!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Description of Program SLOS:
SLOS has been adapted from USLOS (unstructured [grid] lines of sight),
itself an adaptation of LINES_OF_SIGHT and HEMISPHERES_OF_SIGHT, which
are now superseded by this combination of the two purposes (body-normal
lines for tangent-slab radiation calculations, or hemispherical lines
for just one body point at a time) for the 3D structured grid case.
(See also LINES_OF_SIGHT_2D.)
The volume grid is expected to be right-handed, and for the hemisphere
case, the y >= 0 half is expected (with x pointing downstream). An off-
center body point requires both halves for the hemisphere case.
The inner surface is expected to represent a mostly-convex surface. Any
concavities may cause difficulties with hemisphere lines that encounter
the body. Ancillary utility ADJUST_HEMISPHERE_LOS is the initial answer
for such cases.
Body Point Inputs (one entry per line; trailing comments ignored):
Body-normal (tangent-slab radiation calculation) case:
0. 0. 0.
1.628659741950E-01 5.741545696080E-01 0. ! Nose/cone tangency
8.985316209400E-01 1.837811469800E+00 0. ! Cone/shoulder tangency
: : :
If only one body point is detected, a prompt still allows for the
tangent-slab case.
Hemispherical lines (full angular integration) case:
1.0742 -1.81 1.083 [10.] ! x, y, z[, cone angle]
Only one body point is handled for this case. The optional semi-cone
angle allows for writing fewer than the full hemisphere's worth of
lines to avoid unnecessary radiation cone angles outside the viewing
angle of a radiometer instrument. Note that NEQAIR_INTEGRATION can
still integrate over a narrower angle than the indicated data angle.
If the optional angle is omitted, it is taken to be 90 degrees.
Input Structured Volume Grid (PLOT3D Format):
Formatted or unformatted multiblock files are handled, and a single layer
of grid blocks is assumed (normally the case for hypersonic vehicles) so
that the outer surface of all blocks serves as the shock-aligned boundary
with which intersections of lines from the body point(s) are calculated.
Outputs:
While the precursors of SLOS discretize the lines of sight with the same
relative point distributions as found in the nearest off-body grid line,
this variant follows the USLOS approach of using DPLR-type controls nk,
d1, and ds2_fraction that are prompted for.
The lines of sight are written to a single multiblock PLOT3D-type file,
one line per block. These should be checked visually before proceeding
with radiation calculations. Ensuing procedures will interpolate flow
data onto the discretized lines of sight and write results compatible with
the radiation solver. The file name is hard-coded as 'los.g' as part
of simplifying the original nomenclature.
Method:
> A handful of prompts serve to drive the program -- no control file.
> Read the indicated body point (x,y,z) file -- one or more points.
> If a single body point is found, prompt in case hemispherical lines are
NOT intended.
> If all body points are on the centerline (y = 0.), the input volume
grid is assumed to be the starboard (right) half. That half (y > 0.)
also serves for off-center body-normal lines, but for a hemisphere at
an off-center body point, a full volume grid is expected, with the
outer surface closed and convex.
> Build a search tree from the inner multiblock surface, and search
for each body point.
> Generate and store the corresponding unit body-normal vector(s).
> Build a new search tree from the outer boundary, as needed for line-
surface intersection calculations.
> Body-normal line case:
> For each body point B:
> The body normal unit vector is already in hand.
> A search interval along the body-normal line for minimizing
the distance to the outer surface is provided by s = 0 at the
low end and s = the diagonal of the bounding box for the
outer surface mesh.
> The intersection point is then conveniently calculated via
a 1D minimization w.r.t. distance s from B along the body-
normal line (INTSEC6).
> Impose a CFD-type point distribution on the intersected
interval BP, as needed for interpolated flow field data.
DPLR-type use of inputs nk, ds1, ds2fraction will suffice.
> Write the discretized line(s) of sight to the indicated PLOT3D-
type output file.
> Hemisphere lines of sight case:
> The vertices of a triangulated unit hemispherical quadrant define
the lines of sight between the body point and the outer grid
boundary. The number of uniform points specified along an edge
determines the resolution. Typically, 25 such points are used,
giving 650 lines of sight for a centerline body point, or 1300
lines if the body point is off-center. Centerline points require
just half the volume data, and the eventual integration over
solid angle of radiances from the lines of sight can be doubled.
> The initial unit quadrant is transformed so that line 1/vertex 1
corresponds to the body point normal. After The lines for the
first quadrant are generated, the quadrant is rotated 90 degrees
about the normal and the next set of lines is generated. Then if
the body point is off-center, this is repeated twice more.
> Each intersection involves a 2-point line from the body point
through a quadrant vertex, of length derived from the bounding
box of the outer shock surface. This length is overkill for many
intersection search intervals, but that doesn't matter. Each
hemisphere line is intersected and discretized as for the body
normal (tangent-slab) case described above.
> Triangulation choice:
This version of SLOS has the option to read a previously generat-
ed unit triangulation that the Equal_Area_Triangulation variant
of the NPOPT_DRIVER framework can now produce. If nedge is spec-
ified as 25 (say), we look for the file unit_sphere_octant_25.dat
(likewise for other nedge values). If the file is not found, the
spherical_triangulation subroutine is employed as originally.
In principle, the integrations w.r.t. solid angle of radiances
calculated for each line of sight will be more accurate if the
solid angle elements are essentially equal, although it is ex-
pected that the two kinds of triangulation will produce very
similar results for a given value of nedge.
History:
11/02/2018 D.A.Saunders Started adapting USLOS for the structured case.
The option for entering body point indices in
lieu of x/y/z coordinates has been dispensed
with. Use GU to coarsen a surface grid if
necessary. Generation of shock-normal lines
and lines parallel to Ox have also been omitted.
Use the original LINES_OF_SIGHT for these.
11/07/2018 " " Implementation and testing completed.
11/09/2018 " " Use 'los.g' for the output file name as part of
adjusting NEQAIR_Integration to handle the
original nomenclature and this simplified form.
11/09/2018 " " Use 'los.g' for the output file name as part of
11/09/2018 " " Use 'los.g' for the output file name as part of
10/04/2019 " " Resetting the lower intersection search interval
to zero was not being done for each new line.
Earlier testing must not have encountered any
intersection failures that are caused by outer
grid boundaries that are not strictly convex!
10/05/2019 " " Failed intersections prompted another way of
retrying: discretize the part of the line beyond
the failed solution and evaluate line-surface
distances from those points; pick the interval
containing the smallest distance, and redo the
line-surface intersection. REMEMBER THAT THE
ARC LENGTH INTERVAL PASSED TO INTSEC6 SHOULD BE
NORMALIZED BY THE LINE LENGTH IF THAT IS THE
INTERVAL EXPECTED TO CONTAIN THE INTERSECTION,
AS IT IS HERE, USING THE BOUNDING BOX DIAGONAL.
07/21/2020 " " A "small" tolerance in transform_quadrant was
too big. Body points near the nose (normally
not the case for full angular integration)
produced inaccurate rotation axes for the second
quadrant of the transformed unit hemisphere.
This led to irregular heat flux results near the
heatshield apex on the centerline, now fixed.
Avoiding a cross product of parallel unit
vectors (zero length vector) was the root cause.
07/26/2020 " " Reluctant special-casing of the (0,0,0) body pt.
that proves pathological in the sense that the
associated surface grid cell is almost certain
to be a corner cell for which the existing
adjustment that forces the body normal to be in
the symmetry place still doesn't force it to be
along -Ox as it should be for a body of revolu-
tion centered on Ox. Remember that we normally
don't apply the hemisphere LOS method to the
nose region anyway, but it has been needed as
part of comparing radiative heat flux results
with those from a vasdtly more efficient method.
10/08/2020 " " Replace intsec6 with intsec9 as introduced for
line_surface.f90 (interp3d library) used by the
Stardust_Lines utility. This packages a retry
method that is virtually bulletproof.
04/12/2021 " " Handle reading of a unit spherical octant tri-
angulation named unit_sphere_octant_n.dat where
n = nedge is typically 25. (See above for more
details.) If the appropriately named file is
not found, subroutine spherical_triangulation is
invoked as originally. Not much difference is
expected to be found in the integrations with
respect to solid angle subtended at the body
point, but until the option was provided, we
could never know ...
07/11/2022 " " Installed a new option needed for radio blackout
predictions: a line of sight in an arbitrary
direction defined by a cone angle and a clock
angle appended to the body point x y z.
07/13/2022 " " The clock angle was being rotated about Oz when
it should be about Ox.
03/31/2023 " " Use the trigd module for g95 compatibilty.
03/08/2024 " " The buffer length of 64 for the body point data
caused grief! Raise the limit.
Author: David Saunders, AMA, Inc. at NASA Ames Research Center, CA.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Download Latest Version
geomlib.tar.gz (58.3 kB)
