See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259981615

Analytically-derived Aerodynamic Force & Moment Coefficients for Resident

Space Objects in Free-Molecular Flow

Conference Paper · January 2014

DOI: 10.2514/6.2014-0728

CITATIONS READS

8 256

5 authors, including:

Kenneth Arthur Hart

Georgia Institute of Technology
8 PUBLICATIONS   16 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Analytic Free-Molecular Aerodynamics View project

All content following this page was uploaded by Kenneth Arthur Hart on 02 February 2014.

The user has requested enhancement of the downloaded file.

 AIAA 2014-0728
AIAA SciTech
13-17 January 2014, National Harbor, Maryland
AIAA Atmospheric Flight Mechanics Conference

Analytically-derived Aerodynamic Force and Moment

Coefficients of Resident Space Objects in
Free-Molecular Flow
Kenneth A. Hart,∗ Soumyo Dutta,† Kyle R. Simonis,‡
Bradley A. Steinfeldt,§ and Robert D. Braun¶
Georgia Institute of Technology, Atlanta, GA, 30332-1510, USA
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

Fast, high-fidelity trajectory propagation of objects in near-Earth orbits is a key capa-

bility for space situational awareness and mitigating probability of collisions on orbit. This
high-fidelity analysis requires accurate aerodynamics prediction for objects in the free-
molecular regime of flight, but most tools for aerodynamic prediction for this regime either
are found using assumptions or are computationally intensive. Symbolic manipulation
software can be used to analytically integrate expressions for pressure and shear pressure
coefficients acting on a general body in free-molecular regime to derive aerodynamic force
and moment expressions. The analytical aerodynamics prediction method is described and
relations have been developed for the sphere, cylinder, panel, and rectangular prism. The
NASA-developed Direct Simulation Monte Carlo Analysis Code is used to validate the ana-
lytical expressions and it is shown that expressions are accurate within 0.38%. These gener-
alized analytic expressions in terms of angle of attack, sideslip angle, freestream conditions,
wall temperature, and accommodation coefficients allow near-instantaneous computation
of the rarefied aerodynamics and enables space situation awareness analysis.

Nomenclature
A Area, m2
C Coefficient (see subscripts)
erf Error function
f Force, N
H Heaviside function
Kn Knudsen number
l Length, m
n Normal vector
p Pressure, Pa.
R Gas specific constant, J/kg-K
r Position vector
S Region of integration
s Molecular speed ratio
T Temperature, K (see subscripts)
t Tangential vector
u Surface parametrization variable
V Velocity, m/s (see subscripts)
v Surface parametrization variable
α Angle of attack, rad
∗ Graduate Research Assistant, Daniel Guggenheim School of Aerospace Engineering, AIAA Student Member.
† Graduate Research Assistant, Daniel Guggenheim School of Aerospace Engineering, AIAA Student Member.
‡ Undergraduate Research Assistant, Daniel Guggenheim School of Aerospace Engineering, AIAA Student Member.
§ Research Engineer II, Daniel Guggenheim School of Aerospace Engineering, AIAA Member.
¶ David & Andrew Lewis Professor of Space Technology, Daniel Guggenheim School of Aerospace Engineering, AIAA Fellow.

1 of 14

American Institute of Aeronautics and Astronautics

Copyright © 2014 by Kenneth Hart, Soumyo Dutta, Kyle Simonis, Bradley Steinfeldt, and Robert D. Braun. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
 β Sideslip angle, rad
λ Mean free path, m
θ Flow angle, rad
σ Momentum accommodation coefficient (see subscripts)
τ Shear force, Pa.
Subscript
A Axial force
D Drag force
i Incident condition
l Rolling moment
m Pitching moment
N Normal force
n Yawing moment
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

p Pressure
r Reflected condition
ref Reference quantity
S Side force
T Tangential force
w Wall condition
τ Shear pressure
∞ Freestream condition
Superscript
T Transpose
•
ˆ •=
Normalized quantity, ˆ ||•||2

I. Introduction
A. Motivation
pace situation awareness has come to the forefront of public cognizance with recent collisions between
S satellites and orbiting debris in near Earth orbits. With many crucial assets sharing near-Earth orbits
with decommissioned satellites and other space debris, it is important to develop methods that can track
these resident space objects (RSO) with high accuracy and over long periods of time to minimize probabilities
of collision.1
Many objects in the Space Catalog receive significant perturbing forces due to Earth’s atmosphere, which
produces aerodynamic forces that act on these bodies. Currently, there are two approaches to fast modeling
of aerodynamic forces: geometry simplification and time-averaging. In the case of the cylinder, for example,
special cases have been solved analytically by fixing the freestream flow to the longitudinal or lateral axes.
In reality, a cylindrical body will likely rotate and experience a range of freestream flows. Aerodynamics
on a tumbling cylinder have previously been time-averaged across a cycle,2 though these quantities are less
preferable in state-space orbit propagation compared to the instantaneous quantities that analytic expressions
can provide.
In order to develop analytic aerodynamic expressions for complex geometries, the previous research on
primitive geometries must be expanded. This investigation provides validated models of the force and
moment coefficients of such primitive shapes as the sphere, cylinder, panel, and rectangular prism. By
generating analytic models for these primitive geometries without restricting the freestream direction, more
complex geometries can be aerodynamically characterized.

B. Background
Aerodynamic forces and moments are used to catalog the momentum transfer from the flow surrounding
a moving object. All fluid flow fields adhere to three conservation laws: mass, momentum, and energy.3
When a molecule is introduced to a flow and it is immediately indistinguishable from the other molecules,
the flow is considered to be a continuum and governed by the Navier-Stokes equations, a simplification of the

2 of 14

American Institute of Aeronautics and Astronautics

 Boltzmann equation. Recently high-fidelity, analytical expressions for continuum hypersonic aerodynamics
have been developed for a general range of shapes and these have enabled fast propagation of trajectories
for various range of vehicles.4, 5, 6 Although these methods have been originally developed for hypersonic,
continuum flow, the general methodology can be extended to rarefied aerodynamics, especially free-molecular
flow which is the regime of aerodynamics experienced by RSOs at the edge of Earth’s atmosphere.
When the atmosphere is rarefied, the Boltzmann equation cannot be simplified to Navier-Stokes and
molecular mechanics dominate the flow.3 In rarefied flow, the interaction between molecules and their
collisions with the surface of the object have to be accounted and this is accomplished by numerically solving
the Boltzmann equation in a manner that is time intensive. Mean free path (λ), a dimensional quantity that
depends on the number density (related to the mass density) of the local atmosphere, captures the distance
a molecule can travel on average before some form of collision. The Knudsen Number (Kn), as shown in
Eq. (1), is the characteristic value and is a representation of the distance between molecular collisions.
λ
Kn = (1)
l
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

A high Knudsen Number is characteristic of rarefied flow, while low Knudsen number is characteristic of
continuum flow. Figure 1(a) shows the expected Knudsen numbers and the flow type for a 1-m-long vehicle
at different altitudes.3 Based on the Knudsen number classification shown in the figure, the high altitude
region where most RSO orbits reside is characterized by free-molecular flow.

Boundary
Layer
Freestream

(a) Knudsen number vs. altitude for a 1-m-long object.3 (b) Flow behavior from free-molecular to continuum.7
Figure 1. Schematic of Free-molecular regime flow and its extent in atmosphere.

Free-molecular flow is a regime characterized by extreme rarefaction of the atmosphere. A molecule can
travel long distances (relative to the characteristic length of the body) without collisions. Additionally, it
can be assumed that the flow incident on the body is independent of any effect of the particles reemitted
from the surface of the body after collision. Hence, the incident flow is entirely undisturbed by the presence
of the body.8 Figure 1(b) shows a schematic of flow in the free-molecular regime and also describes the
transition from free-molecular to continuum flow. Although the figure is not to scale, it is descriptive of the
free molecular flow and one can see that neither shock waves nor boundary layers have any effect on the flow
incident on the body in free-molecule flow.8

C. Scope of the Work

This investigation builds on previous efforts9, 10, 11 to derive analytic aerodynamic coefficients for primitive
shapes in rarefied flow. The primitives in this paper, including panels, prisms, spheres, and cylinders have
advantageous properties which simplify the process of developing the aerodynamic coefficients, such as a lack
of concavity.

3 of 14

American Institute of Aeronautics and Astronautics

 II. Methodology
A. Physical Model
Models for free-molecular flow build on energy and momentum transferred from the particles in the flow
to an elemental surface area through the particle-surface interaction. The molecule-surface interaction is
problem-dependent and is typically classified by one of two mechanisms: specular or diffuse. Specular
reflection assumes that the molecular-surface interaction is elastic during which the tangential momentum is
preserved and the normal momentum is reversed. Diffuse reflection assumes that the interaction is inelastic
and the velocity of the reflected particle is based on a Maxwellian distribution dependent of the temperature
of the reflected molecule and not the incident momentum.3 Figure 2 shows a schematic of the two types of
reflection mechanisms.
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

(a) Specular reflection (b) Diffuse reflection

Figure 2. Models for molecular-surface interactions.12

The aerodynamic and heating properties due to the molecular-surface interaction is a function of the mo-
mentum and energy of the incident molecule, the reflected molecule, and the wall temperature. The incident
molecule has properties based on freestream conditions, while the surface can be assumed to have a certain
wall temperature. The properties of the reflected molecule, however, can be highly problem-dependent.
The reflected molecule’s properties are often characterized in terms of accommodation coefficients. The
normal momentum accommodation coefficient (σN ) describes the change in normal momentum due to the
molecular-surface interaction, while the tangential momentum accommodation coefficient (σT ) describes the
momentum change in the tangential direction. These quantities are described in Eqs. (2) and (3).8 Note
that p is pressure (normal to the surface) and τ is the shear pressure and they designate conditions of the
incident molecules (subscript i), the reflected molecules (subscript r), or the wall (subscript w). Represented
values σN and σT can be found in found in Ref. 13. Specular reflection simplifies to σN = σT = 0, while
diffuse reflection is σN = σT = 1.3
pi − pr
σN = (2)
pi − pw
τi − τr
σT = (3)
τi

B. Analytic Aerodynamic Coefficients

Based on the above nomenclature, Schaaf and Chambre8 theorized that the pressure coefficient (Cp ) and the
shear pressure (Cτ ) coefficients on an elemental area inclined to the freestream at an angle θ can be described
as shown in Eqs. (4) and (5), which was later presented in modified format by Regan and Anandakrishnan.3
The freestream molecular speed ratio, s, is analogous to the Mach number and defined in Eq. (6), where V∞
is the freestream velocity, R is the gas specific constant, and T∞ is the freestream temperature. Similarly,
Tw is the wall temperature. Eqs. (4) and (5) can be integrated over the surface of a general shape to provide
the aerodynamic force and moment coefficients. It should be noted that unlike previous works regarding
analytic hypersonic continuum aerodynamic coefficients by Grant,4, 5, 6 the tangential momentum component
cannot be ignored in the case of free-molecular flow aerodynamics.

4 of 14

American Institute of Aeronautics and Astronautics

 " r !
1 2 − σN σN Tw 2
Cp = 2 √ s sin (θ) + e−(s sin θ)
s π 2 T∞
(   r ) # (4)
2 1 σN πTw
+ (2 − σN ) (s sin θ) + + s sin θ (1 + erf (s sin θ))
2 2 T
−σT cos θ h −(s sin θ)2 √ i
Cτ = − √ e + πs sin θ (1 + erf (s sin θ)) (5)
s π
V∞
s= √ (6)
2RT∞
In the free molecular regime, the mean free path of the gas is many times larger than the characteristic
dimension of the body. The gas flow is assumed to be of an infinite extent, which is valid for high altitude
considerations. A consequential assumption of the extreme rarefaction is that the indent flow is entirely
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

undisturbed by the presence of the body and, at the body, the approaching gas is assumed to be in Maxwellian
equilibrium. The formulation for the pressure and shear coefficients in Schaaf and Chambre utilize two
accommodation coefficients, σN and σT , which are averaged quantities, so the models presented in Eqs. 4
and 5 are valid so long as these coefficients adequately accommodate the incident and reflected energies.
Specifically, the accommodation coefficients are valid in gases which have significantly less vibrational energy
compared to translational and rotational energy.
The challenge of finding analytical aerodynamics is integrating the pressure coefficient (Eq.(4)) and shear
pressure coefficient (Eq.(5)) expressions over the surface of a body. The elemental force (df ) applied over an
differential area (dA) consists of pressure and shear pressure components, with the pressure working along
the normal vector (n̂) and the shear pressure working along the tangential vector (t̂) as seen in Eq. (7). Note
that the normal vector is defined as positive outwards. Additionally, using conventional flight dynamics
body axes,14 the normalized velocity vector (V̂∞ ) can be defined in terms of the angle of attack (α) and
the sideslip angle (β) as shown in Eq. (8). Then the force and moment coefficients in the x, y, and z body
directions can be found by surface integration as shown in Eqs. (9) and (10). The force coefficients consist of
axial (subscript A), side (subscript S), and normal (subscript N ) forces and the moment coefficients consist
of rolling (subscript l), pitching (subscript m), and yawing (subscript n) moment coefficients. The position
vector of the element (r) is defined from the origin of the coordinate system about which the forces and
moments are taken - in this case the body frame of the vehicle.


df = Cp n̂ + Cτ t̂ dA (7)
T
V̂∞ = − [cos α cos β sin β sin α cos β] (8)
 
CA ZZ
1
 CS  = df (9)
 
Aref S
CN
 
Cl ZZ
1
 Cm  = r × df (10)
 
Aref · lref S
Cn

C. Parameterization Schema
The integrals in Eqs. (9) and (10) are performed over the surface of the body, which must be parametrized
in a fashion to give the position vector of every point on the surface. The position vector for this method
must be parametrized by two independent variables u and v. Additionally, due to the convention used in this
method, u and v must be chosen such that ru × rv is the non-normalized normal vector pointed outwards
∂r ∂r
from the surface, where ru = ∂u and rv = ∂v . Similarly, the differential area, dA, can be shown to be
dA = ||ru × rv ||2 .
The outward unit normal vector is the cross-product of the partial of the position vector with respect to
u and the partial of the position vector with respect to v as shown in Eq. (11). Equation (12) shows that
the unit normal vector and the unit velocity vector can be used to define the local flow angle θ that appears

5 of 14

American Institute of Aeronautics and Astronautics

 in the pressure and shear pressure coefficient expressions. The definition of the local flow angle also helps
resolve the definition of the unit tangential vector, t̂, that defines the direction in which the shear pressure
coefficient is acting. Unlike the normal vector, there are infinite choices for the tangential vector in the plane
tangent to the elemental surface. For example, either ru or rv is an acceptable choice for a tangential vector.
But based on definition of the shear pressure, which is based on the shear force, the proper vector is opposite
of the local flow direction. This is described by Eq. (13). These quantities and the relationships between
them are illustrated in Figure 3.
ru × rv
n̂ = (11)
||ru × rv ||2
sin θ = −V̂∞ · n̂ (12)
 
n̂ V̂∞ · n̂ − V̂∞
t̂ = r  2 (13)
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

1 − V̂∞ · n̂

A variety of shapes are analyzed in this paper for their aerodynamic coefficients, and while there are
many methods to parameterize a surface, the following methods have yielded integrable analytic aerodynamic
coefficients for rarefied flow.

Figure 3. Vector diagram illustrating the relationships between relevant quantities.

1. Triangular Panel
A triangular panel can be created from three position vectors, 4ABC = (rA , rB , rC ), to the vertices where
A, B, and C are defined counterclockwise about the origin. From these three vectors, two can be defined as
r1 = rB − rA and r2 = rC − rA , which are in the plane of the panel. The parameterization of the triangular
panel is given by Eq. (14). A visual of the parameterization of the triangular panel is given in Figure 4.
Though the parameterization given below describes a triangle, the aerodynamic coefficients for the panel are
not dependent on the geometry of the panel.

r = u r1 + v r2 (14)
The limits on u and v in the (u, v)-plane are the positive axes and the line u + v = 1. By defining the
vectors to the vertices of the triangle in a counterclockwise sense, the normal vector is guaranteed positive.

2. Rectangular Panel
A rectangular panel can be defined by the vector to the center of the rectangle and two vectors to the
long and short sides of the triangle. Let rc be the vector to the center and a and b be the vectors to
each side of rectangle. The vectors should satisfy the triple product inequality (a × b) · rc ≤ 0 so that
the normal vector is outward facing. The square panel can be parameterized in the same way, with the
condition that ||a||2 = ||b||2 . The parameterization of the rectangular panel is given by Eq. (15) and u and
v are independently bounded on the interval [−1, 1]. The rectangular panel parameterization is illustrated

6 of 14

American Institute of Aeronautics and Astronautics

 Figure 4. Illustration of the surface parameterization of the triangular panel.

in Figure 5. The rectangular prism is the composition of six rectangular panels and the aerodynamics of
each panel are combined by the method described in Section F.
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

r = rc + u a + v b (15)

Figure 5. Illustration of the surface parameterization of the rectangular panel.

3. Sphere
The sphere is is defined as the collection of points in 3-dimensional space that are equidistant from a center.
In this case, the center of the sphere is assumed to be the center of the reference frame. Since drag is the only
aerodynamic force on the sphere, a solution for the drag on the sphere can be calculated in the wind frame
and then rotated through α and β to find the body-axis coefficients. In this case, the sphere has a single
parameter, the radius (R), and is defined in spherical coordinates as shown in Eq. (16). The bounds on the
parameters u and v are [0, π] and [0, 2π], respectively. Figure 6 illustrates this parameterization scheme.
T
r = R [sin u cos v sin u sin v cos u] (16)

Figure 6. Illustration of the surface parameterization of the sphere.

7 of 14

American Institute of Aeronautics and Astronautics

 4. Cylinder
The cylinder is a rectangle of revolution and is characterized by its radius (R) and length (l). Let the axis
of the cylinder coincide with the first axis of the reference frame and the center of the frame be located at
half the length, then the cylindrical hull can be parameterized by Eq. (17), where u is bounded by [0, 2π]
and v is bounded by [−l/2, l/2]. This surface parameterization is illustrated in Figure 7.
T
r = [v R cos u R sin u] (17)
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

Figure 7. Illustration of the surface parameterization of the cylinder.

The ends of the cylinder are circles parameterized by Eq. (18) with u bounded by [0, R] and v bounded
by [0, 2π]. The sign of the first component of the position vector is equal to the sign of the outward normal
vector of the circle.
 T
l
r= ± u cos v u sin v (18)
2

D. Concavity Constraint
2
Each of the above geometries satisfies the conditions of convex functions, ∂u2 r ≤ 0, ∂v2 r ≤ 0, ∂uv r ≤ 0, thereby
simplifying the process of defining the wetted area of the object. On a non-concave object, there are always
two distinct regions of the surface, the wetted region and the shadowed region. The boundary between these
regions occurs when the sine of the local freestream angle transitions from positive to negative. For a non-
concave shape, this condition is equivalent to sin θ = 0. Using a single condition to describe the boundary
of the wetted region reduces the complexity of solving the integrals of elemental forces by simplifying the
bounds of integration.

E. Surface Parameterization Bounds

On convex geometries, the surface is separated into two disjoint regions where the interface between these
regions occurs when the freestream velocity is orthogonal to the local normal direction. Choosing which
region is the wetted region can be found by calculating the local angle at a single point, however in all cases
in this paper a priori knowledge of the geometry distinguished the wetted area from the shadowed area. The
parameterizations of the objects listed above are not unique, there are an infinite number of parameterization
schema for a single geometry, however they were selected for simplicity in the limits of integration for the
hp iT
wetted area. For example, the sphere could be parameterized by r = R2 − (y 2 + z 2 ) y z , however
the bounds on the inner integral become circular segments. In the (y, z)-plane the region of integration is a
circle, whereas in the (u, v)-plane the region of integration is a rectangle, which reduces the computational
effort to evaluate the integrals. Each of the above geometries are bounded in their (u, v)-planes by a rectangle,
except the triangular panel which is bounded by a right triangle.
The wetted regions of these geometries are not as simple as a rectangle or a triangle. Though the area
of the wetted region is dependent on the direction of the freestream velocity, the shape of the region is
consistent. For example, the wetted area of the cylinder and the sphere are rectangular subregions.

8 of 14

American Institute of Aeronautics and Astronautics

 F. Application of the Heaviside Step Function
Some geometries are piecewise parameterized, such as the rectangular prism where each of the faces is
individually parameterized. Since each panel’s contribution to the overall force and moment coefficients
depends on whether it is wetted by the flow, a novel approach used in this work is to introduce the Heaviside
step function. The Heaviside function behaves like a switch toggling the force and moment coefficients of
each panel. The prism and the cylinder are both piecewise parameterized, and the general form for the
application of the Heaviside function is shown in Eq. (19).
   
CA C A  
 X
 CS  =  CS  H −V̂∞ · n̂i (19)
 

CN i CN i
The same approach is used for the force and moment coefficients. The argument of the Heaviside function
is equivalent to the sine of the local flow angle. For all negative inputs to the Heaviside function, the output
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

is a zero, so the shadowed panels make no contribution to the overall force and moment coefficients. Likewise,
the panels that are wetted by the flow have a positive sine of the local flow angle and their force and moment
coefficients contribute to the overall coefficients. The summation of these contributions is a result of the
linearity of the integral operator.
Although panel methods have been developed in the past, the description of the various panels is included
in this paper because it is a primary component of the method used to analyze the rectangular prism and
the cylinder. The approach to solving for the aerodynamics in this investigation is novel in its application
of the Heaviside step function to closed piecewise parameterized surfaces. The method described below is
applicable beyond the shapes described so long as the shape is closed, piecewise, and convex.

III. Results
A. Validation Method
The aerodynamic analytical expressions derived by the method above are validated against industry-standard
free-molecular aerodynamics prediction tools like Direct Simulation Monte Carlo (DSMC) codes. DSMC
codes solve the Boltzmann equations directly, but are computationally expensive. However, they serve as
a good way to check the accuracy of the analytical expressions. The specific DSMC code used here for
comparison is the NASA-developed DSMC Analysis Code (DAC) that has been tested and used in the field
extensively.15, 16, 17 The version of the code used in the analysis is DAC97-N from October 2012.
DSMC tools like DAC have several types of input files and parameters that have to be adjusted in order
to achieve realistic aerodynamic predictions. On the other hand, the analytical expressions developed from
the Cp and Cτ expressions in Eqs. (4)- (5) are only a function of the molecular speed ratio, wall temperature,
freestream temperature, and the dynamics of the interaction that is captured by the normal and tangential
momentum accommodation coefficients. Hence, one can easily see some of the advantages in complexity
offered by analytical aerodynamic expressions.

B. Convex Geometries
1. Sphere
Analytical expressions for the drag coefficient (CD ) of a sphere were developed using the methodology
above. These expressions are not novel, since Schaaf and Chambre8 also developed similar expressions for
free-molecular flow over a sphere. But this problem is a very simple first step to test the accuracy of the
analytical prediction methodology, since all aerodynamic force and moment coefficients are zero except for the
drag force for flow over a sphere with the center of pressure at the center of gravity of the sphere. Expressions
for the force coefficients, moment coefficients, and stability derivatives of the sphere can be found in this file
embedded in the paper (analyticExpressions.nb) and are presented in a Mathematica notebooka .
Comparisons between DAC results (colored dots) and the analytical expression (lines) are given in Fig. 8.
Diatomic nitrogen was used as the freestream gas in the simulation. Variation of the drag force coefficient
over a range of atmospheric speeds (related to the molecular speed ratio), wall temperatures, and freestream
a For best results use Adobe Acrobat readers The embedded file can be downloaded from the link.

9 of 14

American Institute of Aeronautics and Astronautics

 temperatures are shown in the figure to illustrate the range of conditions for which the analytic model is
valid. Note that the pure specular and pure diffuse bounds are shown. Aerodynamic parameter values for
conditions other than these two extremes will fall in the shaded zone. The baseline flow parameters are
Tw = 300K, T∞ = 973K, and V∞ = 7500 m/s, and each graph shows the effect of shifting these values. The
specular and diffuse flow relations are shown for each of the three sensitivity cases.
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

(a) Atmospheric Speed (m/s) (b) Wall Temperature (K) (c) Freestream Temperature (K)
Figure 8. Validation of analytical expressions for sphere in free-molecular flow.

Figure. 8 shows that the analytical expressions and the DAC predictions match very well. The slight
discrepancies between the two can be attributed to adjustments that need to be made to the DAC compu-
tation grid. In the case of the sphere, drag is the only force acting on the body, so the aerodynamic force is
resolved into the wind frame. All proceeding geometries resolve the aerodynamic forces and moments into
the vehicle body axes.

2. Cylinder
The analytic force and moment coefficients on the cylinder were derived by taking advantage of the axial
symmetry of the body. This means that the freestream can be restricted to the body’s (x, y)-plane, effectively
collapsing the angle of attack (α) to zero and letting the side slip angle (β) vary on the interval [0, π/2].
Every freestream orientation can be x-rotated to lie in the first quadrant of the (x, y)-plane, so these results
can be x-rotated to any other freestream orientation. The analytic results of this equation, for the same
conditions as in Section B.1 above and a length-to-diameter ratio of 2.18, with the side slip angle varied on
the range [0, π/2], are compared to DAC numerical simulation data in Figure 9. The two cases of specular
and diffuse reflections are plotted together to illustrate the range of valid solutions which the analytical
model can be applied.
The specular results show the best agreement between DAC and the analytic model, although there is
still good agreement between the two types of analysis for the diffuse results. The results of the two analyses
differ in cases where there is a small angle between the freestream and the circular face of the cylinder. This
effect is apparent in each of the above plots, and most noticeable in Figure 9(f). A small local freestream
angle corresponds to significant tangential forces, which explains the “tails” on the yawing moment graph.
The effect is also evident on the diffuse curves in the force coefficient graphs.
The tails on the graphs of the yawing moment in the cylinder and the prism under diffuse reflections are
indicative of a discrepancy between the analytic model and DAC. The DAC data tend to show very little
variation on the moment coefficients on these geometries, though these coefficients vary significantly with
the direction of the freestream velocity. For the case of the rectangular prism, which is covered in Sec. C. 2,
the moment coefficients are zero when the freestream flows along one of the body axes, which is shown in the
DAC data. If the freestream velocity’s orientation is perturbed even by a small margin, there is significant
tangential forces on the exposed surfaces, causing the object to rotate. The object will continue to rotate
until the moments due to the tangential forces are balanced, which indicates that the moment coefficient
ought to change with respect to the orientation of the freestream. The analytic models feature this change
in the moment coefficients, while the DAC results do not. Future investigation will determine why the DAC
data does not show such behavior for diffuse reflections.

10 of 14

American Institute of Aeronautics and Astronautics

 (a) Axial Force (b) Side Force (c) Normal Force
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

(d) Rolling Moment (e) Pitching Moment (f) Yawing Moment

Figure 9. Validation of the aerodynamic coefficients of the cylinder.

C. Planar Geometries
1. Panel
Although a simple shape, a panel is considered in this study since it is a fundamental geometry and the
primary shape of stereolithography files, which can be produced by most Computer-aided design (CAD)
software. The analytic results of this equation are for the same conditions as those introduced in Section
B.1, with the angle of attack (α) and side slip angle (β) independently varied on the range [0, 2π]. The
analytical results are once again compared with DAC numerical simulation data. In Fig. 10, the force
coefficients of the panel are illustrated with the angle of attack fixed.

(a) Axial Force (b) Side Force (c) Normal Force

Figure 10. Validation of the aerodynamic coefficients of the panel.

2. Rectangular Prism
The analytic force and moment coefficients on the cylinder were derived by taking advantage of the piecewise
composition of the geometry. By applying the Heaviside function to toggle a panel’s contribution depending
on if it is wetted by the flow or not, the force and moment coefficients are calculated by the sum of the
contributions from all panels. The analytic results of this equation, for the same conditions as in Section

11 of 14

American Institute of Aeronautics and Astronautics

 B.1, are compared with the DAC data. Graphs of the coefficients in the case of Tw = 300K, T∞ = 1100K,
and V∞ = 7500 m/s, equal side lengths, and α = 10◦ are given in Figure 11.

(a) Axial Force (b) Side Force (c) Normal Force

Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

(d) Rolling Moment (e) Pitching Moment (f) Yawing Moment

Figure 11. Validation of the aerodynamic coefficients of the rectangular prism.

D. Analysis
1. Model Accuracy
The equations for the pressure and shear coefficients given in Eqs. (4)- (5) are models of rarefied flow and
are not derived from the governing equation for the flow. To quantify the goodness-of-fit for these models,
the R2 values for the force and moment coefficients are tabulated relative to the DAC results in Table 1.
The median percent error is given in Table 2 as an additional metric for the accuracy of the analytic models.
In cases where the theoretical aerodynamic coefficient is zero, the statistical fit values are not provided. The
diffuse moment coefficients for the cylinder and the prism are not included in the model accuracy analysis
due to the still unresolved reason disagreement between the analytic model and DAC.
Table 1. R2 values for DAC results relative to the analytic model.

Specular Diffuse
Geometry CA CS CN Cl Cm Cn CA CS CN Cl Cm Cn
Sphere 0.667 - - - - - 0.813 - - - - -
Cylinder 0.999 0.999 - - - - 0.994 0.999 - - - -
Panel 0.999 - - - - - 0.999 0.996 0.999 - - -
Rectangular Prism 0.999 0.999 0.994 - - - 0.999 0.997 0.999 - - -

2. Computational Comparison
The large compute time for aerodynamic coefficients is usually a major hindrance for rapid exploration of
designs of rarefied flow vehicles or predicting the trajectories of resident space objects with high accuracy.
The advantages of analytical expression can be easily seen in Table 3 that lists the run times for the analytical
expressions for various shapes. For reference, the analytical computations were tested on a 2.8 GHz Intel

12 of 14

American Institute of Aeronautics and Astronautics

 Table 2. Median absolute percent errors for DAC results compared to the analytic model.

Specular Diffuse
Geometry CA CS CN Cl Cm Cn CA CS CN Cl Cm Cn
Sphere 0.137 - - - - - 0.380 - - - - -
Cylinder 0.085 0.069 - - - - 0.262 0.265 - - - -
Panel 0.153 - - - - - 0.078 0.140 0.231 - - -
Rectangular Prism 0.260 0.089 0.711 - - - 0.213 0.352 0.322 - - -

Core 2 Duo processor. In general, the DAC run times vary based on the resolution of the model, but for
comparison DAC needed 16 minutes on a 2.2 GHz duo-core processor to analyze the simple sphere. The
analytic expression for drag on a sphere ran 870,000 times faster than DAC.
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

Table 3. Mean run time comparison of analytic and numerical methods for determining aerodynamic coefficients.

Geometry Analytic Runtime (ms)

Sphere 1.1
Cylinder 22
Panel 2.4
Rectangular Prism 23

IV. Conclusions
Fields such as space situational awareness could benefit significantly from fast and accurate predictions
of aerodynamic force and moment coefficients of objects in near-Earth orbits. The flow in near-Earth orbits
is typically described as free-molecular flow and current methods of predicting aerodynamic coefficients in
this regime require very computationally intensive methods that cannot be used in conjunction with fast
trajectory propagations tools required for space situational awareness.
However, recent advances in symbolic manipulation software allows one to develop analytical expres-
sions for aerodynamic quantities without making significant assumptions or simplifications. This analytical
aerodynamic expression method has been used for hypersonic, continuum flow in the past and the current
work advances this method for free-molecular flow. These expressions are validated against state-of-the-art
engineering tools, specifically DAC. The results of the analytical method and the numerical tool agreed very
well for force coefficient expressions for spheres, cylinders, triangular panels, and rectangular prisms and had
good agreement over large orientation angle ranges for the moment coefficients. This initial study of analyt-
ically derived rarefied flow aerodynamic coefficients shows promise and future work will include derivation
of analytic expressions for a larger range of primitive shapes, the composition of primitive geometries, and
expressions for concave and arbitrary axisymmetric geometries.

Acknowledgments
The work in this paper was funded by Air Force Research Laboratory Contract No. FA9453-13-C-0205.
The authors want to thank Dr. Michael Grant, whose doctoral work in developing analytical expressions for
hypersonic aerodynamics serves as the basis for the analytical aerodynamics method for free-molecular flow
developed in this paper, and Gerald LeBeau and Katie Boyles at the NASA-Johnson Space Center for their
help with the DAC tool.

References
1 Bar-Shalom, Y., Li, X., and Kirubarajan, T., Estimation with Applications to Tracking and Navigation, John Wiley &

Sons, Inc., New York, NY, 2001.

2 Pilinski, M. D., Argrow, B. M., Palo, S. E., and Bowman, B. R., “Semi-Empirical Satellite Accommodation Model for

Spherical and Randomly Tumbling Objects,” Journal of Spacecraft and Rockets, Vol. 50, No. 3, July 2013, pp. 556–571.

13 of 14

American Institute of Aeronautics and Astronautics

 3 Regan, F. J. and Anandakrishnan, S. M., Dynamics of Atmospheric Re-Entry, American Institute of Aeronautics and

Astronautics, Reston, VA, 1993.

4 Grant, M. J. and Braun, R., “Analytic Hypersonic Aerodynamics for Conceptual Design of Entry Vehicles,” AIAA

2010-1212, 48th AIAA Aerospace Sciences Meeting, Orlando, FL, 2010.

5 Grant, M. J. and Braun, R., “The Extension of Analytic Hypersonic Force Coefficients for Conceptual Design Using the

Divergence Theorem,” AIAA 2012-4580, AIAA Atmospheric Flight Mechanics Conference, Minneapolis, MN, 2012.
6 Grant, M. J., “The Construction of Analytic Hypersonic Pitch Moment Coefficients Using a Curl Transformation,” AIAA

2013-0225, 51st AIAA Aerospace Sciences Meeting, Dallas, TX, 2013.

7 Vallander, S. V., “Collection 1,” Aerodinamika Razrezhennykh Gazov , 1963.
8 Schaaf, S. A. and Chambre, P. L., Flow of Rarefied Gases, Princeton University Press, 1961.
9 Wang, C.-T., “Free Molecular Flow over a Rotating Sphere,” AIAA Journal, Vol. 10, 1972, pp. 713–714.
10 Yoshizawa, Y., “Examination of the Usefulness of the Restricted Hermite Expansion Method for Rarified Gas Dynamics

Problems,” Progress in Astronautics and Aeronautics, Vol. 74, 1980, pp. 371–385.
11 Stalder, J. and Zurick, V., “Theoretical Aerodynamic Characteristics of Bodies in a Free-Molecule-Flow Field,” Tech.

rep., NACA TN 2423, 1951.

12 Davidson, M. W., “Specular and Diffuse Reflection,” Electronic, June 2006.
13 Hurlbut, F. C., Encyclopedia of Applied Physics, chap. Rarefied-Gas Dynamics, Wiley-VCH Verlag GmbH & Co KGaA,
Downloaded by GEORGIA INST OF TECHNOLOGY on February 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.2014-0728

2003, pp. 97–115.

14 Etkin, B., Dynamics of Atmospheric Flight, Dover Publications, Meneola, NY, 2000.
15 LeBeau, G. and Lumpkin III, F., “Application highlights of the DSMC Analysis Code (DAC) software for simulating

rarefied flows,” Computer Methods in Applied Mechanics and Engineering, Vol. 191, 2001, pp. 595–609.
16 Boyles, K., LeBeau, G., Lumpkin III, F., and Blanchard, R., “The Use of Virtual Sub-Cells in DSMC Analysis of Orbiter

Aerodynamics at High Altitude Upon Reentry,” AIAA 2003-1030, 41st Aerospace Sciences Meeting, Reno, NV, 2003.
17 LeBeau, G., “A User Guide For the DSMC Analysis Code (DAC) Software,” Tech. Rep. DAC97-N, NASA JSC, 2012.

14 of 14

American Institute of Aeronautics and Astronautics

View publication stats