American Institute of Aeronautics and Astronautics
1
Integration of Multifidelity Multidisciplinary Computer
Codes for Design and Analysis of Supersonic Aircraft
Karl A. Geiselhart
1
and Lori P. Ozoroski
2
NASA Langley Research Center, Hampton, Virginia 23681, USA
James W. Fenbert
3
and Elwood W. Shields
4
Alliant Techsystems, Inc., Hampton, Virginia 23681, USA
and
Wu Li
5
NASA Langley Research Center, Hampton, Virginia 23681, USA
This paper documents the development of a conceptual level integrated process for design and
analysis of efficient and environmentally acceptable supersonic aircraft. To overcome the
technical challenges to achieve this goal, a conceptual design capability which provides users with
the ability to examine the integrated solution between all disciplines and facilitates the application
of multidiscipline design, analysis, and optimization on a scale greater than previously achieved, is
needed. The described capability is both an interactive design environment as well as a high
powered optimization system with a unique blend of low, mixed and high-fidelity engineering tools
combined together in the software integration framework, ModelCenter. The various modules are
described and capabilities of the system are demonstrated. The current limitations and proposed
future enhancements are also discussed.
1
Aerospace Engineer, Aeronautics Systems Analysis Branch, Mail Stop 442.
2
Aerospace Engineer, Aeronautics Systems Analysis Branch, Mail Stop 442.
3
Senior System Engineer, Aeronautics Systems Analysis Branch, Space Division.
4
Senior System Engineer, Aeronautics Systems Analysis Branch, Space Division.
5
Senior Research Engineer, Aeronautics Systems Analysis Branch, Mail Stop 442.
American Institute of Aeronautics and Astronautics
2
Nomenclature
A
e
= equivalent area
C
L
= lift coefficient
dp/p = (the calculated pressure – the ambient pressure)/(the ambient pressure)
EXTR = extraction ratio
FPR = fan pressure ratio
L/D = lift to drag ratio
nmi = nautical miles
OPR = overall pressure ratio
OML = outer mold line
SFC = specific fuel consumption
T
3
= compressor exit temperature, °R
T
4
= combustor exit temperature, °R
TTR = throttle ratio
V
app
= approach velocity, kts
V
jet
= jet velocity, ft/s
I. Introduction
HE design of an efficient, environmentally acceptable, and economically viable supersonic transport remains one
of the most challenging problems for aircraft designers. The solution to this design problem does not reside
within one discipline but will only be found by investigating the complex interactions between various
disciplines. The ability to apply integrated design in the conceptual stage is the only way to ensure that these, often
conflicting areas, can effectively be explored to achieve the demanding design goals.
There are many examples of Multi Disciplinary Optimization (MDO) applications to supersonic aircraft
throughout the literature.
1-13
There are also many examples of conceptual aircraft design reports where the authors
described going “deep” in a particular discipline, focusing on a single cruise point low-boom and/or low-drag design
in their process.
14-30
Many of these are often byproducts of tool and method development and the testing of
optimization algorithms and/or schemes. There are fewer instances focused on supersonic design for low-boom
concepts with shape optimization tied to overall vehicle performance.
1,2,4
This current effort considers the lessons of
the past and the need for a multi-user, robust, flexible, workhorse system capable of integrating more disciplines at
multiple levels of fidelity. The present capability includes a comprehensive suite of functional modules ranging from
setting up user displays and directories, selecting levels of desired analysis for different types of problems, selecting
a wide range of automated plots to view during execution, to detailed inputs controlling the actual analysis codes
and design optimization algorithms.
The development of an effective integrated process for multifidelity, multidisciplinary, design optimization and
analysis for a low-boom and low-drag supersonic aircraft concept has been completed. This integrated process
includes propulsion system design and analysis, mission performance and takeoff analyses, and community noise
assessment. The process also includes low-fidelity codes for aerodynamic performance that can be used for cruise
point performance and equivalent-area based sonic boom analyses. The same low-fidelity codes can also be used to
generate low speed polars for takeoff and landing analyses and for generating polars for a full mission performance
analysis. In addition to the low-fidelity analysis codes, there is a mixed-fidelity method for low-boom equivalent-
area (A
e
) based design using computational fluid dynamics (CFD) codes, and a high-fidelity sonic boom analysis
using off-body CFD pressure distributions for boom propagation. For A
e
based low-boom design, a low-boom target
A
e
is needed. The integrated process includes several options for A
e
target generation: 1) it can be generated by the
classic Seebass-George-Darden boom minimization theory, 2) it can be created interactively by manually adjusting
an f-function or the actual A
e
distribution for a favorable ground signature and loudness, or 3) developed through the
use of genetic optimization to minimize loudness for a numerically optimal low-boom target. The integrated process
allows easy formulations of optimization of the overall aircraft at the systems level and for the evaluation of various
optimization strategies.
The overall implementation details of the integrated process using ModelCenter
®
9.0
31
are given in Section II,
followed by sections describing the details of various disciplinary analysis assemblies. Example capabilities of the
process are demonstrated followed by conclusions and future plans.
T
American Institute of Aeronautics and Astronautics
3
II. Implementation in ModelCenter
The analysis and design environment has been built in ModelCenter 9.0 taking full
advantage of the new process flow capabilities included within the latest ModelCenter
release. Figure 1 shows the high level model that has been developed for supersonic
applications. The implemented process flow model is quite different from the data flow
model previously developed for this application as documented in Ref. 32. One
significant advantage of the process flow modeling option is that it allows independent
components to execute in parallel. In addition, the implementation of logic nodes
provides the ability to make seamless changes in the analysis path of the integrated
process, through either user selection or analytic determination. Process flow also allows
conditional links that are especially useful when integrating multifidelity analysis codes.
For example, total equivalent area can be generated by high-fidelity or low-fidelity
methods, and the solutions from both methods can be linked to the same downstream
analysis component. One can get either a low-fidelity or high-fidelity analysis result by
simply selecting the desired path. The adoption of the process flow capabilities within
ModelCenter has been instrumental in developing the flexible and multifidelity
capabilities presented herein. The current model also implements data objects to make
large amounts of data available to multiple components in the model, greatly reducing
the number of required data links in the model. In the current process, data objects are
simply file variables, containing geometric, aerodynamic, configuration, and flight
condition data. This development has also greatly improved model organization and
maintenance over the prior capability documented in Ref. 32.
In the early stages of low-boom and/or low-drag design, full mission performance
and community noise constraints may not be as important as low-boom and low-drag
design goals. Moreover, the cost of these analyses can be prohibitive even when
conducting low-fidelity optimization. Therefore, a tiered approach is used for running
low-fidelity analyses and using low and mixed-fidelity analyses for design optimization.
In the first tier, enough data for a single cruise point analysis are generated. A low-
fidelity weights analysis is included in the first tier so that Breguet range analysis results
can be used as part of the objective or constraint. In the second tier, the data needed for a
detailed takeoff and landing analysis are added. In the third tier, the data required to fly a
full mission are added. All aerodynamic data needed for low and mixed-fidelity low-
boom design are generated in all three tiers. This tiered approach allows the user to
perform optimization or parametric analyses at varying levels of cost per function
evaluation by switching an analysis option. A low-boom and/or low-drag optimization
could be performed for cruise only, with no constraints on takeoff or landing field
lengths, or at cruise combined with low speed aerodynamics to allow constraints on
takeoff and landing field lengths, approach velocity, etc. at a slightly higher cost. The
requirement for a full set of aerodynamic data for a complete mission performance
analysis will significantly add to the overall cost and community noise computations are
even more costly.
III. Propulsion System Analysis Assembly
Propulsion system analyses are performed using the Numerical Propulsion Simulation System
33
(NPSS) and an
improved version of the Weight Analysis of Turbine Engines
34
(WATE) computer code (WATE++) is used for the
propulsion system weight and flow path. NPSS and WATE++ are used to generate the propulsion system weight,
nacelle geometry, performance data for mission and noise analyses, and the input for low-fidelity plume shape and
for CFD based engine simulation. Following the tiered approach philosophy, NPSS and WATE++ generate weights
and data for the tier I cruise point analyses. NPSS and WATE++ will generate the engine state tables and geometric
data for the Aircraft Noise Prediction Program
35
(ANOPP) used for community noise analysis. Additionally, the
model can be run with or without NPSS analysis. If the model is set to skip NPSS analysis, the existing propulsion
system weight and nacelle geometry are scaled by thrust for wave drag, skin friction, CFD, and overall aircraft
performance and noise analyses. There is a component in the propulsion assembly that generates the input to
Vehicle Sketch Pad
36
(VSP) for the nacelle geometry. The current system uses a minimum size nacelle based on
Figure 1. Top level
process in ModelCenter.
American Institute of Aeronautics and Astronautics
4
various engine flowpath parameters from WATE++. Future improvements in this system will include new inlet,
nacelle and nozzle design modules currently being developed at NASA. Integration of full NPSS models within this
system is a major step forward in propulsion-airframe integration at the system design level for supersonics where
traditionally these propulsion design parameters have been uncoupled from the rest of the airframe.
IV. Geometry Assembly
In the initial stage of conceptual design, VSP
is used to model the geometry. VSP is an easy-to-use parametric
geometry modeler developed at NASA in recent years. A point definition of the VSP geometry, similar to
PLOT3D’s
37
format for representing rectangular grids, is used as the standard geometry format for all analysis
codes. Due to differences in geometry requirements for various codes, a geometry format conversion code hrm2geo
has been developed to convert VSP geometry for both low-fidelity and high-fidelity analyses. In particular, input
files for the wave drag code, the equivalent area calculation code, skin-friction drag code, linear aero codes,
Cart3D
38
, and VGRID
39
(a CFD volume meshing code) can all be generated from VSP geometry using hrm2geo.
This allows both low and high-fidelity codes to use the same underlying geometry model for all analyses. VSP can
also automatically export its
parametric geometry as a closed
triangulation mesh suitable for
Cart3D CFD analysis.
Initially, a baseline geometry
can be laid out using the
standalone VSP graphical user
interface. A ModelCenter Plugin
has been developed to load a VSP
parametric geometry into
ModelCenter and to expose the
desired geometric parameters of
interest. Any future modification
of the exposed geometry variables
can quickly be modified within the
VSP plug-in. This provides a very
flexible means of changing design
variables for shape optimization
within ModelCenter. Additionally,
the geometry can be visually
examined during the ModelCenter execution process to monitor the impact of geometric design variable changes to
the aircraft shape. This is particularly helpful during optimization or parametric variation. Figure 2 shows a
ModelCenter Geometry View window in which the current VSP geometry has been selected for display as a shaded
solid. Other options include selections for displaying the original and intermediate geometries as wireframes or
shaded solids.
In addition to the basic outer mold line (OML) geometry, additional components allow the user to set up
geometric constraints related to the vehicle, such as sizes and locations of landing gear, control surfaces, fuel tanks,
passenger cabin, and various ground clearance angles. This information then becomes available to constrain the
OML optimization or to regenerate non OML geometry information necessary for the analysis process.
V. Wing Design and Lift Matching Assembly
Once the propulsion system has been developed and the geometry is modeled, the process includes options for
doing a wing camber surface design
40
for the given planform to minimize the drag for a fixed C
L
. This camber
surface design process can be skipped if airfoil section parameters from the VSP geometry are included in the
overall system level optimization. In addition, the user can select to skip any wing camber optimization if desired.
A lift matching process is then executed to determine the angle of attack for the required C
L
and the given tail
deflection. The camber surface design and lift matching module is currently composed of modified linear theory
tools. A future improvement to this module will include the addition of a process for CFD based camber designs.
This automated CFD analysis has already been implemented in a ModelCenter process but has not been included in
the current build for wing camber design.
Figure 2. ModelCenter geometry view of VSP model.
American Institute of Aeronautics and Astronautics
5
VI. A
e
Based Low-Boom Design Assembly
The low-boom design module capabilities have been developed to provide both automated and user interactive
operation, for both A
e
target development and A
e
target matching. There are currently two options to generate target
equivalent areas. The Hybrid code
41
uses the George-Seebass-Darden boom minimization theory to generate target
A
e
distributions for low-boom design. In addition, a
parametric A
e
target generation tool was developed to
generate A
e
targets for low-boom design that allow trade-
offs between A
e
volume requirements for a configuration
and the PLdB level of the ground signature.
42
Additionally, the process contains a method for shaping
fuselage and fuselage like components to match the A
e
of
the configuration to the target A
e
.
43
Future development
plans include the addition of methods for off body
pressure target generation and matching as an additional
approach for low-boom design. Figure 3 provides a high
level view of the A
e
based low-boom design assembly.
Equivalent area calculations are available at low,
mixed, and high-fidelity levels. The supersonic modified
linear aerodynamics computer program, LTSTAR
44
is
used to generate the low-fidelity A
e
due to lift and
HWAVE, a streamlined and modified version of the
Harris far-field wave drag program
45
, is used for volume
A
e
. Modifications to the original Harris far-field wave
drag code have removed some of its restrictive
requirements for geometry definition. CFD based equivalent areas can also be calculated. Two high-fidelity CFD
codes were selected based on speed and accuracy. Cart3D is a high-fidelity inviscid analysis package that uses
adaptively refined Cartesian meshes for conceptual and preliminary aerodynamic design. Cart3D runs relatively fast
and has less restrictive requirements on the input geometry definition than other CFD codes. USM3D
46
is an
Euler/Navier-Stokes solver for unstructured, tetrahedral meshes. USM3D is slower, but can be used with SSGRID
47
to shear and stretch the VGRID generated grid to enable computing off-body pressure distributions for propagation
to the ground. The processes for both CFD codes are completely automated and include options for selecting
processors, using restart files, automated interactive generation of many plots for visualizing grids and flow field
properties to examine the results if desired. The low-boom design methodology using the low and mixed-fidelity
capabilities are documented in detail in Ref. 14.
VII. Aerodynamic Analysis Assembly
Currently, aerodynamic analyses are
divided into two categories, low-fidelity
and high-fidelity. At this time, low-fidelity
analysis methods are used during
optimization and the high-fidelity results
can be used as a final verification. The
low-fidelity aerodynamics data is generated
using a collection of tools
40,45,48
with
runtimes varying from fractions of a
second to several seconds depending on the
tier. The low-fidelity analysis module is
shown in Figure 4. The high-fidelity CFD
tools and processes are generally the same
as those used in the high-fidelity A
e
analysis process. Future enhancements to
this module are expected to include
automated methods for using high-fidelity
results to correct low-fidelity results.
Figure 4. Aerodynamic analysis assembly.
Figure 3. A
e
based low-boom design assembly.
