3 PDF
3 PDF
PII: S0997-7546(15)00051-5
DOI: http://dx.doi.org/10.1016/j.euromechflu.2015.04.003
Reference: EJMFLU 2887
Please cite this article as: G. Zuppardi, L. Morsa, R. Savino, M. Sippel, T. Schwanekamp,
Rarefied aerodynamic characteristics of aero-space-planes: A comparative study of two
gas–surface interaction models, European Journal of Mechanics B/Fluids (2015),
http://dx.doi.org/10.1016/j.euromechflu.2015.04.003
This is a PDF file of an unedited manuscript that has been accepted for publication. As a
service to our customers we are providing this early version of the manuscript. The manuscript
will undergo copyediting, typesetting, and review of the resulting proof before it is published in
its final form. Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
Rarefied aerodynamic characteristics of aero-space-planes: a
comparative study of two gas-surface interaction models
Gennaro Zuppardia(*), Luigi Morsaa, Raffaele Savinoa
Abstract:
Accurate computations of aerodynamic forces are very important for the design of aero-space-planes.
In rarefied flow, the computation of momentum and energy transfer between the flow and the
vehicle's surface occurs in two steps: incidence and re-emission of gas molecules. While incidence is
well understood, the re-emission process is still today not yet completely clear, thus many models
have been developed. In the present paper, the effects of the re-emission models by Maxwell and by
Cercignani-Lampis-Lord have been compared by means of direct simulation Monte Carlo (DSMC)
codes. Two different study cases have been considered: a complete aero-space-plane and a wing
profile. Computer simulations have been carried out using two DSMC codes to investigate hypersonic
flows at an altitude of 120 km where for both the wing section and the vehicle the flow is in
transitional regime. The results pointed out that the influence of the interaction models, considering
specular, semi-specular and full diffusive re-emission is pretty strong, while the Cercignani-Lampis-
Keywords:
________________________
(*) e-mail: [email protected]
1. Introduction
The computation of aerodynamic characteristics of aero-space-planes in high altitude flight, or in rarefied flows,
relies on the Direct Simulation Monte Carlo (DSMC) method [1], [2], [3]. A correct evaluation of the aerodynamic
force and moments of a spacecraft along its ascent and/or re-entry path provides useful information for the proper
design of the propulsion and attitude control systems, for instance to ensure trim, stability and manoeuvrability. In
rarefied flow, accurate computations rely on a proper choice of the gas-surface interaction model. This influences the
computation of momentum and energy exchanged with the surface and therefore the aerodynamic forces, moments and
heat flux.
It is well known that, in rarefied flow, the gas–surface interaction is considered as composed of two different
mechanisms: incidence and re-emission. While incidence is well understood and the related computation has been
proved to be accurate, the re-emission phase is still today not yet completely understood; thus many models (Maxwell
[1], [2], [3], Cercignani-Lampis [1], [2], [3], Schamberg [4], [5], Nocilla [6]) have been proposed, trying to describe the
molecule remission from the surface of the body. Every model is based on simplifying assumptions and relies on
“engineering” parameters (or accommodation coefficients) providing against the ignorance about the physical
mechanism of the re-emission process. Lord [1], [2], [7] provided a DSMC implementation of the Cercignani-Lampis
Most of the DSMC computations currently are based on the assumption of full diffusive interaction. According to
Bird [1], [2], this assumption should be reviewed whenever a smooth surface has been exposed for long time to ultra-
high vacuum, a surface is heated, the weight of the gas molecule is much smaller than the weight of the surface
material, the translational energy of the impinging molecules is larger than several electron-volts. Knetchel and Pitts [8]
verified experimentally that the accommodation coefficients of the normal and tangential momentum components
depend on a number of factors: gas composition, impact energy of the impinging molecules, surface temperature,
surface material, roughness of the surface, angle of incidence. These measurements are particularly interesting because
The present paper is the follow-on of a former article by Zuppardi [9] where the Maxwell and the Schamberg
models were compared with DSMC computations carried out only at the impact point on an elementary surface
changing the incidence angle. In the present work, the Maxwell and CLL models are compared considering:
1. the impact point on an elementary surface, for a basic, direct comparison of the re-emitted quantities,
2. a wing section of SpaceLiner, an advanced visionary concept of suborbital, hypersonic, winged passenger
transport, currently under investigation at DLR, Space Launcher System Analysis [10], [11], [12]. In this
2
case computations are aimed at the evaluation of the influence of the two interaction models on global and
3. the complete SpaceLiner vehicle in clean and flapped configurations, to assess the effect of the interaction
model on the aerodynamic coefficients, on the lift-to-drag ratio and on the pitching moment around the
center of gravity of the vehicle. The latter is particularly important to assess the vehicle's longitudinal
Two dimensional and three dimensional computer tests were carried out by the DS2V-64 bits [13] and DS3V [14]
codes, including the Maxwell and CLL models with specular, semi-specular and full diffusive molecule-surface
interactions. The models assume constant accommodation coefficients on the whole surface, i.e. independent from the
incidence angle. Unfortunately, the lack of experimental data in real conditions does not make possible a proper
evaluation of the models; thus the scope of the paper is only to verify and quantify the differences in results obtained
with the two interaction models and with different accommodation coefficients.
The influence of the gas-surface interaction on the aerodynamic coefficients has been already considered for simple
A direct, quantitative comparison of the results from a DSMC code implementing the Maxwell and CLL models has
been already carried out by Padilla and Boyd [18], who investigated aerodynamic forces and heat transfer on a flat plate
in a space environment, for a range of angle of attack and with several accommodation coefficients. To the authors
knowledge, a systematic study on complex configurations such as winged spacecrafts, has not been considered before.
As discussed before, the gas-surface interaction includes incidence and re-emission of the molecule upon and from
the surface. Molecules impinging upon the surface are considered in equilibrium at temperature Ti and molecules re-
emitted from the surface are considered in equilibrium at temperature Tr that can be different from the surface
temperature Tw. An evaluation of the level at which Tr is adjusted toward Tw is provided by the thermal
3
Ei − E r
αE = (1.a)
Ei − E w
E is the molecule transitional energy, subscripts i, r and w are for incident, re-emitted and re-emitted at wall
temperature. According to Bird [1], [2], the energy accommodation coefficient can be approximated by the fraction of
molecules reflected diffusely. The accommodation coefficients are defined also in terms of the normal (αn) and
pi − p r
αn = (1b)
pi − p w
τi − τ r
αt = (1.c)
τi
All accommodation coefficients range from 0 for no accommodation (specular reflection) to 1 for complete
accommodation (full diffusive re-emission). Knetchel and Pitts [8] obtained the following formulae, evaluating αn and
where ϕ is the complementary of the incidence angle ϑ (ϕ=π/2- ϑ ). Equations 2 are least square curves fitting
experimental results obtained for satellites or spacecrafts close to Earth re-entry. For completeness, Fig. 1 shows the
profiles of αn and αt as functions of ϑ and clearly verifies the strong dependence of the momentum accommodation
4
0 10 20 30 40 50 60 70 80 90
1 1
αt
0.8 0.8
αt
0.6 0.6
0.4 αn αn 0.4
0.2 0.2
Ei=10 eV
Ei=40 eV
0 0
0 10 20 30 40 50 60 70 80 90
ϑ [deg]
Fig. 1. Profiles of normal (αn) and tangential (αt)momentum accommodation coefficients as functions
The DSMC implementation of both the Maxwell and the Cercignani-Lampis-Lord model starts from the
computation, for each collision, of the incident velocity Vi (ui, vi, wi) of each molecule impinging upon the surface.
According to Bird [1], [2], the incident velocity is composed of the free stream velocity V∞ (V∞x, V∞y, V∞z) and the
molecular thermal velocity C (U, V, W), set in turn at random from the most probable molecular velocity (c) at
temperature T∞:
k
c(T∞ ) = 2 T∞ (3)
m
ui = V∞x + U (5.a)
vi = V∞y + V (5.b)
wi = V∞z + W (5.c)
5
The component ui is normal to the surface, the components vi and wi are tangent to the surface.
According to Maxwell, the reflection is governed by the “classical” specular and/or diffusive models. The specular
model assumes that the normal component of the reflected velocity is reversed: ur = -ui, and the tangential components
do not change: vr=vi, wr=wi. In the DSMC implementations, the diffusive model is considered always full diffusive at
the surface temperature. Components of the reflected velocity are computed at random from the most probable
U r = c(Tw ) ln R (6.a)
The normal and tangential components of the stress are linked to the incident and reflected momentum; the resultant
p = (1 + f )pi + (1 − f )p w (7.a)
τ = (1 − f ) τi (7.b)
where f is the fraction of particles re-emitted specularly; therefore a specular re-emission is defined by f=1 and a full
Besides the Maxwell specular and full diffusive models, the Cercignani-Lampis model seems to be the most
successful for DSMC application. The model satisfies the reciprocity condition, i.e. the principle of detailed balance;
this means that there is a relationship between the probability of a gas surface interaction with a set of incident and
reflected velocities and the probability of the inverse interaction. This model is based on the accommodation coefficient
for the kinetic energy (αEn) associated with the normal component of velocity, and the tangential momentum
accommodation, respectively. The model assumes that there is no coupling between the normal and tangential
components of the velocity during the reflection process. It provides the probabilities that an incident molecule with
normal velocity component Ui is reflected with normal velocity component Ur. The tangential component Vi (or Wi) is
6
⎛ 2U r ⎞ ⎧⎪ 2(1 − α En )1 2 U i U r ⎫⎪ ⎧ U 2 + (1 − α ) U 2 ⎫
⎪ r En i ⎪
P( U r , U i ) = ⎜⎜ ⎟⎟I 0 ⎨ ⎬ exp⎨ ⎬ (8.a)
⎝ α En ⎠ ⎪⎩ α En ⎪⎭ ⎪⎩ α En ⎪⎭
⎡ {V − (1 − α t )1 / 2 Vi }2 ⎤
P( Vr , Vi ) = ( πα t ) −1 2 exp ⎢ − r ⎥ (8.b)
⎣⎢ αt ⎦⎥
th
where the components of velocity are made non-dimensional by c(Tw) and I0 is a Bessel function of zero order. Eq.8.b
holds also for P(Wr, Wi). Considering that the energy accommodation coefficient is related only to the kinetic energy,
in the present computations, αEn is approximated by αn: αEn≅αn. As the Cercignani-Lampis model relies on the
accommodation coefficients αn and αt, it overcomes the limitation of the unreal Maxwell specular and full diffusive
models.
The DSMC implementation by Lord consists in the following equations computing the components of the reflected
velocity:
Vr = (1 - α )1 2 Vi + r cos θ (9.b)
Wr = r sin θ (9.c)
θ = 2πR (9.d)
r = − α ln(R ) (9.e)
and α=αn for the normal component (Ur), α=αt(2-αt) for the tangential components (Vr and Wr).
The CLL model verified [19] to provide a realistic boundary condition with incomplete accommodation. In
addition it reduces to the specular and to the diffusive model when both αn and αt are zero or unity, respectively. The
7
p = pi + pr (10.a)
τ = τi − τ r (10.b)
It is well known that the DSMC method [1], [2], [3] is currently the only tool for the solution of rarefied flow fields
from slip flow to free molecular regimes. The Navier-Stokes equations fail in low density regimes due to the failure of
the “classical” laws by Newton, Fourier and Fick, computing the transport parameters.
DSMC considers the gas as made up of particles (molecules and atoms). It is based on the kinetic theory of gases
and computes the evolution of millions of simulated particles, each one representing a large number (say 1015) of real
particles in the physical space. Intermolecular and molecule-surface collisions are taken into account. The
computational domain is divided in cells. The cells are used for selecting the colliding molecules and for sampling the
macroscopic fluid-dynamic quantities. The most important advantage is that the method does not suffer from numerical
instabilities and does not rely on similarity parameters, like the Mach and the Reynolds numbers. On the other hand, the
method is inherently unsteady; a steady solution is achieved after a sufficiently long simulation time.
1) a code, written specifically for the purpose of this paper, including the equations 3 to 10 for the computation of
the pressure and tangential stress at the molecule impact point on an elementary surface, changing the incidence angle.
The model does not consider intermolecular collisions; the code is provided of a graphical routine for the visualization
of the re-emission lobes in the meridian and azimuth planes. The output pressure and shear stress are averaged on the
2) DS2V-64 bits [13] to compute, besides the 2-D global aerodynamic coefficients, also the pressure, shear stress
3) DS3V [14] to compute the global aerodynamic coefficients of SpaceLiner in clean and flapped configuration.
Both DS2V and DS3V consider air as made up of five neutral reacting species (O2, N2, O, N and NO) and rely on
the built-in Gupta-Yos-Thompson [20] chemical model, consisting of 23 reactions. Both codes are “sophisticated”. As
widely reported in literature [21-24], these codes implement computing procedures providing efficiency and accuracy
higher than those from a “basic” DSMC code. A sophisticated code, in fact, is based on two sets of cells (collision and
sampling cells) with the related cell adaptation and implements methods promoting nearest neighbour collisions. This
type of code generates automatically computational parameters such as numbers of cells and of simulated particles by
8
the input numbers of megabytes and of free stream number density. It uses a radial weighting factor in solving axial-
symmetric flow fields and provides optimal time step. Finally, the same collision pair cannot have sequential collisions.
Besides being “sophisticated”, DS2V and DS3V are also “advanced”, allowing the user to evaluate the quality of a
simulation. The user can verify, by the on line visualization of the ratio between the molecule mean collision separation
(mcs) and the mean free path (λ or m.f.p.) in each computational cell, that the number of simulated particles and
collision cells are adequate. In addition, the codes allow the user to change (or to increase), during a run, the number of
simulated particles. The ratio mcs/λ has to be less than unity everywhere in the computational domain. Bird [13], [14]
suggests 0.2 as a limit value for an optimal quality of the run. In addition, the codes give the user information about the
stabilization of the runs by means of the profile of the number of simulated particles as a function of the simulated time.
The stabilization of a DSMC calculation is achieved when this profile becomes jagged and included within a band
4. Case Studies
Figure 2(a) shows the rendering of SpaceLiner-7.1, an advanced concept for a suborbital, hypersonic, winged
passenger transport, which is currently under investigation at the German Aerospace Center (Deutsches Zentrum für
Luft- und Raumfahrt or DLR [10, 11]). The fuselage length (L) and the wing span of the vehicle are 63 m and 40 m,
respectively. By referring to a Cartesian system, whose origin is located in the nose stagnation point, the x-axis is along
the fuselage and the z-axis is in the symmetry plane, the center of gravity is located at xcg=34.5 m and zcg=1.28 m.
Figure 2(b) shows the vehicle with flaps deflection of δ=-35 deg, approximated by an unstructured body grid. The
surface mesh was generated by the Rhinoceros software, including 1877 and 1961 surface triangles for the clean and
flapped configurations, respectively. The wing profile, located at y=11.5 m from the symmetry plane, is highlighted in
Fig.2(a) and is shown in Fig.2(c). The chord (c) and the maximum thickness of the airfoil are about 23.8 m and 1.1 m,
respectively.
The considered flight conditions could likely be a point of interest of a typical mission profile of the SpaceLiner at an
altitude of 120 km. The free stream velocity (V∞) is 4500 m/s [12], the thermodynamic parameters and air composition
-8 3 15 -3 -3
are provided by the US standard Atmosphere: T∞=329 K, ρ∞=2.24×10 kg/m , N∞=5.15×10 m , p∞=2.56×10 [Pa]
λ∞=3.5 m and the molar fractions of molecular oxygen, molecular nitrogen and atomic oxygen are 0.0864, 0.7294 and
0.1842, respectively. The wall temperature (Tw), constant along the whole body, is 300 K. The free stream Mach
number is M∞=11, the free stream Reynolds and Knudsen numbers, based on the fuselage length and the airfoil chord,
9
are ReL∞=283, KnL∞=0.06 and Rec∞=107, Knc∞=0.15; the values of the Knudsen numbers indicate that the flow fields
around both the whole vehicle and the airfoil are in low density regime.
(a) (b)
(c)
Fig. 2. (a) Rendering of SpaceLiner-7.1, (b) unstructured body grid with deflected flaps used in DS3V
simulations (δ=-35 deg), (c) wing section at y=11.5 m from the symmetry plane
Tests at the impact point were carried out in the interval of incidence angle (ϑ) 5-85 deg with a spacing of 5 deg
5
Each run simulated 10 molecule-surface collisions. The tests considered, for both models, specular (f=1, αn=αt=0),
Simulations on the airfoil were carried out varying the angle of attack AoA in the range 0-50 deg with a spacing of 5
deg, considering specular, semi-specular and full diffusive Maxwell models and only the semi-specular CLL model.
The computational region for the 2-D simulations, performed by DS2V-64 bits code, was a rectangle: Lx2D=29 m,
Lz2D=8 m. Using this code, a better geometrical definition of the surface was possible. In fact, the wing section was
approximated by 250 linear panels on the upper surface and 250 linear panels on the lower surface.
The simulations on the whole vehicle were carried out in the same interval of angles of attack (0-50 deg) and
involved only the specular, semi-specular and full diffusive Maxwell models. The 3-D simulations, performed by the
DS3V code, took advantage from the symmetry of the body and of the flow field (no sideslip angle); the flow field was
computed about only half vehicle. The computational region was a parallelepiped: Lx3D=77 m, Ly3D=28 m, Lz3D=34 m.
10
5. Accuracy of the computations
DS2V and DS3V codes automatically generate the computational grid on the basis of the available megabytes and of
free stream number density. The user can control the number of cells by setting the input number of divisions and
elements in each division; the higher is the number of elements and/or the lower is the number of molecule per cell, the
higher is the number of the cells after the adaptation process. The codes suggest an optimal number of molecules for
In the present work the adaptation process for each 2-D test (DS2V) was undertaken considering 30 molecules/cell
and about 250 molecules/cell for the collision and sampling cells, respectively; the number of collision and sampling
4 6 7
cells were about 9×10 and 3×10 , respectively. The number of simulated particles was about 5×10 . The quality of the
-3
results is guaranteed by the relatively low value (order of 2.0×10 ) of the mcs/λ, and by the ratio between simulated
time (ts) and time (t∞) to travel the airfoil chord at the free stream velocity (ts/t∞ ≅5).
For the three-dimensional computations (DS3V), the adaptation process relied on 8 molecules for each collision cell
7
and on 30 molecules for each sampling cell. Each simulation, with and without flap, was carried out with about 1.6×10
simulated particles. The maximum value of the mcs/λ ratio, averaged on all runs, was about 0.6 and ts/t∞ ≅50. For the
sake of completeness, Fig.3 shows the 2-D spots, displayed during the DS3V run, of the ratio mcs/λ at the most severe
test condition for DSMC, i.e. for SpaceLiner with flap and at α=50 deg. The maximum value of mcs/λ is less than unity
and in most of the flow field close to the suggested optimal limit value of 0.2.
Fig. 3. 2-D spot of the ratio mcs/λ in the meridian plane of the flow field around SpaceLiner with flap:
α=50deg
11
Even though DS2V and DS3V are well known and widely accepted by the international DSMC community, a set of
preliminary runs have been also carried out in order to check the reliability of the grids stated automatically by the
codes. A sensitivity analysis has been carried out for both DS2V and DS3V, in terms of the global aerodynamic
coefficients computed at an intermediate angle of attack (α=25 deg). The present analysis follows a similar procedure
[25] in which the influence of the number of collision and sampling cells was evaluated on the DS2V results. The test
case considered in [25] was the Orion capsule in axi-symmetric flow during re-entry (h=85 km, V∞=7600 m/s) and the
merit parameter was the heat flux distribution. The present analysis extends the former one by including also DS3V.
Tables 1(a) and 1(b) report some aerodynamic coefficients of the airfoil and of the SpaceLiner vehicle computed as
functions of the number of collision cells for DS2V and of the input number of elements for DS3V (DS3V does not
provide in output the number of collision and sampling cells). In order to highlight the differences, data are shown up to
the forth decimal digit. Data in the first rows of the tables are computed by the DS2V and DS3V default grid structures;
both tables verify that the aerodynamic coefficients of the airfoil and of SpaceLiner practically do not change.
Table 1(a)
AoA=25 deg
Table 1(b)
AoA=25 deg
12
6. Results and discussion
For a direct comparison of the Maxwell and CLL models, preliminary tests have been carried out considering just the
impact point on an elementary surface. Figure 4 shows the re-emission lobes from the Maxwell and the CLL models at
the intermediate angle of incidence of ϑ =40 deg. The re-emission lobes from both models are drawn in the meridian
plane (x-y) and in the azimuth plane (x-z). The graphical representation of the lobes relies on Eqs.5.a,b,c for the
computation of the incident velocity components and on Eqs.6.a,b,c and 9.a,b,c for the computation of the re-emitted
velocity components from the Maxwell full diffusive and CLL models, respectively. All re-emitted velocities are scaled
Figures show that the re-emission lobes corresponding to full diffusive re-emissions are quite similar for the two
models and the lobes for specular re-emission in the meridian plane are pretty similar. This is an important result
because, being the models completely different, each model can support the other one. The only relevant difference is
that the CLL model does not compute re-emission in the azimuth plane for a specular re-emission. For completeness,
the typical shape of the lobes computed with the CLL model are shown also for semi-specular (αn=αt=0.5) re-emission.
As expected, also the re-emitted pressure (pr) and tangential stress (τr) by the two models, scaled by the free stream
dynamic pressure, are comparable (see Figs. 5(a) and 5(b)). The tangential stress are almost equivalent.
Maxwell f=1
f=0
V∞ V∞
ϑ ϑ
Cercignani-Lampis-Lord
V∞ αn = 1 αt = 1 V∞ αn = 0 αt = 0
ϑ ϑ
13
Fig. 4. Re-emission lobes of the Maxwell and the CLL models: ϑ =40 deg
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
2 2 1 1
Maxwell, f=0
CLL, α n= α t=1
1.6 Maxwell, f=0.5 1.6 0.8 0.8
CLL, α n= α t=0.5
Maxwell, f=1
1.2 CLL, α n= α t=0 1.2 0.6 0.6
τr
pr
(a) (b)
Fig. 5. Profiles of the re-emitted pressure (a) and tangential stress (b) by the Maxwell and CLL models
To quantify the difference between the models, the ratios between the results of the CLL and Maxwell models are
reported in Table 2. The ratios involving the tangential stresses for full diffusive re-emission are not reported because
they are not meaningful. In agreement with what shown in Fig.5 and excluding the low (5, 10 deg) and the high (80, 85
deg) angles of incidence, the ratios are all of the order of magnitude one, even for specular re-emission. More
specifically, in agreement with the almost coincident full diffusive re-emission lobes, the ratios of pressure are exactly
equal to one. The same happens for the ratio of the tangential stresses for specular and semi-specular reflection. This
implies that no significant differences are found using the two models. The present results confirm the equivalence of
the Maxwell and CLL models already found by Padilla and Boyd [18]. This is an important remark because the
computer tests in [18] were performed in rarefied conditions different from the present ones. In fact, Padilla and Boyd
carried out computations on flat plate at test conditions corresponding to 140 km orbit around Venus.
14
Table 2
Figures 6(a) to 6(d) show the influence of the gas-surface interaction model on the wing section aerodynamic
coefficients: lift Cl (a), drag Cd (b), longitudinal moment (the reduction pole is the leading edge) Cm (d), aerodynamic
efficiency E (Cl/Cd) (d) in clean configuration (δ=0 deg). These figures show that, also in this case, the semi-specular
Maxwell and CLL models produce very similar results and, as expected, the coefficients for both models are
intermediate between the full diffusive and specular cases. This is confirmed also by Figs. 7(a), 7(b) and 7(c), showing
15
the profiles of pressure (a), skin friction (b) and heat flux (c) along the wind side of the wing at an intermediate angle of
attack AoA=20 deg. The influence, on the computed pressure, of the interaction model is negligible and, as expected,
the skin friction and the heat flux for both models is one half of the corresponding value from the full diffusive model.
Tables 3, 4 and 5 summarize the results in terms of computed coefficients from the Maxwell specular, Maxwell
semi-specular and CLL semi-specular model divided by the same coefficients from the Maxwell full diffusive model.
According to the previous figures, the effect of the re-emission model are relevant and the ratios increase with the angle
of attack. For example, in the interval of angle of attack 0-50 deg, Clf=1/Clf=0 changes from 0.77 to 1.68, Cdf=1/Cdf=0
changes from 0.07 to 0.98 and Cmf=1/Cmf=0 changes from 0.45 to 1.10. This is explained by the fact that, as shown in
Fig.5(a), increasing the angle of incidence, the specular re-emitted pressure increases much more than the full diffusive
one and, consequently, all aerodynamic coefficients (lift, drag, moment) are larger. Furthermore, the specular, re-
emitted tangential stress is different from zero, while the full diffusive one is always zero, see Fig.5(b).
The ratios for the semi-specular Maxwell model show a similar trend but to a lesser extent. Due to the above seen
equivalence of the Maxwell and the CLL models, similar values of the ratios are also found for the semi-specular CLL
model.
16
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
1.2 1.2 1.5 1.5
Maxwell, f=1 Maxwell, f=1
Maxwell, f=0.5 Maxwell, f=0.5
1 CLL, α n= α t=0.5 1 1.25
CLL,α n=α t=0.5
1.25
Maxwell, f=0 Maxwell, f=0
0.8 0.8 1 1
Cd
Cl
0 0 0 0
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
AoA [deg] AoA [deg]
(a) (b)
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0 0 3.5 3.5
Maxwell, f=1
-0.1 -0.1 3 Maxwell, f=0.5 3
CLL, α n= α t=0.5
-0.2 -0.2 Maxwell, f=0
2.5 2.5
-0.3 -0.3
2 2
Cm
-0.4 -0.4
1.5 1.5
-0.5 -0.5
Maxwell, f=1 1 1
-0.6 -0.6
Maxwell, f=0.5
CLL, α n= α t=0.5 0.5 0.5
-0.7 -0.7
Maxwell, f=0
-0.8 -0.8 0 0
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
AoA [deg] AoA [deg]
(c) (d)
Fig. 6. Profiles of the lift (a) and drag (b), longitudinal moment coefficients (c) and aerodynamic efficiency
(d) of the wing section in clean configuration (δ=0 deg) by the Maxwell and CLL models
17
0 5 10 15 20 25 0 5 10 15 20 25
0.5 0.5 0.2 0.2
Maxwell, f=1 Maxwell, f=0.5
Maxwell, f=0.5 CLL, α n= α t=0.5
0.4 CLL, α n= α t=0.5 0.4 Maxwell, f=0
Maxwell, f=0 0.16 0.16
0.3 0.3
p [Pa]
τ [Pa]
0.12 0.12
0.2 0.2
0.08 0.08
0.1 0.1
0 0 0.04 0.04
0 5 10 15 20 25 0 5 10 15 20 25
s [m] s [m]
(a) (b)
0 5 10 15 20 25
700 700
Maxwell, f=0.5
600 CLL, α n= α t=0.5 600
Maxwell, f=0
500 500
q [W/m2]
400 400
300 300
200 200
100 100
0 0
0 5 10 15 20 25
s [m]
(c)
Fig. 7. Profiles of pressure (a), shear stress (b) and heat flux (c) along the wind side of the airfoil by the
18
Table 3
19
Table 4
20
Table 5
Ratios of the airfoil aerodynamic coefficients between the CLL semi-specular (αn=αt=0.5)
21
6.3. SpaceLiner
Figures 8(a)–8(d) show the profiles of the lift CL (a) and of the drag CD (b) coefficients, of the aerodynamic
efficiency E (c), of the longitudinal moment coefficient CMcg (d), the reduction pole is the vehicle center of gravity, as
functions of the angle of attack. Figure 8(e) shows the stability index I, defined as I=(xcg-xp)/L, where xp is the x-
coordinate of the pressure center. All results refer to the vehicle without flap deflection and with flap deflection δ=-35
The increments of the lift and drag coefficients, produced by the flap deflection, are stronger for the specular model,
as discussed in the case of the wing section. The most relevant differences, produced by the gas-surface interaction
model, are evident in the aerodynamic efficiency and in the longitudinal moment coefficient profile, therefore in the
longitudinal stability. It is well know that the longitudinal stability is defined by the conditions dCMcg/dα<0 and I<0.
SpaceLiner with no flap and with a specular interaction is unstable (dCMcg/dα>0, I>0) up to α=40 deg. On the other
hand the vehicle is longitudinally stable in the whole interval of angle of attack for full diffusive and semi-specular
interaction models. Considering the results with flaps, even in the condition of longitudinal stability, the equilibrium is
The interaction model strongly influences the angle of maximum efficiency. This angle, for specular, semi-specular
and diffusive models is 5, 20 and 25 deg with no flap deflection and 10, 15 and 20 deg with deflected flap, respectively.
As discussed before, the value of the maximum efficiency angle of attack is important because at this angle a longer
downrange and a larger maneuverability is possible along the atmospheric re-entry flight path.
Once again, in order to quantify the influence of the surface interaction model, Tables 6(a),(b) and 7(a),(b) report the
ratios of the aerodynamic coefficients and stability index as functions of the angle of attack, computed by the specular
and semi-specular models with respect to the full diffusive model. Tables 6(a) and 7(a) show the results corresponding
to the configuration with no flap deflection. Tables 6(b) and 7(b) refer to the case of deflected flap. These results show
that the interaction model has a large effect. For example, the ratio CMcgf=1/CMcgf=0 ranges from -0.80 to 0.38 when
flap deflection is zero and from 3.29 to 0.94 when the flap is deflected. The ratio If=1/If=0 is between -4.54 and 11.10
when flap deflection is zero and between -0.16 and 3.30 with deflected flap.
22
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
2.4 2.4 3.5 3.5
No flap, f=1 No flap, f=1
Flap, f=1 3 Flap, f=1 3
2 No flap, f=0.5 2 No flap, f=0.5
Flap, f=0.5 Flap, f=0.5
2.5 2.5
1.6 No flap, f=0 1.6 No flap, f=0
Flap, f=0 Flap, f=0
2 2
CD
CL
1.2 1.2
1.5 1.5
0.8 0.8
1 1
0 0 0 0
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
AoA [deg] AoA [deg]
(a) (b)
0 5 10 15 20 25 30 35 40 45 50
5 5
No flap, f=1
Flap, f=1
4 No flap, f=0.5 4
Flap, f=0.5
No flap, f=0
3 Flap, f=0 3
E
2 2
1 1
0 0
0 5 10 15 20 25 30 35 40 45 50
AoA [deg]
(c)
23
0 5 10 15 20 25 30 35 40 45 50 5 10 15 20 25 30 35 40 45 50
0.05 0.05 0.12 0.12
0.08 0.08
0 0 0.04 0.04
0 0
-0.05 -0.05
-0.04 -0.04
CMcg
Ι
-0.08 -0.08
-0.1 No flap, f=1 -0.1 No flap, f=1
Flap, f=1 -0.12 Flap, f=1 -0.12
No flap, f=0.5 No flap, f=0.5
-0.15 -0.15 -0.16 Flap, f=0.5 -0.16
Flap, f=0.5
No flap, f=0 -0.2 No flap, f=0 -0.2
Flap, f=0 Flap, f=0
-0.2 -0.2 -0.24 -0.24
0 5 10 15 20 25 30 35 40 45 50 5 10 15 20 25 30 35 40 45 50
AoA [deg] AoA [deg]
(d) (e)
Fig. 8. Profiles of lift (a) and drag (b) coefficients, aerodynamic efficiency (c), longitudinal moment
coefficient (d) and stability index (e) of the SpaceLiner with flap and with no flap as functions of the angle of
attack and the Maxwell models
24
Table 6(a)
Ratios of the SpaceLiner aerodynamic coefficients and stability index by the Maxwell
specular (f=1) and the Maxwell full diffusive (f=0) models: δ=0 deg
25
Table 6(b)
Maxwell specular (f=1) and the Maxwell full diffusive (f=0) models: δ=-35 deg
26
Table 7(a)
Maxwell semi-specular (f=0.5) and the Maxwell full diffusive (f=0) models: δ=0 deg
27
Table 7(b)
Ratios of the aerodynamic parameters and stability index of SpaceLiner by the Maxwell
semi-specular (f=0.5) and the Maxwell full diffusive (f=0) models: δ=-35 deg
28
7. Conclusions and future developments
The gas-surface interaction models by Maxwell and Cercignani-Lampis-Lord have been compared in hypersonic,
rarefied flow, considering the aerodynamic characteristics of the SpaceLiner vehicle. Preliminary comparisons have
been carried out analyzing the normal and tangential stresses linked to the re-emitted molecules at the impact point on
an elementary surface. The computer simulations have been carried out considering specular, semi-specular and full
diffusive models. The results show that: i) the outputs of the Maxwell and Cercignani-Lampis-Lord models are very
similar, when the same molecules re-emission behavior is considered, in agreement with what already found in
literature, ii) the influence of the re-emission model (changing from specular to full diffusive) is relevant. The effects
increase with the angle of attack and are more important when a complete flapped vehicle configuration is considered.
More specifically, considering different interaction models, the angle of maximum efficiency is very different and the
The present results suggest considering in details the influence of the accommodation coefficients, supposed constant
in this preliminary work. Future developments are necessary to investigate the dependence of the accommodation
References
1. G. A. Bird, Molecular Gas Dynamics and Direct Simulation Monte Carlo, Clarendon Press, Oxford, Great Britain (1998).
2. G. A. Bird, The DSMC Method, Version 1.1, Amazon, ISBS 9781492112907, Charleston, USA (2013)
3. C. Shen, Rarefied Gas Dynamic: Fundamentals, Simulations and Micro Flows, Springer-Verlag, Berlin, Germany (2005)
4. R. Schamberg, A New Analytical Representation of Surface Interaction for Hyper-thermal Free Molecule Flow, Rand Corp., RM-
5. R. Schamberg, Analytic Representation of Surface Interaction for Free Molecular Flow with Application to Drag of Various
Bodies, Aerodynamics of the Upper Atmosphere, Rand Corp., R-339, Santa Monica, CA, 1959, pp 12-1 – 12-41
6. S. Nocilla, The Surface Re-Emission Law in Free Molecule Flow, Proceedings of the 3rd International Symposium on Rarefied
7. Lord R.G., Some Extensions to the Cercignani-Lampis Gas- Surface. Scattering Kernel, Phys. Fluids A 3 (4), pp 706-710 (1991)
8. E.D. Knechtel, W.C. Pitts, Normal and tangential momentum accommodation for Earth satellite conditions, Acta Astronautica,
9. G. Zuppardi, DSMC Implementation of the Schamberg Model: Comparison with the Maxwell Model, Proceedings of the 24th
29
10. M. Sippel, A. van Foreest, C. Bauer, F. Cremaschi, System Investigation of the SpaceLiner Concept in FAST20XX, AIAA Paper
2011-2294 (2011)
11. M. Sippel et al., Technical Maturation of the SpaceLiner Concept, 18th AIAA/3AF International Space Planes and Hypersonic
Systems and Technologies Conference, 14-18 September 2012, Tours, France, http://arc.aiaa.org/doi/abs/10.2514/6.2012-5850
12. G. Zuppardi, L. Morsa, M. Sippel, T. Schwanekamp, Aero-Thermo-Dynamic Analysis of the SpaceLiner-7.1 Vehicle in High
Altitude Flight, Proceedings of the 29th International Symposium on Rarefied Gas Dynamics, Xi’an, China (2014).
13. G.A. Bird, The DS2V Program User’s Guide Ver. 3.2, G.A.B. Consulting Pty Ltd, Sydney, Australia (2005).
14. G.A. Bird, The DS3V Program User’s Guide Ver. 2.5, G.A.B. Consulting Pty Ltd, Sydney, Australia (2006).
15. D.M. Pilinski, Dynamic Gas-Surface Interaction Modeling for Satellite Aerodynamic Computations, Diss. Ph. D. Thesis, Univ.
16. M.M. Moe, S.D. Wallace, K. Moe, Refinements in the Determination of Satellite Drag Coefficients: Method for Resolving
17. K. Moe, M.M. Moe, Gas-Surface Interactions in Low-Earth Orbit, Planetary and Space Science 53, 793-801 (2005)
18. J.F. Padilla, I.D. Boyd, Assessment of Gas-Surface Interaction Models for Computation of Rarefied Hypersonic Flow, J.
19. M.S. Woronowicz, D.F.G. Rault, Cercignani-Lampis-Lord Gas-Surface Interaction Model: Comparison Between Theory and
20. R.N. Gupta, J.M. Yos, R.A. Thompson, A Review of Reaction Rates and Thermodynamic Transport Properties for an 11-Species
Air Model for Chemical and Thermal Non-Equilibrium Calculations to 30,000 K, NASA TM 101528 (1989).
21. G.A. Bird, Sophisticated Versus Simple DSMC, Proceedings of the 25th International Symposium on Rarefied Gas Dynamics,
22. G.A. Bird, M.A. Gallis, J.R. Torczynski, D.J. Rader, Phys Fluid. 21, 017103 (2009).
23. M.A. Gallis, J.R. Torczynski, D.J. Rader, G.A. Bird, J. Comput. Phys. 228, 4532-4548 (2009).
24. G.A. Bird, The DS2V/3V Program Suite for DSMC Calculations, Proceedings of the 24th International Symposium on Rarafied
25. L. Morsa, G. Zuppardi, R. Votta, A. Schettino, Influence of Chemical Models on Heat Flux for EXPERT and Orion Capsules,
30