on developing parallel versions of the
new time accurate codes. The parallelization of PARC3D is a
NASA Glenn in-house effort. A speedup of 5.6 over a single Cray
Y-MP processor has been achieved using 16 CPUs (model 590)
which are part of the NASA Glenn IBM RS6000 workstation
cluster. 2-D NPARC is being parallelized under a grant with
Indiana University Purdue University at Indianapolis (IUPUI).
Preliminary data show speedup ratios similar to 3-D NPARC.
The accompanying figure shows the result of inviscid
calculations during an inlet unstart transient. These time
accurate 2-D NPARC calculations were performed using the
variable-diameter-centerbody, mixed-compression inlet with
freestream conditions of Mach 2.5. To induce the unstart, a
step increase in freestream temperature is applied at time,
T=0.0, while holding the compressor face average Mach number
constant at 0.29. The calculations, generated by Dr. J.K.
Chung, took 300 CPU seconds on a single Cray Y-MP processor
using a 3.555x10-6 second time step.
Kevin J. Melcher
phone: (216) 433-3743