COMPARING POTENTIAL FLOW SOLVERS FOR AERODYNAMIC
CHARACTERISTICS ESTIMATION OF THE T-FLEX UAV
Fanglin Yu1 , Julius Bartasevicius1 & Mirko Hornung1
1 Institute of Aircraft Design, Technical University of Munich, Boltzmannstrasse 15, 85748 Garching
Phone: +49.89.289.15988, Fax: +49.89.289.15982
Abstract
The FliPASED project aims to design an actively-controlled wing with an aero-servo-structural optimization
toolchain to achieve drag reduction, and test the new design on the T-FLEX demonstrator, which is a 7m
wingspan unmanned aircraft with a V-Tail. Accurate and fast predictions of aerodynamic characteristics is vital
in this preliminary design optimization process. This paper discusses the applicability of several low order
aerodynamic tools to such aircraft while comparing them with a higher order CFD tool. The lifting-line based
XFLR5 and PAWAT, the vortex-lattice based AVL, Tornado, PyTornado and VSPAERO, the 3D panel based
FlightStream, and the commercial CFD tool STAR-CCM+ were used. The results include a convergence study,
computational performance of each tool and comparison of calculated force and moment coefficients. The
user experience of these tools are shared.
Keywords: Aerodynamics, UAV, T-FLEX, FLiPASED, FLEXOP
1. Introduction
As the climate impact of air traffic-related emissions gets more concerned, increasing the efficiency
of aircraft and reducing the emission is of high interest. The EU-funded project Flight Phase Adaptive
Aero-Servo-Elastic Aircraft Design Methods (FliPASED) aims to improve aerodynamic performance
and cut down fuel consumption of aircraft by a concept for active wing shape adaptation [1]. Such
concept can reduce the drag by maintaining optimal lift distribution during various flight conditions
through actively deflecting control surfaces. Based on the concept, an active-controlled wing design
can be achieved by a multidisciplinary design optimization (MDO) toolchain which incorporates aero-
dynamics, structural design, aeroelastic simulation and control design. The wing design will be tested
on the T-FLEX demonstrator, which is a UAV developed in the previous EU project Flutter Free FLight
Envelope eXpansion for ecOnomic Performance Improvement (FLEXOP) [2].
The aerodynamic solver is vital in the wing design process for accurate and fast estimation of the
aerodynamic characteristics of the wing, especially drag. Computational fluid dynamics (CFD) tools
can yield results with high fidelity, but are too computational expensive to be applied in such case,
even with today’s computation power. Therefore only low-order aerodynamic tools can make the
design optimization computational feasible. There are plenty of low-order aerodynamics tools based
on different theories, e.g. Lifting-line theory (LLT), Vortex lattice method (VLM) and panel method. An
overview of some available tools and their theory can be found in the review from Technical University
of Delft [3].
To find out the most suitable tool to be used in the active-controlled wing design, this study is con-
ducted. The low order aerodynamic tools, Athena Vortex Lattice (AVL) [4], XFLR5 [5], PyTornado
[6], Tornado, VSPAERO, PAWAT, FlightStream are compared regarding aspects such as prediction
accuracy of aerodynamic characteristic, computational performance and applicability to the T-FLEX
UAV . In the comparison, the data from the CFD tool STAR-CCM+ serve as the reference.
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
1.1 Description of the Demonstrator
The T-FLEX demonstrator is a 65 kg take-off weight, 7m wingspan unmanned aircraft with a swept
wing and a V-Tail and is powered by a jet engine (Figure 1). A custom airfoil (internal name try6) in
two versions- 10 percent thickness for the root and 8 percent for the wingtip- has been developed for
the wing. A symmetrical 7.5 percent thickness airfoil is used for the tail surfaces.
Figure 1 – T-FLEX Subscale flight demonstrator during landing phase.
The geometry of the aircraft is summarized in table 1.
Table 1 – Geometry of T-FLEX UAV
Wing span, m: 7.0 Tail projected span, m: 1.27
Wing area, m2 : 2.478 Tail area, m2 : 0.39
Wing aspect ratio: 19.77 Tail aspect ratio: 4.2
Wing incidence, deg: -0.5 Tail incidence, deg: -4.33
Wing 0.25c sweep, deg: 19.14 Tail 0.25c sweep, deg: 19.83
Wing taper ratio: 0.5 Tail taper ratio: 0.52
Wing twist, deg: -2 Tail dihedral, deg: 35
Number of wing control surfaces: 8 Number of tail control surfaces: 4
Fuselage length, m: 3.42
Fuselage maximum height, m: 0.315
Fuselage maximum width m: 0.3
2. Aerodynamic Modelling
The increase in computational power has enabled complex numerical methods to be used for solving
aerodynamic problems. Starting with Prandtl’s Lifting Line approach up to the three-dimensional
Direct Numerical Simulations, all methods can be solved on high-performance personal computers
these days. However, out of the full range of methods some are more popular than others due to a
good balance between their precision, complexity, reliability and computational cost. This group of
methods are the new implementations of the Lifting Line Theory, Vortex or Doublet Lattice Method
and 3D Panel Method. The mathematics of these methods will not be explained here, as this is not
2
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
the goal of this study. However, descriptions of these methods are provided with references for an
interested reader.
2.1 AVL
AVL is a program for performing aerodynamic analysis of rigid aircraft of arbitrary configurations [4].
It uses the VLM method to model the lifting surfaces. One advantage of AVL is the implementation of
the slender body theory for fuselage modelling. Because of an intrinsic limitation of VLM, AVL is only
suitable for inviscid calculation at small angles of attack and sideslip. Beside the global aerodynamic
coefficient, flight stability characteristics can be acquired from eigenmode analysis in AVL.
2.2 Tornado
Tornado is a Vortex Lattice Method for linear aerodynamic wing design applications in conceptual
aircraft design or in aeronautical education [7]. The method is built in MATLAB [8] and is based on
the description as provided by Moran [9].
The VLM implementation in Tornado ignores the thickness effects of the airfoil, but includes the cam-
ber. In Tornado, modelling of control surfaces is possible. Experimental functions that generate a
Trefftz-plane analysis can be used. Different options for mesh creation (linear panel distribution, co-
sine panel distribution, etc.) are available. A graphical interface is available which can plot coefficients
of interest, display geometries and mesh.
Initially, Tornado was designed only to include linear aerodynamics. However, the code has been
updated to include viscous effects as well [10].
If required, stability derivatives can be calculated using central-difference approximation around the
trim condition. It is also possible to calculate trimmed polars.
2.3 PyTornado
PyTornado is an aerodynamic tool for conceptual aircraft design. Short computation times make
it possible to easily obtain estimates of aerodynamic loads and to benchmark different concepts
[6]. Although a similar name as Tornado, PyTornado has been implemented from scratch within the
European research project AGILE. A Vortex Lattice Method is implemented in this code. It has a
user interface, pre- and post-processing in Python and a calculation core routine in C++ [11], which
guarantees a user friendly interface and computational efficiency. It can be used as a standalone
aerodynamic solver or can be integrated into a MDO toolchain. The deformation feature, which is
under development, could be potentially used for aeroelastic analysis.
2.4 XFLR5
XFLR5 is a software tool designed specifically with model sailplanes in mind [5][12]. Therefore, it
focuses on wings operating at low Reynolds numbers. The tool uses XFoil [13] (XFoil v6.99 since
XFLR5 v6.55) to calculate the 2D aerodynamics of an airfoil. Non-linear Lifting Line Theory (based on
the NACA technical note 1269 [14]), Vortex Lattice Method with quadrilateral rings (as recommended
by Katz and Plotkin [15]) or 3D Panel Method (based on Maskew [16]) can be used for 3D wing and
tail analysis. Body analysis is not recommended by the author [12].
Unlike the usual VLM solvers, the VLM method implemented in XFLR5 provides a viscous drag
correction. In such case, lift-related characteristics (lift distribution, induced drag) are kept inviscid
and after local lift distribution is calculated, viscous drag correction using 2D airfoil polars is applied.
The lift distribution is not changed. This method is also used during this study. However, the author
of the software raises awareness that such correction is not scientifically sound, as using 2D polars
ignores any spanwise effects [12].
2.5 VSPAERO
VSPAERO [17] is the aerodynamic analysis tool integrated within the conceptual aircraft design pack-
age OpenVSP [18]. The tool has two methods available - the Vortex Lattice Method with a simple
stall prediction methodology (not used in this study) and a 3D Panel method [19]. Propellers can be
included in the simulation. The tool also incorporates the possibility to calculate the parasite drag
using the component build-up method. In the current study, only the VLM method is used.
3
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
2.6 PAWAT
The Preliminary Design Tool for Propeller-Wing Aerodynamics (PAWAT) is an aerodynamic tool for
the conceptual design of aircraft [20]. The calculation of the steady state lifting surface aerodynamics
in PAWAT is based on a modified three-dimensional nonlinear lifting line theory with a fixed wake
model employing nonlinear airfoil data to model nonlinear and viscous effects to a certain extent [20].
PAWAT is also capable of modelling propellers and it allows investigations of the interaction effects
between wing and propeller.
The method is built in MATLAB [8]. The description of the lifting line method used is described by
Phillips and Snyder [21].
2.7 FlightStream
FlightStream is a novel surface vorticity solver capable of using structured or unstructured surface
meshes. As a vorticity-based solver, the code can be expected to be substantially more robust and
stable compared to pressure-based potential-flow solvers and less sensitive to surface perturbations,
and it also allows the use of coarser meshes with an acceptable level of fidelity [22].
To account for viscous effect, integral boundary layer was implemented in FlightStream and was
coupled with inviscid solver via displacement of the inviscid boundary equal to the displacement
thicknesses of the local boundary layers. More features like prediction of flow separation and stall
characteristics are also enabled by this implementation.
2.8 STAR-CCM+
Simcenter STAR-CCM+ is a multiphysics computational fluid dynamics (CFD) software. In this study,
it is used to provide the reference data for comparison.
For these simulations, the geometry was prepared within the CAD software SolidWorks. The T-FLEX
geometry was imported and the bullet-shaped domain was created (15 spans in radius and 45 spans
in length, Figure 2). The trailing edges of the lifting surfaces were trimmed to total thickness of 0.2
percent of the reference chord length of the lifting surface. The geometry was then transferred to
STAR-CCM+. In-order to reduce the computational effort, symmetry condition was used.
A polyhedral mesher was used. For turbulent simulations, a prism layer mesher was applied. Custom
mesh controls were set on the different boundaries (domain, fuselage, wing and tail) as well as on
the trailing and leading edges. Wake controls (2 spans in length and 15◦ in spread angle) were used
on the fuselage, wing and tail (Figure 3).
Steady inviscid and steady fully turbulent simulations were done. In the latter, Reynolds-averaged
Navier-Stokes (RANS) equations were solved. Spalart-Allmaras turbulence model was used together
with all y+ treatment. Near wall thickness of the prism layer was therefore adjusted for y+ = 1 condi-
tion. Total thickness of the prism layer was adapted to fit the complete boundary layer at AoA = 2◦
(Figure 4).
Figure 2 – The domain used in STAR-CCM+ Figure 3 – The wake control applied on the
simulations. geometry.
One has to emphasize that most of the tools simulated the wing and tail. Fuselage was included only
in simulations of STAR-CCM+ and FlightStream. A study was done with STAR-CCM+ to investigate
4
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
Figure 4 – Prism layer mesh around the wing cross section.
the influence of the fuselage on the spanwise lift distribution. A small influence was noted at low
angles of attack. However, at high angles of attack the fuselage does change the flow at the wing
root.
3. Practical Recommendations
3.1 Geometry modelling
Special care must be taken when defining the aircraft geometry within different aerodynamic or struc-
tural modelling tools. As such tools are usually used during the preliminary design stage, where no
CAD models are available, the aircraft geometry must be defined via geometry parameters. From the
tools tested during this study, only FlightStream supports direct import of CAD models.
In the current case, a collection of parameters describing the overall geometry were available from
early stages of the project. These were initially used for all the aerodynamic modelling software.
However, it became clear soon, that the parameters not only differed from the actual build geometry
(due to modelling and manufacturing errors), but also the different software tools interpreted these
parameters differently.
As a result the geometry modelled in each of the tools was exported in a .stl format. Then the actual
build geometry was 3D scanned and all the geometries were compared within a CAD software. The
initial missalignment of the geometries within different tools are shown in the Figure 5.
(a) Isotropic view. (b) Top view.
Figure 5 – The visual comparison of the geometries modelled in different tools. Bright red - 3D
scanned geometry; green - the CAD geometry after aligning it with the scan; yellow - geometry as
initially defined in the project; blue - XFLR5; dark red - VSPAERO.
After this was done, it was ensured that the geometries modelled in the different software tools were
identical in the end.
This problem can be particularly troublesome for the tools, which do not document in detail the
interpretation of the geometry parameters. Also, the use of V-Tails introduced additional geometrical
complexity. In such cases, a clear understanding of the definitions like the incidence of the tail
5
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
(incidence around the aircraft y-axis or around the quarter-chord line of the V-Tail surface?) or span
(projected on the horizontal plane or along the surface?) is a must.
3.2 User interface
The user interface is an important criterion for evaluating software tools. The intuitive and user-
friendly interface can lower the barrier to entry for new users. This section briefly summarises the
user experience of the different tools in this regard.
Out of the 7 tools that are presented here, only AVL, Tornado and PyTornado do not have a graphical
user interface. For these tools, menu options are chosen from the terminal and command line inputs
must be done. Only in the results processing step (plotting) graphical interaction is allowed.
The remaining four softwares all have a complete graphical user interface (Figure 6). All of them are
self-explanatory and comparably easy to use.
The results, generated by the solvers, can be accessed interactively. It should be noted that while
all the generated data (each simulation data point) is accessible in XFLR5 via the graphs menu,
data access with the other three solvers can be somewhat more complicated, if specific graphs are
required. For example, in XFLR5 all previous simulations can be accessed for the same project all
the time, but this is not possible in the viewer of VSPAERO, where only the results from a single
simulation can be viewed at once. Similarly, PAWAT also only displays results of a single data point.
(a) XFLR5. (b) OpenVSP.
(c) PAWAT. (d) FlightStream.
Figure 6 – Graphical user interfaces of the compared tools.
The results from the different tools can be exported in various format - tool specific and generic. But
all the investigated tools support export in text format, which enables users to conduct postprocessing
in any preferred external processing routine. The content of exported data varies among tools but all
the tools can export basic aerodynamic information easily.
3.3 Automation integration possibilities
As an intrinsic requirement from the MDO toolchain, the aerodynamic solver has to be executed
automatically in batch mode without GUI.
6
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
AVL and PyTornado only provide a terminal interface which is not intuitive to use compared to other
tools, but it makes it easily scripted and thus integrated into a toolchain. As Matlab can be ran in a no
GUI mode, Matlab based TORNADO and PAWAT could also be good options for a MDO toolchain.
VSPAERO can also be run from the command line. XFLR5 supports scripting in the GUI mode to run
a batch analysis.
4. Results
In this section, results from all 7 tools are compared against the inviscid and viscous CFD data. Lift,
drag, moment and lift distribution results are discussed separately.
4.1 Mesh convergence and computational performance
To make sure that the computational results are independent of the mesh, convergence studies have
been performed. The number of panels have been increased in chordwise and spanwise direction
separately until the difference in resulting lift coefficient was less than 1 percent of the absolute value.
The comparison for all the VLM-based tools can be found in Figure 7. Solver times for each mesh
study was also noted, the comparison can be seen in Figure 8.
Figure 7 – Mesh convergence study for the Figure 8 – Computational time comparison for
VLM-based tools. the VLM-based tools.
While XFLR5 and VSPAERO require the highest number panels for convergence, it was Tornado that
took the longest to compute. On the contrary, PyTornado required the lowest number of panels for
convergence while also being the most time efficient tool.
PAWAT, being a lifting line method tool, requires two orders of magnitude fewer panels to reach
convergence criteria and therefore is not displayed in the plots. It takes up to 3 seconds to perform
calculation with any mesh.
For FlightStream, the mesh is generated following the official guideline. The mesh used in this study
has 29100 cells and 78991 vertices, which is much more than the panels used in the VLM-based
method. The calculation time of the simulation with viscous coupling is around 1 minute.
4.2 Mesh convergence for STAR-CCM+
STAR-CCM+ being a CFD tool requires significantly more resources for computations. Therefore,
these simulations were performed on the Leibniz-Rechenzentrum Linux Cluster CoolMUC-2 [23].
Two nodes with 28 cores each were used.
For the mesh independence studies, the number of cells were gradually increased from 0.9 to 22.0
million and global lift, drag and moment coefficients at AoA = 2◦ were compared (Figure 9 and Table
2). In the end, a grid with 12.3 million cells was chosen for the study.
The prism layers were removed for inviscid simulations while keeping the rest of the mesh parameters
the same. This resulted in a mesh size of 5.9 million cells.
7
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
Figure 9 – Investigation of the mesh independence performed with the STAR-CCM+ software.
Table 2 – Comparison of lift, drag and moment coefficients during the mesh independence study.
Ncells CD ∆CD , % CL ∆CL , % Cm ∆Cm , % tsolver , s
0.9M 0.0247 0.1942 0.1400 139
1.6M 0.0225 -9.8 0.2004 3.1 0.1388 -0.9 237
4.6M 0.0207 -8.6 0.2037 1.6 0.1346 -3.1 705
6.4M 0.0204 -1.6 0.2040 0.1 0.1332 -1.0 958
8.5M 0.0201 -1.3 0.2045 0.3 0.1330 -0.2 1403
12.3M 0.0199 -1.3 0.2053 0.4 0.1324 -0.4 2008
22.0M 0.0197 -1.0 0.2049 -0.2 0.1319 -0.4 4284
4.3 Global aerodynamic coefficients
4.3.1 Lift
The lift coefficient data is plotted with respect to the angle of attack in Figure 10 as well as in Figure
11 for the linear part of the slope. The lift curve slope coefficients CLα and zero angle lift coefficients
CL0 are shown in Table 3.
Figure 10 – Lift coefficient CL with respect to the Figure 11 – Lift coefficient CL with respect to the
angle of attack α. angle of attack α.
Significant reduction in lift is apparent when comparing the turbulent simulations to Euler simulations.
This is expected, as the viscous boundary layer on the top surface of the wing reduces the effective
camber line, therefore reducing the aerodynamic angle of attack. Interestingly, most of the tools show
better alignment with the turbulent simulations than with the inviscid ones, even though only PAWAT
and FlightStream take viscosity into account when calculating lift.
8
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
Table 3 – Comparison of lift curve slope CLα , zero angle lift coefficient CL0 , minimum drag coefficient
CDmin , pitching moment curve slope Cmα and zero angle pitching moment coefficient Cm0 for different
aerodynamic modelling tools.
FlightStream
PyTornado
VSPAERO
Turbulent
Tornado
PAWAT
XFLR5
Euler
AVL
CLα 0.106 0.111 0.103 0.107 0.105 0.104 0.104 0.104 0.104
CL0 0.206 0.248 0.214 0.180 0.122 0.185 0.198 0.205 0.215
CDmin 0.020 0.005 0.015 0.016 0.001 0.002 0.015 0.012 0.002
Cmα -0.027 -0.028 -0.030 -0.047 -0.050 -0.032 -0.032 -0.026 -0.030
Cm0 0.132 0.141 0.117 0.103 0.193 0.214 0.147 0.128 0.159
When only the linear part of the lift curve is concerned, the calculated curve slope agreed with each
other. The zero angle of attack lift shows deviations among tools. Tornado differs most from the other
tools. Taking into account that all the tools are meant for preliminary design phase, the differences
between them could be categorised as being insignificant.
The nonlinear part of the curve is predicted by both PAWAT and FlightStream. Even at high angle of
attack, the lift curve from FlightStream matches quite well with CFD turbulent result. However, as no
CFD simulations above 14◦ were done, the CLmax could not be estimated.
The spanwise normalized lift distribution for α = 2◦ is plotted in Figure 12. As only the shape of
the distribution is of importance here, the local lift coefficients are normalized with respect to the
maximum local lift coefficient for the same tool.
Figure 12 – Spanwise normalized lift distribution for α = 2◦ . The local lift coefficients are normalized
with respect to the maximum local lift coefficient of the same tool.
9
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
The normalized lift distributions between the turbulent and Euler simulations are almost identical. The
estimated maximum local lift location is similar for all the tools. The overall shape is very similar with
some discrepancies at the root and tip areas. The differences between the STAR-CCM+ results and
the other tool results at the wingtip might be due to the poor discretization when extracting the lift
distribution from STAR-CCM+.
4.3.2 Drag
Figure 13 shows the inviscid drag polar. While all the VLM tools and the panel-based method Flight-
Stream agree mainly, the differences compared with the STAR-CCM+ Euler simulation are noticeable
even at low lift coefficient.
One has to note that the inviscid drag extracted from STAR-CCM+ here is the pressure drag compo-
nent acting on the aircraft. Strictly speaking, this is not equal to the induced drag by definition. The
separation of induced and profile drag from CFD is not straight-forward, and if Euler simulations are
used the induced drag due to viscous effects are then ignored. Nowadays there exist some methods
to extract these two drag components from CFD [24], but they were not implemented at the time of
writing this article.
The total drag coefficient shown in Figure 14 includes both viscous and inviscid drag. Significant
differences can be seen in between the tools that correct for viscous drag (STAR-CCM+ (turbulent),
FlightStream, PAWAT, XFLR5, VSPAERO) and the ones that do not (Tornado, AVL, PyTornado).
Figure 13 – Inviscid drag coefficient CDi with Figure 14 – Total drag coefficient CD with respect
respect to the angle of attack α. to the angle of attack α.
Different methods were used to correct the viscous drag in different software tools. Variation of
viscous drag is clearly visible in the Figure 14.
Both PAWAT and XFLR5 correct viscous drag based on 2D airfoil polar data. For XFLR5, 2d viscous
drag is interpolated from local wing lift coefficient. The interactive boundary layer, which is a coupling
method between potential flow and viscous flow on surfaces, is not implemented in the VLM available
in XFLR5 [5]. The consequence of underestimation of viscous drag is confirmed in the Figure 14.
In PAWAT, equations are established for wing segments based on the aerodynamic force derived
from three-dimensional vortex lifting law and the aerodynamic force derived from nonlinear airfoil
characteristics of the segment and the segment area [20]. An iterative procedure is needed to solve
the equations. Total drag coefficient from PAWAT matches quite well with CFD data.
In FlightStream, the integral boundary layer is coupled with the inviscid surface solver to account for
viscous drag. Even though, the total drag seems to be underestimated.
4.3.3 Pitching moment
The pitching moment coefficient with respect to angle of attack is shown in Figure 15.
As already shown in Table 3, there is a good (differences up to 10 percent) agreement of the pitch-
ing moment slopes of the linear part in between the STAR-CCM+, FlightStream, VSPAERO and
PyTornado. The zero angle pitching moment coefficient is predicted well by both FlightStream and
10
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
Figure 15 – Pitching coefficient Cm with respect to the angle of attack α .
VSPAERO, with the latter having even smaller deviation from the STAR-CCM+ results. This is unex-
pected, since the fuselage, which should have an influence on the pitching moment, is not modelled
in the OpenVSP.
A pitch-up trend can be noted at high angles of attack from the STAR-CCM+ simulations. This is due
to the early stall at the tip section of the wing (Figure 16). Only FlightStream captures the effect of
the pitch-up, even though not as pronounced as with the STAR-CCM+.
Figure 16 – Pressure coefficient distribution at α = 14◦ . The line marks the location where the wall
shear stress changes direction.
It should be noted that while PAWAT provided good calculations for lift and drag, the results for
pitching moment are not satisfactory.
5. Conclusions
The selection of aerodynamic tools is highly dependent on the purpose and the aircraft configuration.
Based on the user experience and simulation results of T-FLEX UAV, following recommendations are
given.
11
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
• If a fast polar calculation is of main interest, VSPAERO(OpenVSP), XFLR5 and PAWAT could
be good options. All of them have an intuitive user interface for modelling, calculation and
postprocessing. The simulations run relatively quickly.
• If the boundary layer and nonlinear effects are of concern, FlightStream provides the best re-
sults due to the coupled integral boundary layer solver.
• Pitching moment coefficient curves, as calculated by VSPAERO (linear part) and FlightStream
(nonlinear part as well), match very well with the STAR-CCM+ results.
• If an aerodynamic tool is needed for a MDO task, AVL, PyTornado and VSPAERO(OpenVSP)
could be good choices. The input file for these tools can be easily automatically prepared. The
execution of these tools can be invoked automatically and fast.
The aforementioned recommendations should be applied with caution to similar configurations as
T-FLEX which is an UAV equipped with a swept, high aspect ratio wing and a V-tail.
6. Acknowledgements
The work presented has been conducted within the framework of project FliPASED (grant agreement
No. 815058) funded from the European Union’s Horizon 2020 research and innovation program.
7. Contact Author Email Address
To contact the author, please use the following email address: fanglin.yu@tum.de
8. Copyright Statement
The authors confirm that they, and/or their company or organization, hold copyright on all of the original material
included in this paper. The authors also confirm that they have obtained permission, from the copyright holder
of any third party material included in this paper, to publish it as part of their paper. The authors confirm that
they give permission, or have obtained permission from the copyright holder of this paper, for the publication
and distribution of this paper as part of the ICAS proceedings or as individual off-prints from the proceedings.
References
[1] FLIPASED Consortium. FLIGHT PHASE ADAPTIVE AERO-SERVO-ELASTIC AIRCRAFT DE-
SIGN METHODS | FLiPASED Project | H2020 | CORDIS | European Commission. URL: https:
//cordis.europa.eu/project/id/815058 (visited on 05/13/2021).
[2] FLEXOP Consortium. FLEXOP Project Homepage. URL: https://flexop.eu/news (visited
on 10/24/2018).
[3] B Peerlings. A review of aerodynamic flow models, solution methods and solvers-and their
applicability to aircraft conceptual design Literature study report. 2018.
[4] Mark Drela. AVL. URL: http : / / web . mit . edu / drela / Public / web / avl/ (visited on
05/05/2021).
[5] Andre Deperrois. XFLR5. URL: http://www.xflr5.com/ (visited on 05/05/2021).
[6] Alessandro Gastaldi and Aaron Dettmann. PyTornado. URL: https://pytornado.readthedocs.
io/en/latest/index.html (visited on 05/05/2021).
[7] Tomas Melin. “A Vortex Lattice MATLAB Implementation for Linear Aerodynamic Wing Applica-
tions”. Royal Institute of Technology (KTH), Dec. 2000. URL: https://www.researchgate.
net/publication/238671899.
[8] MATLAB - MathWorks - MATLAB & Simulink. URL: https://uk.mathworks.com/products/
matlab.html (visited on 05/16/2021).
[9] J. Moran. An Introduction to Theoretical and Computational Aerodynamics. Dover Books on
Aeronautical Engineering. Dover Publications, 2003. ISBN: 9780486428796. URL: https://
books.google.de/books?id=4eVP3yWZ1LgC.
12
Comparing potential flow solvers for aerodynamic characteristics estimation of the T-FLEX UAV
[10] A. Da Ronch et al. “Benchmarking ceasiom software to predict flight control and flying qualities
of the B-747”. In: 27th Congress of the International Council of the Aeronautical Sciences 2010,
ICAS 2010 4.September (2010), pp. 2906–2912.
[11] Tomas Melin. Implementation Of A Vortex Lattice Method In A Heterogeneous Programming
Language Environment. 2018. URL: https://www.researchgate.net/publication/
327733453.
[12] André Deperrois. Guidelines for XFLR5: Analysis of foils and wings operating at low Reynolds
numbers. Tech. rep. February. 2013, pp. 1–72. URL: http : / / scholar . google . com /
scholar ? hl = en{\ & }btnG = Search{\ & }q = intitle : Analysis + of + foils + and +
wings+operating+at+low+Reynolds+numbers{\#}0.
[13] Mark Drela. XFOIL. URL: http://web.mit.edu/drela/Public/web/xfoil/ (visited on
05/05/2021).
[14] James C Sivells and Robert H Neelly. “Method for calculating wing characteristics by lifting-line
theory using nonlinear section lift data”. In: National Advisory Commitee for Aeronautics 865
(1947), pp. 75–93.
[15] Joseph Katz and Allen Plotkin. Low-Speed Aerodynamics. 2nd. New York: Cambridge Univer-
sity Press, 2001. ISBN: 978-0-521-66219-2.
[16] Brian Maskew. “Program VSAERO theory document”. In: Nasa Cr-4023 (1987). URL: http://
scholar.google.com/scholar?hl=en{\&}btnG=Search{\&}q=intitle:Program+
VSAERO+Theory+Document{\#}0.
[17] David Kinney. Using VSPAERO. URL: http : / / openvsp . org / wiki / doku . php ? id =
vspaerotutorial (visited on 06/03/2022).
[18] OpenVSP. Software Package, Version 3.22.0. 2021. URL: http://openvsp.org/ (visited
on 05/05/2021).
[19] David Kinney. VSPAERO. . . What’s New? 2021.
[20] H.-J. Steiner. Preliminary Design Tool for Propeller-Wing Aerodynamics Part II: Theory. Bauhaus
Luftfahrt., 2010.
[21] W F Phillips and D O Snyder. “Modern Adaptation of Prandtl’s Classic Lifting-Line Theory”. In:
JOURNAL OF AIRCRAFT 37.4 (2000). DOI: 10.2514/2.2649. URL: http://arc.aiaa.
org.
[22] Erik D. Olson and Cindy W. Albertson. “Aircraft high-lift aerodynamic analysis using a surface-
vorticity solver”. In: vol. 0. American Institute of Aeronautics and Astronautics Inc, AIAA, 2016.
ISBN: 9781624103933. DOI : 10.2514/6.2016-0779.
[23] CoolMUC-2 - Leibniz-Rechenzentrum (LRZ) Dokumentation. URL: https://doku.lrz.de/
display/PUBLIC/CoolMUC-2 (visited on 05/31/2022).
[24] George Loubimov and Michael Kinzel. “A novel approach to calculating induced drag from
computational fluid dynamics”. In: AIP Advances 11.7 (2021). ISSN: 21583226. DOI: 10.1063/
5.0051080.
13
