| |||||||||||||||
|
|
![]() ![]() Re-Distributed by http://www.tpub.com
nacelle. While this paper concentrates on nacelle design, it is important to note that the implementations
are sufficiently general to allow for the redesign of any other part of the airplane.
TLNS3D-MB
CDISC is incorporated as a system of subroutines called by the main TLNS3D-MB driver. The CDISC
routines are only called when the residual has dropped to a value chosen by the user. Once called,
CDISC first extracts all relevant volume grids for the cowl surface from the grid array. These volume
grids are then oriented to place the nacelle centerline on the x-axis. The volume grid is now on a sort of
"virtual" workbench on which the design changes for the current design cycle can be easily performed.
At each chosen streamwise design station, positioned circumferentially around the nacelle, the geometry
is changed to offset the discrepancy between the analysis and the target pressures. The portions of the
nacelle between target stations are changed in a linear fashion. Finally, circumferential smoothing is
applied as desired. The volume grids are then perturbed to accommodate this new nacelle surface.
Finally, the new grids are restored to their proper orientation and are placed back into the grid array for
further analysis. This new analysis proceeds until a user specified drop in the residual between design
iterations is achieved, after which a new design cycle is begun.
The TLNS3D-MB solver, used herein, incorporates mesh sequencing and multigrid, but requires
point-matched grids. The code allows a choice of turbulence models. The Spalart-Allmaras one-equation
model14 was used for the examples in this paper.
One limitation in the above approach is the difficulty in constructing point-matched multi-block grids
for complex configurations. For this reason, the overset grid approach is more attractive although there
might be reservations among some in the aerospace community regarding its accuracy, particularly in
regard to flux conservation. It should be noted, however, that most patched grid multi-block codes use
non-conservative interpolation for block interfaces as well.
OVERFLOW
Except for the process of monitoring when to start and stop analysis and design, the entire CDISC
system is largely run as a post-processor outside OVERFLOW. Control passes out of OVERFLOW
whenever a user specified criterion (typically, the residual drop on the grid involving the component to
be designed) is reached. Once out of OVERFLOW, the volume grid and restart files are used to compute
the pressure on the fan cowl. The relevant grids are then oriented on to the "virtual" workbench
described above and the CDISC geometry modifier is invoked. The grid is then automatically adjusted
to accommodate the geometry change. The cowl grids are then restored to their proper orientation.
Next, all relevant grids are re-projected and recomputed. For example, the pylon might have to be
reprojected back onto the new nacelle surface. Next the pylon grid will have to be restretched and the
nacelle fringe points re-projected back onto the pylon grid. Finally, the new grid system must be run
through the mesh interpolation program, PEGSUS15, in order to update the interpolation stencils
between the various grids, before OVERFLOW can be restarted for analysis of the new geometry. The
analysis continues until the next residual drop is achieved.
The OVERFLOW code also allows a choice of turbulence models. The Baldwin-Barth model16 was
used for the examples given in this paper.
Results
|
|
Privacy Statement - Press Release - Copyright Information. - Contact Us |