CAVITY_MAP Description:
This program maps a CFD surface solution in and around a cavity or gouge
on to an idealized cavity grid for comparison with other calculations.
The earlier Ensemble program assumed the cavities were rectangular, but
no such assumption is made here. Instead, an idealized rectilinear
form of the cavity is used as an ancillary input. This has four cavity
walls and floor at arbitrary angles, plus surrounding surface patches.
The rectilinear outer envelope comes from Boeing's Cavity Heating Tool.
Mapping the CFD solution to the Cavity Heating Tool grid turns out to be
what was really required here (not the mapping to a unit cavity done in
Ensemble), so that CFD fidelity can indicate if the engineering tool
is conservative or not. Such a mapping has been added as a further
output, without eliminating the unit cavity mapping of both the CFD and
outer envelope solutions. This version does more than interpolate at the
coarse mesh vertices: it averages the CFD soln. over the areas of each
coarse cell. It also generates slices of both the mapped/smoothed CFD
data and the input CHT data for line plot comparisons.
Strategy of original outer/inner envelope approach, & the alignment scheme:
> Impose CFD-type dense grids on the idealized surface patch edges.
This is the outer envelope for the actual damage.
> Construct a scaled-down, shallow form of the idealized walls and
floor, sitting on the smooth OML and centered sensibly. This acts
as an inner envelope for the cavity portion.
> Align the CFD surface grid with the working axes by calculating the
principal axes of the CFD cavity patches (masses <-> cell areas
acting at cell centroids lead to moments of inertia about the center
of mass; the eigenvectors of the inertia matrix form rows of the
needed rotation matrix).
N.B.: This version no longer uses cavity L, W, D to distinguish the
principal axes, because as the dimensions become less distinct, and
with sloping walls in the picture, the area-based moments of inertia
are less likely to make the distinctions reliably.
Instead, all possible permutations of the tentative rotation matrix
are applied, and the one that best transforms a third (derived)
reference point and the center of mass is chosen.
> For each line connecting corresponding points of the two sets of
idealized wall and floor patches, find the intersection with the
realigned CFD surface grid. The CFD solution at that intersection is
the desired mapped solution on a unit cavity, which is topologically
the same as the dense, idealized cavity and can be derived from it by
normalizing the arc lengths as (u,v) in the unit square and using u
and v for the non-constant x, y, or z appropriately.
> For the smooth OML patches downstream & adjacent to the unit cavity,
employ the ADT search and interpolation techniques used in Ensemble.
> There appears to be no point in retaining the option to perform
ensemble averaging of multiple solutions.
N.B.: This version also permits use of plain ADT searches for all
idealized patches. It appears that the inner/outer envelope method for
the cavity matches may not be necessary, and it suffers in places where
the intersection lines are almost parallel to the CFD surface. This
possibility forced suppression of possible line extrapolation during the
line/surface intersection calculations. Method 2 is now recommended.
Assumptions:
> Shuttle coordinates: X -> downstream, Y -> right wing, Z -> up
> The idealized outer envelope comes from the Cavity Heating Tool and
is in a standard 10-patch form. No assumptions are made about the
(i,j) convention, however.
> The x,y,z coordinates have the same units. A hidden scaling factor
applied to the CFD data is optional at the end of control file.
Control file format ('cavity_map.inp'):
Cavity_map control file
dplr_cavity_soln.dat ! CFD surface solution (Tecplot ASCII file)
1 ! Flow variable representing heating bump factor
1 10 33 ! CFD patch # and (i, j) of reference upstrm. pt.
8 10 33 ! CFD patch # and (i, j) of reference dnstrm. pt.
1.0 ! Reference heating value; < 0 => diagnostics
! List of CFD patches inside the cavity
1, 2, 10:25
heating_tool_grid.dat ! Idealized cavity grid (Tecplot ASCII)
2 ! 1 = inner/outer envelope method; 2 = plain ADT
[-0.0254 ! Optional scale factor applied to the CFD xyzs;
! negative means divide rather than multiply.]
Notes:
(1) The reference CFD points should be at the upstream and downstream
cavity lips near the mid-points. The line joining them after
realignment will be in the Z = 0 plane.
(2) The reference heating value allows flow values to be converted to
bump factors; use 1.0 if the flow values are already bump factors;
use < 0 to turn on output of intermediate grids and diagnostics
(3) The list of CFD cavity patches can use commas and colons.
(4) See the optional scale factor at the end of the control file above.
Input Surface Solution Format (Tecplot ASCII, structured, CFD & Idealized):
TITLE = ""
VARIABLES = "x" "y" "z" "BF" ["..."]
ZONE T="G1"
I=66, J=66, K=1, ZONETYPE=Ordered
DATAPACKING=BLOCK
DT=(DOUBLE DOUBLE ... (variable # of them) ... DOUBLE DOUBLE )
6.647333145E+00 6.575638294E+00 6.489704609E+00 6.390774727E+00 6...
: : : : :
POINT data packing is also permitted. One or more flow variables may be
present.
Output Results (Tecplot ASCII files):
(1) CFD solution mapped to a unit cavity with CFD-type grid density.
(2) CFD solution mapped to the plain outer envelope grid (CHT grid).
This was added belatedly as what was really wanted.
However, the realignment benefits from use of CFD-type resolution.
Results have been non-dimensionalized by the ideal cavity length.
(3) Two pairs of slices and dices of mapped/smoothed CFD data on the
CHT grid and of any functions input with the CHT grid.
The slice & dice abscissas are arc length, not X or Y, for better
display of results along vertical cavity walls.
Each cut is a separate zone.
The output formats are similar to the input formats (Tecplot ASCII).
The output file names are derived from the input CFD file name.
History:
05/13/05 D. Saunders Final version of Ensemble (rectangular cavities).
6/14:6/21 " " Initial adaptation as Cavity_Map.
06/22/05 " " Map the plain outer envelope (probably from the
cavity heating tool) by making sure the indexing
is in standard order then normalizing as for the
form used with the interpolated CFD solution.
06/29/05 " " The transpose of the rotation matrix should be
applied for realignment. This was hidden during
tests that mapped the outer envelope to itself.
07/01/05 " " Added the option to use the plain ADT mapping
method everywhere. Added mapping to the plain
outer envelope grid.
07/03/05 " " Introduced a second CFD reference point in order
to take out a misalignment that is inevitable if
the ideal cavity walls are not symmetric.
09/21/05 " " Steve Alter urged a control-volume averaging
approach for the interpolations to the coarse
Cavity Heating Tool mesh. This is now done by
dividing the coarse cells into many sub-cells.
10/19/05 " " Output line plots from the CHT grid (CFD + CHT).
10/29/05 " " Mapping a copy of the coarse, linear CHT data to
the CHT grid highlighted the lack of continuity
at patch edges caused by the area-based averaging
to cell centers. Therefore, force equality along
common edges by averaging in map_to_envelope.
11/02/05 " " Make arc length the abscissa for slices & dices.
This was a little awkward, requiring temporary
copies of slice X & Ys, etc.
02/06/06 " " Internal procedure deallocate_blocks is now a
part of Tecplot_io.f90.
02/16/06 " " Steve Alter's subscript check caught something in
in subroutine slice_patch. The slice arrays
passed from the higher level are big enough but
were wrongly dimensioned in the routine. Using
(*) is simpler than passing the right length.
04/27/06 " " Using the same slice & dice arrays for the CFD
and CHT solutions is bad inside the Tec I/O pkg.
if the numbers of functions differ. We need to
deallocate %q arrays and reallocate them if a
CHT solution is present.
05/02/06 " " Divide the dimensional outputs by cavity LENGTH.
06/01/06 " " Permute_R had a glitch affecting the case where
cavity W > L > D.
08/14/06 " " Revised to use Tecplot 360 version of the
Tecplot_io package.
08/24/06 " " The mapped.* file's variable names beyond 4
were not being copied from the CFD file.
07/26/07 " " Use of L, D, W to distinguish the principal axes
is not reliable if any two are close to each
other, and sloping walls add to the fuzziness of
the calculations. Instead, all 48 permutations
of the initial rotation matrix are now tried, and
the one that best transforms a third, derived
reference point is chosen.
An optional scale factor may be included at the
end of the control file. It is applied to the
CFD coordinates. If it is negative, it is used
to divide rather than multiply the coordinates.
07/29/07 " " Offset the derived reference point to help break
ties, and include the transformed center of mass
in the test for the best rotation.
07/30/07 " " Tabulate floor averages for 85/15% and 50/40/10%
splits.
07/31/07 " " Tabulate floor averages for the CHT solution too.
08/05/07 " " CFD uncertainties are now added according to the
Length/Depth ratio (floor and walls) or half the
excess over 1 (outside, after averaging). The
line plot files now contain an extra variable.
08/10/07 " " Bill Wood revised the uncertainty definitions:
walls and OML now use |BF - 1|. Also, the short/
long cavity cutoff changed from 10 to 15 L/D.
08/15/07 " " Before the principal axis calculations, align the
reference points with the X axis. This makes no
difference in some cases (suggesting consistency)
yet can help other cases (possibly with more of a
rotation about OX).
08/16/07 " " Determine preliminary and final rotations about
line joining reference points by 1D optimization.
08/20/07 " " Remove possible bad scaling of the CFD data by
shifting it to the cavity patch center first.
However, it doesn't help - moments of inertia
work with X - Xcg, etc. anyway.
Now that the third 2-D rotation can be found via
1-D optimization, just go with that now.
This does depend on the reference points' being
on a line parallel to the length axis.
08/08/13 " " All ADT variants are now in one module with
generic build_adt and search_adt interfaces.
Author: David Saunders, ELORET Corporation/NASA Ames Research Center, CA
Now with ERC, Inc. at NASA ARC.
Download Latest Version
geomlib.tar.gz (58.3 kB)
