• 1
AMU_-2001-1004
PREDICTION OF BUSINESS JET AIRLOADS USING THE
OVERFLOW NAVIER-STOKES CODE
Elias Bounajem"
Cessna Aircraft Company, Wichita, Kansas 67215
and
Pieter G. Buning"
NASA Langley Research Center, Hampton, Virginia 23681
Abstract
The objective of this work is to evaluate the
application of Navier-Stokes computational
fluid dynamics technology, for the purpose of
predicting off-design condition airloads on a
business jet configuration in the transonic
regime. The NASA Navier-Stokes flow
solver OVERFLOW with Chimera overset
grid capability, availability of several
numerical schemes and convergence
acceleration techniques was selected for this
work. A set of scripts which have been
compiled to reduce the time required for the
grid generation process are described.
Several turbulence models are evaluated in
the presence of separated flow regions on
the wing. Computed results are compared to
available wind tunnel data for two Mach
numbers and a range of angles-of-attack.
Comparisons of wing surface pressure from
numerical simulation and wind tunnel
measurements show good agreement up to
fairly high angles-of-attack.
Introduction
Computational Fluid Dynamics (CFD)
technology has seen remarkable advances
* Member, AIAA
"*Associate Fellow, AIAA.
Copyright © 2001 by the American Institute of Aeronautics and
Astronautics, Inc. All rights reserved.
in the last few years. CFD is used today on a
regular basis to support design efforts in
such areas as aircraft design, propulsion
system design and integration, combustion,
ship design and the automotive industry, to
name a few.
The aircraft industry is a primary customer of
CFD technology. Depending on the task at
hand, the aircraft designer has a wide range
of sophistication to chose from. This range
includes: full modeling of viscous effects as
available in Navier-Stokes type codes; Euler
codes, transonic small disturbances and full
potential codes when viscous effects are not
of primary influence; and finally, linearized
potential flow codes for shock free flows or
when only an approximate answer is being
sought. These codes are implemented with
different types of grids: multiblock, patched
or overset grids if a structured grid approach
is considered, Cartesian and unstructured
grids for unstructured methods. The
literature is rich with publications identifying
the pros and cons of each of these methods.
The bulk of CFD work in the aircraft design
arena has been in the cruise flight regime.
Under these conditions, the airflow is still
attached to the surface with small regions of
separated flow, and CFD calculations are
known to be quite reliable in predicting
aircraft performance characteristics.
Areas which have not received as much
attention are those that deviate from the
1
American Institute of Aeronautics and Astronautics
cruise condition. These situations are
encountered when the entire flight envelope
of the airplane is of interest. Here,
combinations of angle-of-attack, Mach and
Reynolds number can be challenging for any
CFD code. High angle-of-attack will cause
the flow to separate. As the high subsonic
regime is approached, the flow will have
numerous shock waves on the wing, pylon,
etc. The interaction of shock waves with the
boundary layer complicates flow patterns
and makes CFD prediction significantly more
difficult. With the presence of large
separated flow regions, the focus shifts more
and more towards turbulence modeling.
Here, the validity of the model's underlying
assumptions becomes critical to the quality
of the solution the CFD code can provide.
The present work aims at evaluating the
applicability of Navier-Stokes CFD
technology to predict off-design airloads on a
business jet configuration. OVERFLOW
[1,10], a NASA research flow solver which
uses the Chimera overset grid approach,.
has been selected for this evaluation.
OVERFLOW offers a wide array of
numerical schemes and turbulence models,
and has been accepted throughout the
aircraft industry as one of the leading Navier-
Stokes codes available.
Grid Generation
The overset grid approach has the
advantage of allowing grid generation of
aircraft components separately, thus
providing grids that conform to the
component topology. Through a hole cutting
procedure and boundary point interpolation,
excess overlap between component grids is
blanked out.
The geometry modeled here is a business
jet configuration with aft fuselage mounted
pylon/nacelles. Figure 1 shows the different
surface grids that make up the current
model. In all fifteen grids, including the outer
box grids, were needed to define the
geometry.
Surface Grids
The first step in the grid generation process
is to generate surface grids for the various
AIAA-2001-1004
components. Five components were
identified: wing, fuselage, pylon, nacelle and
wing-body fairing. For the wing, pylon, and
nacelle, the CATIA CAD model was
imported in IGES format into the NASA
Langley GridTool software [2], where a
surface grid was generated. The fuselage
surface grid was generated directly in CATIA
to improve surface grid quality resulting from
projection onto the original geometry. As for
the wing-body fairing (Figure 2), it was
deemed appropriate to represent it with a
combination of collar grids and therefore no
computational surface grid was generated
for this component.
Collar Grids
The next step in the process is to make use
of the collar grid approach [3] in the overlap
region between the various components.
This approach is used to ensure good quality
grids in these areas, to allow for inter-grid
communication through interpolation and to
capture the viscous effects in the juncture all
in the same grid. Collar grids are generated
by identifying the intersection curve between
components and growing a grid onto the
adjacent components.
Collar grids were generated at the pylon-
nacelle and pylon-fuselage junctures. For
the wing-body fairing, a collar grid was
generated at the fuselage-fairing intersection
and at the wing-fairing intersection. A third
grid originating at the model centerline
covered the remainder of the fairing.
The NASA Ames Chimera Grid Tools (CGT)
software package [4] was used to generate
the collar grids, to add wakes to the wing
and pylon grids, and to add a wing tip cap
grid [14]. Another cap grid was required for
the pylon shelf, which extended beyond the
nacelle exhaust plane. To increase
communication with the nacelle grid, the
pylon cap grid was extended upstream and
projected onto the nacelle surface (Figure
3).
Volume Grids
From the body-fitted surface grids, volume
grids were generated with HYPGEN [5-6].
The initial spacing off the surface is equal to
0.00035 inches, which corresponds to a y÷
2
American Institute of Aeronautics and Astronautics
value of 1 at 10 percent chord from the wing
leading edge.
Communication between the component grid
outer boundaries, and extension of the
computational domain to the far field is
accomplished with two levels of Cartesian
box grids, following an approach used for
other geometries [7-8].
Grid Communication
The PEGSUS 4.0 code [9], developed by
CALSPAN at AEDC, is used to remove
excess overlap between grids and find
interpolation stencils for inter-grid boundary
points. This process effectively connects all
of the overset component and box grids into
one system.
Scriptin.q the Process
It is recognized that this grid generation
process is iterative. Several attempts may be
made to ensure that the surface is
adequately represented and that the number
of orphan points, grid points with no
adequate interpolation stencil, is minimized.
Also, during development phase, the
configuration is constantly changing as more
refinements are introduced. With this in
mind, a set of scripts that performs the tasks
highlighted above is highly desirable. This
allows for making changes to the various
components and then regenerating the
entire grid system with a minimum amount of
user input.
The present work uses a modified set of
scripts, originally developed for a generic
business jet configuration. These scripts
encompass surface grid generation and
refinements, volume grid generation, hole
cutting and interpolation stencil identification.
An example illustrating the script approach is
the generation of the wing-body fairing collar
grids. As shown in Figure 4, this process
starts with wing and fuselage surface grids,
the fairing definition in the form of reference
grids (created in GridTool from the IGES
CAD definition), and grid lines representing
the intersection of the fairing with each of the
fuselage, wing, and symmetry plane.
The first step uses the SURGRD surface
grid generation code [15] from CGT to
AIAA-2001-1004
create collar surface grids, shown in Figure
2. Second, volume grids are generated with
HYPGEN. Finally, CGT utilities are used to
add reflected symmetry planes and smooth
the wake region of the wing-fairing collar.
Flow Solver
The OVERFLOW Navier-Stokes flow solver
is used in this analysis. This code uses an
implicit approximate factorization algorithm
to solve the thin-layer formulation of the
Navier-Stokes equations. For these
calculations, central differencing with
second- and fourth-order artificial dissipation
is used for the Euler terms. Trial runs with
Roe's upwind differencing scheme did not
show improvement over central differencing.
Local time stepping, grid sequencing, and
multigrid are used to accelerate
convergence [10].
While steady-state acceleration techniques
were used for these simulations, the mid-
range and high angle-of-attack cases were
largely separated. Total lift coefficient varied
by 0.02 from a mean value in some cases.
It is recognized that a more thorough
analysis of the unsteady aspect of these
flowfields is needed; however, for the
airloads analysis process only averaged
steady loads are desired.
Processinq Requirements
The grid system for this configuration has a
total of 3.9 million points. Solutions were
generated on an SGI Origin 200 with four
processors and 1.8 GB of memory. Each
case required about 260 MB of memory and
30 hours to converge on a single processor,
except for higher angles-of-attack. These
were processed further to get the solution
stabilized within a certain band.
Results and Discussion
A total of six flow conditions are examined in
the current study. These consist of a series
of three angles-of-attack (low, mid-range
and high) at two transonic Mach numbers
above cruise. For each flow condition,
results were obtained using the Baldwin-
Barth [11] and Spalart-AIImaras [12] one-
equation turbulence models, and the k-
omega two-equation turbulence model [13].
3
American Institute of Aeronautics and Astronautics
Attempts to use the Menter SST model [16]
were unsuccessful due to wing root grid
issues. Surface pressure coefficients on the
wing are presented at two stations, one
inboard and one outboard. The Reynolds
number matches that at which the
experimental data was collected in the wind
tunnel.
At low angle-of-attack, where the flow
remains attached to the surface, solutions
using the Baldwin-Barth model show surface
pressures in very good agreement with wind
tunnel data (Figure 5), with shock location
and strength well predicted. Spalart-AIImaras
results are similar. The k-omega model
predicts a shock location significantly aft of
the one-equation models. This behavior is
similar to that shown in Reference 13 for
transonic flow over the RAE 2822 airfoil.
For the mid-range angle-of-attack (Figure 6),
Baldwin-Barth results show shock location
and strength are well predicted on the upper
surface inboard station for both Mach
numbers and on the lower surface for the
higher Mach number case. The upper
surface outboard station pressure
comparison shows that the shock position is
about 5 percent chord aft of that measured
in the wind tunnel. Here the flow separates
behind the shock. For the most part the
lower surface predictions are good. Toward
the trailing edge, the calculated pressure
coefficient shows significant unsteadiness,
varying with where the solution is stopped.
This is more noticeable in the high Mach
number case. This behavior is not present in
the Spalart-AIImaras results, where the flow
is better behaved at the trailing edge. The
predictions on the upper surface of the
inboard station are slightly better than
Baldwin-Barth. Again, the k-omega model
predicts the shock location aft of the other
models.
In the high angle-of-attack case (Figure 7),
where a considerable amount of separation
is present, the upper surface rooftop and
shock location predicted by both one-
equation models are fairly good. However,
pressures aft of the shock at the inboard
station do not match the trends of the wind
tunnel data. A lower pressure at the upper
surface trailing edge leads to an acceleration
of the flow on the lower surface approaching
AIAA-2001-1004
the trailing edge. Results for intermediate
stations (not shown here) indicate that this
pattern disappears a short distance outboard
of this station. This could be a result of
solution convergence difficulty at these
extreme conditions. Further investigation of
this issue is required. Although the k-omega
model continues to predict a shock location
too far aft as in the previous cases, the post-
shock pressure levels match wind tunnel
results. The lower surface predictions for all
models remain good.
Support for the aerodynamic loads process
is provided by supplying design loads for
various components of the aircraft such as
the fuselage, nacelle, pylon, etc. This is
done by providing total component load
(e.g., pylon normal force coefficient), or by
providing a running load.
Here, running loads are computed from the
CFD solution by dividing the body into a
number of segments and integrating the
pressure to obtain either a force or a
moment coefficient in these segments. Data
can be presented either as a cumulative total
starting from a specific point, or separately
for each of the segments, as in Figure 8 for
the fuselage. Individual component loads
can also be used to determine the fraction of
the aircraft total load that is being carried by
a particular component (Figure 9).
If accurate, CFD solutions are invaluable for
the aerodynamic loads process because
they provide distributed surface pressures
which can be analyzed component-by-
component. These supplement wind tunnel
data from extensive pressure taps or
component balances, both expensive
experimental techniques.
Conclusion
A Navier-Stokes flow solver, OVERFLOW,
has been successfully used for the
prediction of aerodynamic loads on a
business jet configuration. Three turbulence
models were evaluated at above-cruise
Mach numbers and low to high angles-of-
attack. Overall, results show the Baldwin-
Barth and Spalart-AIImaras models providing
the closest match to experimental results.
The k-omega turbulence model predicts the
shock location and flow separation farther aft
4
American Institute of Aeronautics and Astronautics
than the one-equation models. Further
investigation of isolated flow patterns at the
middle and high angle-of-attack cases is
needed.
References
1. P.G. Buning, et al., "OVERFLOW User's
Manual, Version 1.8," NASA Langley
Research Center, Hampton, VA, Feb. 1998.
2. J. Samareh-Abolhassani, "GridTool: A
Surface Modeling and Grid Generation
Tool," NASA CP-3291, 1995.
3. S.J. Parks, P.G. Buning, J.L. Steger, and
W.M. Chan, "Collar Grids for Intersecting
Geometric Components Within the Chimera
Overlapped Grid Scheme," AIAA 91-1587,
June 1991.
4. W.M. Chan, "Manual for Chimera Grid
Tools," NASA Ames Research Center,
Moffett Field, CA, Oct. 1998.
5. W.M. Chart and J.L. Steger,
"Enhancements of a Three-Dimensional
Hyperbolic Grid Generation Scheme," Appl.
Math. and Comput., Vol. 51, pp. 181-205,
1992.
6. W.M. Chan, I.-T. Chiu, and P.G. Buning,
"User's Manual for the HYPGEN Hyperbolic
Grid Generator and HGUI Graphical User
Interface," NASA TM 108791, Oct. 1993.
7. D.G. Pearce, et al., "Development of a
Large-Scale Chimera Grid System for the
Space Shuttle Launch Vehicle," AIAA 93-
0533, Jan. 1993.
AIAA-2001-1004
8. R.L. Meakin, "Moving Body Overset Grid
Methods for Complete Aircraft Tiltrotor
Simulations," AIAA-93-3350, July 1993.
9. N.E. Suhs and R.W. Tramel, "PEGSUS
4.0 User's Manual," AEDC-TR-91-8, Arnold
Engineering Development Center, Arnold
AFB, TN, Nov. 1991.
10. D.C. Jespersen, T.H. Pulliam, and P.G.
Buning, "Recent Enhancements to
OVERFLOW," AIAA 97-0644, Jan. 1997.
11. B.S. Baldwin and T.J. Barth, "A One-
Equation Turbulence Transport Model for
High Reynolds Number Wall-Bounded
Flows," AIAA 91-0610, Jan. 1991.
12. P.R. Spalart and S.R. AIImaras, "A One-
Equation Turbulence Model for Aerodynamic
Flows," La Recherche Aerospatiale, No. 1,
1994, pp. 5-21.
13. J.E. Bardina, P.G. Huang, and T.J.
Coakley, 'q'urbulence Model Validation,
Testing, and Development," NASA TM
110446, April 1997.
14. S.E. Rogers, H.V. Cao, and T.Y. Su,
"Grid Generation For Complex High-Lift
Configurations," AIAA 98-3011, June 1998.
15. W.M. Chan and P.G. Buning, "Surface
Grid Generation Methods for Overset Grids,"
Computers and Fluids, Vol. 24, No. 5, 1995,
pp. 509-522.
16. F.R. Menter, 'q-wo-Equation Eddy
Viscosity Turbulence Models for Engineering
Applications," AIAA J., Vol. 32, Nov. 1994,
pp. 1299-1310.
5
American Institute of Aeronautics and Astronautics