5
5
This paper conducts a numerical study of four sets of aerodynamic control surfaces,
called grid fins, mounted on a full Launch Abort Vehicle geometry using a Cartesian Euler
Solver with embedded boundaries and adjoint-driven adaptive meshing. Since Cartesian
methods are insensitive to complex geometry, this makes them suitable for the highly
complex resulting configuration. Numerical results are compared against sub- trans- and
supersonic wind tunnel data in order to examine our ability to accurately predict the force
and moment increments afforded by these unconventional control surfaces. A database
of 1152 separate cases were ran including 12 different Mach numbers from 0.5 - 2.5, 15
different angle of attacks (0 - 15◦ ), and 6 different geometries. Overall, the simulation
data show good agreement with tunnel runs and similar rankings of configurations and
trends are found between wind tunnel and simulation results suggesting a large potential
for Cartesian Euler Solvers in accurately predicting force and moment increments for grid
fins on the Launch Abort Vehicle. In particular, pitching moment was predicted accurately
over the entire Mach-alpha space.
Nomenclature
M∞ = free-stream mach number
CA = axial force in missile axis
CN = normal force in missile axis frame
Cm = pitching moment in missile axis frame (with respect to c.g.)
α = angle of attack, deg
CP = pressure coefficient
E = total output error in functional value
J = functional value
e = relative error in functional value
ψ = adjoint variable
R = Euler residual terms
(+) = circumferential positioning of grid fins. first fin aligned w/ vertical
(×) = circumferential positioning of grid fins. first fin 45 degrees w/ vertical
LAV = Launch Abort Vehicle
AM s = LAV abort motors
ACM s = attitude control motors
CM = Crew Module
I. Introduction
Grid fins, also known as lattice fins, are unconventional aerodynamic control surfaces first proposed by
1
Belotzerkovsky et al. They consist of an outer frame which supports an inner lattice of intersecting
∗ Aerospace Engineer, Science and Technology Corporation, Hampton, VA., AIAA Member
† Aerospace Engineer, NASA Ames Research Center, CA, AIAA Associate Fellow
1 of 21
2 of 21
II. Methods
II.A. Geometry
The geometry used for the study was based on the “26aa” CFD watertight model representation of the Orion
LAV shown in Fig. 2. A few of the key components modeled in the geometry include the heatshield, abort
motors, feedline (or raceway), and the umbilical notch cutout. These can be found in Fig. 3. As a result of
these circumferential asymmetries, the configurations were simulated with no assumptions of symmetry.
A total of six different configurations were studied. The first
configuration used the baseline 26aa geometry with no grid fins at-
tached (seen in Fig. 2). The configurations employing the grid fins
are shown in Fig. 4. For all the cases, four fins are used which are
all spaced at 90 degrees. The (+) configurations align the fins at an
offset of 45 degrees with the Abort Motors (AMs) while the (×) con-
figurations align the fins directly aft of the (AMs) in the streamwise
direction. Each of the configurations shown was triangulated us-
ing ∼ 1.2 million triangles. The final configuration consisted of one
isolated unswept grid fin. This work considers only the unpowered
LAV. While power-on experimental data was taken for the fins in the
(+) configuration, corresponding data for fins in the (×) orientation Figure 2. 26aa LAV baseline configuration
was not obtained in the experiment due to high load factors on the (no fins)
grid fins from the Abort Motor jets. Fig. 3 shows a more complete
view of the abort motors, raceway, and umbilical notch cutout.
3 of 21
Figure 4. LAV grid fins configurations. (a) represents the + configuration with the fins located at the compass points.
(b) (+) configuration with 22.5 degree swept grid fins. (c)(×) configuration with 90 degree circumferential spacing and
45 degree offset with from vertical (aligned with the abort motors). (d)(×) configuration with 22.5 degree swept fins.
Figure 5. Fin geometries considered. (a) CAD model unswept fin. (b) CAD model of swept fin with a sweep angle of
22.5 degrees relative to the unswept fin. (c) representative surface triangulation including ≈ 56, 000 triangles.
Fig. 5 displays both the swept and unswept CAD models of the fins along with the discretized surface
triangulation. The fins were meshed directly from the Pro Engineer CAD parts. However, the baseline
4 of 21
5 of 21
The goal in the error estimation procedure is to obtain a cell-wise estimate of this relative error (e(QH ))
from the flow solution. Following the development in Nemec et al.,12 an expression for the functional on the
fine mesh can be obtained as a function of the functional on the coarse mesh in terms of the adjoint variable,
ψ, and the Euler residual terms.
QH
h denotes a reconstruction of the flow solution from the coarse mesh to the fine (embedded) mesh. The
remaining error term in equation 3 can be used to determine a bound on the local error in each cell of the
embedded mesh.
6 of 21
J = CN + 0.2CA (5)
After experimenting with several functional formulations in Nemec et al.7 for a similar LAV geometry,
it was determined that equation 5 results in good mesh convergence for both forces and moments. Fig. 7
shows the mesh adapted grids for a subsonic and supersonic case.
Figure 7. Adapted meshes for swept geometry. Fig. (a) shows a typical mesh of 7.6 million cells after 15 adaptation
cycles. This case was run at M∞ = 0.7 at α = 10 degrees. Fig. (b) shows a typical supersonic case for a typical mesh
size of 7.4 million cells and 15 adaptation cycles. This case was simulated at M∞ = 1.8 and α = 10◦ .
The final refined mesh for the isolated grid fin configuration can be seen in Fig. 8.
Figure 8. Fin only configuration; mesh shown after 15 adaptions. M∞ = 1.3, α = 10 degrees.
7 of 21
Figure 9. Isolated grid fin configuration for 3 different Mach numbers all computed at 0◦ angle of attack. The contours
are colored by square root of Mach number.
Figure 10. Single grid fin computations for Drag Coefficient and Pitching Moment Coefficient for 0, 5, and 15◦ angle of
attack. The transonic drag rise due to the flow choking inside the lattice structure can be clearly seen.
8 of 21
Figure 11. Final adapted meshes for selected LAV cases. 15 adaptation cycles were used for all cases with each case
having ≈ 8 million cells. surface CP contours are also shown.
9 of 21
Fig. 13 displays numerical results for axial force for the 4 configurations at three different flight regimes.
The simulation results are shown against the wind tunnel results. Each plot shows a fins on and fins off case.
The simulation results over-predict the subsonic and transonic drag rise, both with and without fins present.
At higher supersonic mach numbers (not all shown), axial force is predicted accurately by the simulation.
10 of 21
Likewise, simulation results for Normal Force are plotted for the 4 configurations in Fig. 14.
11 of 21
Fig. 15 displays baseline (no grid fins) comparisons for normal force, axial force, and pitching moment
(about the center of gravity) at 5, and 10 degrees angle of attack. The axial force rise at subsonic and
transonic speeds is over-predicted by the CFD results, while at speeds M ach > 1, strong agreement is found
between CFD and experiment. Normal Force, and Pitching Moment comparisons match over the entire
range.
12 of 21
Fig. 18 in the appendix displays axial force and pitching moments for all fin configurations at 10 degrees
angle of attack. The over-prediction of axial force for subsonic, and transonic regimes remains consistent
across all of the configurations.
Table 1 displays percent differences in flow coefficients between CFD and experiment. The CFD results
over-predict axial force. At higher Mach numbers, the axial force differences between the two geometries is
reduced. Similar trends are shown for pitching moments. These differences decrease substantially for Mach
numbers above 1.
Table 1. Comparison of Aerodynamic Load differences between the simulation and the wind tunnel data for the
baseline configuration with no fins. Tabulated maximum differences from α = 0 − 15 degrees. The increments are
computed by 100 ∗ ((Simulation − Experiment)/Experiment).
Pitching moment coefficients are shown in Fig. 16 for the 4 configurations. Improvement is seen between
the computational and experimental results. Importantly, the pitching moment increments due to the fins
are in good agreement with the experiment.
13 of 21
14 of 21
Overall, the fins provide between 74 to 330 % increase in pitch. All configurations show a similar trend of
maximum pitch increase at Mach = 0.5, a slight decrease in pitch stability in the transonic range, and then
a small increase in pitch stability at larger Mach numbers. At subsonic and transonic Mach numbers, the
swept fins produce slightly more control stability relative to the unswept fins, while at higher supersonic Mach
numbers the benefit is nearly indistinguishable.A similar result is found with regard to the configuration. At
the low and transonic Mach numbers, the (×) configuration produces slightly more control authority than
the (+) configuration, while at higher Mach numbers the difference is small.
Fig. 17 shows a comparison of experimental and CFD results across a Mach number range of 0.5 to 2.5 at
α = 3◦ . The relative rankings of the configurations are the same between both experiment and simulation.
Interestingly, in both experiment and simulation, the (×)-swept configuration leads in pitching moment up
to M∞ = 2.0, but is overtaken by the (×)-unswept at higher Mach numbers. The pitching moment increase
in the transonic range is predicted closely by the AERO package. Over the complete range of Mach numbers,
both experiment and CFD simulation predict a slight drag reduction as a result of the swept fins. The benefit
seems to diminish at larger Mach numbers.
15 of 21
16 of 21
Table 4. Simulation prediction of Axial Force, Normal force, and Pitching moment for the swept fin cases. Increments
from baseline (no fins) are shown. The entries of the table represent (Baseline with fins - Baseline) for the case of
swept fins. α = 10 degrees.
For the swept cases, the (×) configuration provides slightly higher stability with comparable axial force
values to the (+) configuration.
17 of 21
M∞ Sweep ∆CA
Unswept ∼ 36.6
0.5
Swept ∼ 35.0
Unswept ∼ 43.3
0.9
Swept ∼ 36.5
Unswept ∼ 38.2
1.1
Swept ∼ 31.9
Unswept ∼ 26.9
2.0
Swept ∼ 21.6
Unswept ∼ 24.8
2.5
Swept ∼ 20.4
Table 6 shows aerodynamic coefficients for the swept and unswept fins for the (+) configuration. For
subsonic and transonic regimes, the swept fins provide the highest stability (as measured by pitching mo-
ment), while providing lower drag increases than the unswept cases. Similar trends are found in Table 7 for
the (×) configuration cases.
Table 6. Simulation prediction of Axial Force, Normal force, and Pitching moment for (+) configuration cases.
Percentage increments from baseline (no fins) are shown. The entries of the table represent 100*((Baseline with fins -
Baseline))/Baseline. α = 10 degrees.
18 of 21
IV. Conclusion
A total of 1152 separate cases were simulated at 12 different Mach numbers from 0.5 - 2.5, 15 different
angles of attack from 0 - 15◦ , and six different configurations using a cut-cell Cartesian grid approach with
embedded boundaries. Adjoint-driven mesh refinement was used to provide a grid system optimized for
surface loads of interest given a reasonable cell budget. Simulations were generally in good agreement with
the wind tunnel data suggesting accurate prediction of the load increments due to the fins using the adjoint-
driven grid refinement with less than 8 million cells per case. Similar trends were seen in the transonic regime
and the ranking of configurations was consistent. Notably, pitching moment and normal force is in excellent
agreement across the database, while axial force is in good agreement to experiment with some tendency to
over-predict the experiment in the subsonic regime. Overall, the swept fins show highest pitching moment
benefit for both the (+) and (×) configurations in both the simulation and the experiment. Preliminary
results show that the (×) configuration provides a slight increase in stability in subsonic and transonic flight.
Future studies will need to look at the aerodynamic effect and hinge moments of the plumes impinging upon
the fins during LAV powered flight.
19 of 21
Figure 18. Force coefficients vs. Mach number. All fin configurations are shown for 10◦ angle of attack. The over-
prediction of axial force for subsonic, and transonic regimes remains consistent across all configurations.
Acknowledgments
Special thanks to Michael Mendenhall at Nielson Engineering & Research, Inc. and Reynaldo Gomez at
Johnson Space Center for providing the geometry and the wind tunnel data. This work was supported by
the NASA Ames Research Center Contract NNA10DF26C.
20 of 21
Moscow, Mashinostroeniye (Russian), 1987, (see also: Belotzerkovsky, S.M. et al. Wings with internal framework. Machine
Translation, FTID(RS)T-1289-89, Foreign Technology Div., 1987).
2 Orthner, K. S., Aerodynamic Analysis of Lattice Grids in Transonic Flow, Master’s Thesis, Dept. of Aeronautics and
2009-1105, 47th AIAA Aerospace Sciences Meeting, Orlando, Florida, Jan 2009.
4 Debiasi, M, Yan, Z., Loon, C.T., Swept-back Grid Fins for Transonic Drag Reduction, AIAA Paper 2010-4244, 28th AIAA
Method, AIAA Paper 2005-0877, 43rd AIAA Aerospace Sciences Meeting, Reno, NV, Jan. 2005. (Also NASA Technical Report
NAS-05-008.)
7 Nemec, M., Aftosmis, M. J., and Wintzer, M., Adjoint-Based Adaptive Mesh Refinement for Complex Geometries, AIAA
Paper 2008-0725, 46th AIAA Aerospace Sciences Meeting, Reno, NV, Jan 2008.
8 Aftosmis, M. J., Berger, M.J., Nemec, Robust and Efficient Cartesian Mesh Generation for Component-Based Geometry
Generation, 1998.
10 Aftosmis, M. J., Berger, M.J., A Parallel Multilevel Method for Adaptively Refined Cartesian Grids with Embedded
Boundaries, AIAA Paper 2000-808, 38th Aerospace Sciences Meeting and Exhibit, Reno, NV, Jan 2000.
11 Venditti, D. A., and Darmofal, D. L., Grid Adaptation for Functional Outputs: Application to Two-Dimensional Inviscid
Meshes, 18th AIAA Computational Fluid Dynamics Conference, Miami, FL, June 2007
21 of 21