CFD of x0.

7 scaled-down Northrop Gruman X47-B

UCAV
Haseeb Ur Rehman Syed Tauqeer Ul Islam
School of Aerospace and Mechanical Engineering Rislapur, School of Aerospace and Mechanical Engineering Risalpur,
Pakistan Pakistan
National University of Science and Technology National University of Science and Technology
Islamabad, Pakistan Islamabad, Pakistan
maseha1098@gmail.com tislam@cae.nust.edu.pk

Abstract— In this project, basic CFD analysis is performed on x0.7 scaled-down Northrop Grumman X47-B UCAV to evaluate
primary performance parameters. Model was acquired from literature and then it was refined to repair irregular surfaces and edges
using Space Claim and then it is made into a solid body previously being a surface body. Control surfaces were separated too and
made into a solid body as well. Then meshing was done, refined at leading edges and near the surface of the aircraft using inflation.
This was done in the Mesh part of ANSYS. Then by applying the boundary conditions in the Fluent Setup, results were obtained
which came out to be quite conforming to the general aerospace trend for Lift, drag and moments with angle of attacks and sideslips
and it shows that the UCAV is flight worthy as it is capable of producing sufficient lift at cruise velocity and angle of attack.

Keywords—UCAV, Scaled-down, CFD, Mesh, CAD, Inflation, FEM, Flow Similarity, Discretization, Spalart Allmaras, SDRs

I. INTRODUCTION
Economic methods of computational fluid dynamics (CFD) with sufficient precision are complementary to experimental
methods and play an increasingly important role in the simulation of maneuvering conditions. Examples are conditions that
cannot be simulated in a wind tunnel or are too dangerous to perform in flight test. One can get stability and control
characteristics for such conditions using CFDs. Before using the calculated stability and control characteristics, CFD methods
must be well validated and evaluated against the state of the art in the wind tunnel and flight test data or at least against any one
of the given.
Properly selecting the geometric detail level to use in a CFD simulation has a huge impact on profitability. The inclusion of
more geometric details leads to a more complex and therefore more expensive grid generation task. Adding more geometric
details will also result in grids with more cells. Since the calculation time and therefore the calculation cost is directly dependent
on the number of cells in the grid, adding more geometric details will result in more expensive simulations. However, omitting
these geometric details can lead to inaccurate or even incorrect results.
In aircraft aerodynamic design work, accurate prediction of the maximum lift coefficient, lift-to-drag ratio, Reynolds number
dependence, and improvement of their reliability in actual flight performance are important issues. These are the most
challenging tasks, because the flow characteristics in the multi-element airfoil include many types of boundary layer transition,
such as Tollmien-Schlichiting (TS) wave, bending instability, leading line instability, leading edge contamination and flow
separation, retransition, interaction, as well as its complicated geometry. Also, in general, like most wind tunnel tests, they are
done on a limited basis. Numerical Reynolds conditions using subscale models, the prediction contains many uncertainties due
to the Reynolds number effect of flow physics. On the other hand, due to the current development of CFDs that solve the
Reynolds mean Navier-Stokes (RANS) equations and computer technologies, it has been possible to make accurate predictions
for the cruise phase with minimal wind tunnel testing. Although an accurate prediction of aerodynamic forces is required even
for the design of a high lift system, it is not easy to achieve the same prediction accuracy with respect to the cruise phase due
to the complicated configurations and flow characteristics described above.
At the moment, Pakistan does not have its own UCAVs as development is yet to start in this field and by analyzing and
performing CFD of this fine and exquisitely designed state of the art UCAV, will definitely trigger more of such study and
innovation on this side of aerospace industry and help develop more of such UCAVs here locally. This industry is just increasing
and increasing worldwide and if we do not heed any importance to this side then there is a major possibility that we might be
left behind in this war of technology.
Here, a baseline CFD analysis of a UCAV is done based on previous knowledge and results are compared to the general aviation
and aerospace trends to check the fly-worthiness of the UCAV and whether the model is suitable and viable or not.
II. METHODOLOGY

A. Model Acquirement
The model was acquired from already available archive as X47-B is an established UCAV. Then, it was not what we required
it to be because our project encompasses much lighter and smaller scale UAV, and with a different mission profile and ultimately
different purpose too.

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

 Figure 1: CAD model initial CATIA V5
B. Modelling
The model, initially, contained surface bodies as they use terms in the CAD modelling lingo, but to have a valid CFD analysis
of the model, it was necessary to have a solid body model instead of a surface body model which needed quite a bit refining at
the start. Solid body is required because a surface body cannot be subtracted from another solid body which will be required
when enclosing the model in the pressure far-field. For this, every face was split and isolated and then it was checked whether
there were any overlapping faces or not. It was done in CATIA V5 software. Then check was made for any spaces or gaps
between adjacent faces so that there should not be any when final product is reached. Lastly, all the faces were joined back to
each other and along with that, it was being checked whether any two faces fail to join as that will again make the solid back to
surface with a gap in between the faces.

Figure 2: CAD model Space Claim

When solid body was achieved, then scale-down step was done and the whole UCAV was scaled-down to factor of 0.7. It means
that every dimension of the model was reduced to 70 % of its original length. For CFD purpose, inlet and exhaust of the engine
was close for simplification as it requires another mass flow in and out condition during the setup and processing of the CFD.
Also, during the meshing phase, mesh penetrates the inlet where there is no requirement to calculate any parameter and hence
results in unnecessary calculations.

Figure 3: Isometric view CATIA V5

 Table 1: Major geometric parameters of the UCAV
Fuselage Length 8.3 m
Wingspan 13.25 m
Planform Area 44 m2
Mean Aerodynamic Chord 3.9 m
(MAC)
Weight 34445 N
Centre of Gravity 0.4 MAC
For control surfaces separation, different platforms were used to separate and rotate them. To separate the elevators and elevons,
CATIA V5 was used. First the faces of the control surfaces were separated and isolated then, they closed from the faces they
were separated from and also faces on the UCAV were closed too. When control surface became separate body, then they were
moved a little bit away from the main fuselage and wing body so as to give them freedom to rotate. For further rotation and
translation of control surfaces and further modifications, ANSYS workbench Space Claim was used. Space Claim also helped
in cleaning and further refining the model.
Since the UCAV that is being worked on is a flying wing configuration, i.e., it is tail-less, hence there is no rudder or at least
no simple or conventional rudder is present on the UCAV. For yaw control, there are control surfaces present on both wings
up and down which provide yaw to the UCAV through drag differential. These are called Split Differential Rudders (SDRs).
They too are ought to be modelled. For that, in CATIA V5, surface was separated from the body and given an angle and
rotated about front hinge. Then the hollow part formed due to face separation was filled by making missing face there.
Similarly, downward SDR was also modelled same way just rotating downward.

Figure 1: SDR modelled

C. Meshing
Meshing part was done in the Fluent Mesh section where default meshing was done on inlet and outlets while face meshing
was done on walls of the UAV. Size of the meshing elements was varying depending upon level of detailed parameters required
in that region. For instance, more change is observed on the leading edges of wings hence sizing was put there of 0.01 m.
Similarly, more changes are to be observed on the control surfaces so 0.02 face sizing was put on the elevator and elevons as
well as Spit Differential Rudders.
 Figure 5: Surface mesh sizing

Now, Y plus is the dimensionless length and is defined as minimum length from the surface of the body first parametric value
is measured whether it be velocity, temperature or pressure etc. To reduce this value, Inflation was added in the meshing in
which far-field sphere was selected as first scoping geometry while aircraft surfaces were selected as boundary scoping
geometry. Type of inflation selected was ‘first layer thickness’ and first layer thickness value were given 0.05 while no. of
layers and growth rate were subject to change for every case.
During the meshing part, enclosure faces were named so as to constrain them in the Setup section. One face in front of the UAV
was named ‘Inlet’ while rest five faces were named ‘outlet’ as enclosure was very closely bound to the UAV and flow could be
assumed to have released from any of the sides except inlet. The surface of the solid UCAV model was named “walls”.

Figure 6: Far-field Mesh Visualization

Another setup of enclosure led to naming only two named selections. The outer hollow Sphere was named as far-field while
rest of the faces were named aircraft walls or simply aircraft. Radius of the far-field was set to 75 m so as not to interfere with
the flow and the aircraft surface.

Figure 7: Refined mesh on leading and trailing edges

D. Fluent Setup
Here, the environment is established around the mesh generated in the last step. For better results the air was set to ideal gas
which carried extra calculations of energy equation along with the viscous Spalart Allmaras equation. All the reference values
were calculated from the inlet and from that view the reference area was given to be 44 m2 which is the planform area of the
model. Also reference length was given 4 m which is mean aerodynamic chord of the UAV model. These values are already
mentioned in Table 1.
For each angle of attack case, the direction of the relative air flow was changed from setup instead of changing the pitch of
UCAV itself as it would have required to re-mesh the whole model which would be counterproductive albeit a bit accurate.
Mach No. was given 0.45 which is the cruise Mach no. of the UAV from design. This was converted into components in x and
z axes with cosine component of Mach No. in x direction and sine component in z direction. Simulation was done on several
angle of attacks from zero degrees with interval of 2 degrees or in various cases, 3 degrees, up until 18 degrees where stall was
observed for simple geometry. The stall angle kept changing slightly as the control surfaces came into play.
Method was set to COUPLED with all of the subsequent calculations set to SECOND ORDER UPWIND and controls were
revised slightly according to the setup requirements. Simple initialization was done using the tab computing from INLET.
E. Grid Independence Study
The computational domain has been carefully designed to obtain a blockage factor of less than 0.1%. Appropriate meshes were
created for each configuration, discretizing the computational domain in space and time through an adequate number of nodes
and elements, where Reynolds-Averaged-Navier-Stokes (RANS) equation was solved due to its ability to model complex
geometries and ensure high accuracy in calculating the normal and shear stresses acting on computational models. An
independent mesh analysis was performed to obtain the maximum precision in the results, without compromising the time of
each simulation. Comparison of different mesh sizes in CL and CD at same angle of attack are evident in following figures:

III. RESULTS AND DISCUSSION

Results are quite conformant with the trend as CL does increase linearly until stall point where it starts to drop
drastically. CL max came out to be 0.7 at almost 15 degrees angle of attack. Such low CL is because of the symmetrical nature
of wing used in the UCAV model but it does satisfy the trend. Similarly, Cm is according to the general trend too as Cm-alpha
comes out to be negative and Cm is linearly decreasing with increasing alpha. Then at stall, it becomes almost constant.
Following Table shows performance values derived from the CFD model:
Table 1: Performance values obtained from the CFD results
CLmax 0.75
CL - α 2.49/rad
Cm - α -0.48/rad
Cl (roll) - β -0.0285/rad
Cn - β 0.02/rad
L/Dmax 13
The model is generating lift sufficient to sustain maximum weight of the UCAV even at small angle of attacks but certainly is
not sufficient for maximum performance output. This is because the max L/D ratio is somewhere between 3- and 4-degrees
angle of attack while cruise trim angle is slightly changed at zero elevator and elevons deflection.
Following no. of cases were performed:
 Case No. of cases Range (Interval)

Zero control surfaces 9 From 0⁰ to +18⁰ α (2 degrees)

Zero control surfaces 7 From -10⁰ to +20⁰ β (5 degrees)

Elevators -20⁰ 10 From 0⁰ to +20⁰ α (2 degrees)

Elevators +20⁰ 9 From 0⁰ to +18⁰ α (2 degrees)

Elevators & Elevons -20⁰ 11 From 0⁰ to +22⁰ α (2 degrees)

Elevators & Elevons -10⁰ 11 From 0⁰ to +22⁰ α (2 degrees)

Elevators & Elevons +10⁰ 10 From 0⁰ to +20⁰ α (2 degrees)

Elevators & Elevons +20⁰ 9 From 0⁰ to +18⁰ α (2 degrees)

SDRs 20⁰ 10 From 0⁰ to +20⁰ α (2 degrees)

SDRs 20⁰ 5 From -10⁰ to +10⁰ β (5 degrees)

Ailerons 10⁰ 10 From 0⁰ to +20⁰ α (2 degrees)

Elementary plots with zero control surface deflections came out to be as following:
A. Aerodynamic Model
Aerodynamic model gives basic drag and lift coefficients through aerodynamic equations and it also evaluates amount of trim
angle of attacks, elevator deflections and lift coefficients are required for sustained flight at different velocities.
Now, from basic lift and moment equations:
𝐶𝐿 = 𝐶𝐿0 + 𝐶𝐿𝛼 . 𝛼 + 𝐶𝐿𝛿𝑒 . 𝛿𝑒
𝐶𝑚 = 𝐶𝑚0 + 𝐶𝑚𝛼 . 𝛼 + 𝐶𝑚𝛿𝑒 . 𝛿𝑒

Also at trim conditions, Cm will become zero while CL will become CL_trim.
By applying trim conditions and then slightly manipulating above equations, we get following forms:
𝐶𝐿𝑡𝑟𝑖𝑚 − 𝐶𝐿0 = 𝐶𝐿𝛼 . 𝛼 + 𝐶𝐿𝛿𝑒 . 𝛿𝑒
−𝐶𝑚0 = 𝐶𝑚𝛼 . 𝛼 + 𝐶𝑚𝛿𝑒 . 𝛿𝑒

2𝑊
Here, 𝐶𝐿𝑡𝑟𝑖𝑚 =
𝜌𝑣 2 𝑆

To get the aerodynamic model plots, above equations have to be converted into matrix form with known on one side and
unknown on the other. By doing so, we get following equation:
𝐶𝐿 − 𝐶𝐿0 𝐶𝐿 𝐶𝐿𝛿𝑒 𝛼
[ 𝑡𝑟𝑖𝑚 ]=[ 𝛼 ][ ]
−𝐶𝑚0 𝐶𝑚𝛼 𝐶𝑚𝛿𝑒 𝛿𝑒

Or,
𝛼 𝐶𝐿 𝐶𝐿𝛿𝑒 −1 𝐶𝐿𝑡𝑟𝑖𝑚 − 𝐶𝐿0
[ ]=[ 𝛼 ] [ ]
𝛿𝑒 𝐶𝑚𝛼 𝐶𝑚𝛿𝑒 −𝐶𝑚0

Now all the values on the right side are known and by varying CL_trim with velocity i.e., by making it a function of velocity, we
will get values of angle of attack and elevator deflection at different flight velocity. Following are the results of these
calculations:

B. Surface Y plus Reduction

Yplus value was initially discovered to be in thousands range which was then reduced to minimum as per the requirement of
Spalart-Allmaras model by refining near boundary mesh and increasing no. of mesh elements in the boundary layer formation
region. This was done by adding inflation in the meshing part and thus adding a size gradient from enclosure to UCAV surface.
Following Y plus contours were observed for steady level flight and at stall point after refining the mesh.

At 0 Angle of Attack:

And at 16 Angle of Attack:

C. Results with control surfaces deflections

Following are performance parameters found through CFD with control surface deflections:
 IV. CONCLUSION
After modelling the UCAV and setting it up for computational simulation, results are analyzed and compiled to compute
performance parameters and profiles and furthermore, stability check was also done through pitching moment vs alpha curve.
The lift curve slope shows that at considerably low alpha, trim lift coefficient can be achieved at cruise velocity. The lift
generated at higher alphas is sufficient for the aircraft for take-off and sustained flight. Pitching moment curve shows negative
slope which is according to the general trend for longitudinally stable aircraft. Also, changing the elevator deflection changes
pitching moment coefficient at the same alpha and hence changing the trim alpha.
Rolling moment curve has negative slope with beta too which again justifies lateral response of the UCAV and shows
conformance with general trend. Yawing moment curve has positive slope with beta which corroborates directional compliance
with the trend.
Maximum lift-to-drag ratio is achieved at reasonably lower angle of attack which is a desired characteristic in any aircraft. Drag
polar shows a trend-following dome too, and from there maximum lift-to-drag ratio point can be evaluated using geometry and
drawing straight line from origin of the drag polar tangent to the curve and point it touches the curve is maximum lift-to-drag
ratio point.
V. REFERENCES
[1] Boelens, O. J. (2012). CFD analysis of the flow around the X-31 aircraft at high angle of attack. Aerospace Science and Technology, 20(1), 38-51.
[2] Bravo-Mosquera, P. D., Cerón-Muñoz, H. D., Díaz-Vázquez, G., & Catalano, F. M. (2018). Conceptual design and CFD analysis of a new prototype of
agricultural aircraft. Aerospace Science and Technology, 80, 156-176.
[3] Wisnoe, W., Nasir, R. E. M., Kuntjoro, W., & Mamat, A. M. I. (2009, May). Wind tunnel experiments and CFD analysis of Blended Wing Body (BWB)
Unmanned Aerial Vehicle (UAV) at mach 0.1 and mach 0.3. In International conference on aerospace sciences and aviation technology (Vol. 13, No.
AEROSPACE SCIENCES & AVIATION TECHNOLOGY, ASAT-13, May 26–28, 2009, pp. 1-15). The Military Technical College.
[4] Prisacariu, Vasile. "CFD Analysis of UAV Flying Wing." INCAS Bulletin 8, no. 3 (2016): 65.
[5] Ishida T, Kawai S, Nakahashi K. A high-resolution method for flow simulations with Cartesian mesh method. AIAA Paper 2011-1296; 2011.
[6] Uddin H, Kramer RMJ, Pantano C. A Cartesian-based embedded geometry technique with adaptive high-order finite differences for compressible flow
around complex geometries. J Comput Phys 2014;262:379-407.
[7] Richtmyer RD, Morton KW. Difference Methods for Initial Value Problems. 2nd ed. London: Wiley-Interscience; 1967.
[8] Volpe G. Performance of compressible flow codes at low mach number. AIAA J 1993;31:49-56.
[9] Harlow FH, Welch JE. Numerical calculation of time-dependent viscous incompressible flow with free surface. Phys Fluids 1965;8:2182-9.
[10] Ho Y-H, Lakshminarayana B. Computation of unsteady viscous flow using a pressure-based algorithm. AIAA J 1993;31:2232-40.
[11] Javadia K, Darbandia M, Taeibi-Rahni M. Three-dimensional compressible-incompressible turbulent flow simulation using a pressure-based algorithm.
Comput Fluids 2008;37:747-66.
[12] Darwish M, Sraj I, Moukalled F. A coupled finite volume solver for the solution of incompressible flows on unstructured grids. J Comput Phys
2009;228:180-201.
[13] Shterev KS, Stefanov SK. Pressure based finite volume method for calculation of compressible viscous gas flows. J Comput Phys 2010;229:461-80.
[14] Chen ZJ, Przekwas AJ. A coupled pressure-based computational method for incompressible/compressible flows. J Comput Phys 2010;229:9150-65.
[15] Chorin AJ. A numerical method for solving incompressible viscous flow problems. J Comput Phys 1967;2:12-26.
[16] Turkel E. Preconditioned methods for solving the incompressible and low speed compressible equations. J Comput Phys 1987;72:277-98.
[17] Van Leer B, Lee WT, Roe PL. Characteristic time-stepping or local preconditioning of the euler equations. AIAA Paper 91-1552; 1991.
[18] Choi YH, Merkle CL. The application of preconditioning in viscous flows. J Comput Phys 1993;105:207-33.
[19] Turkel E. Review of preconditioning methods for fluid dynamics. Appl Numer Math 1993;12:257-84.
[20] Weiss J, Smith WA. Preconditioning applied to variable and constant density flows. AIAA J 1995;33:2050-7.
[21] Lee D. Local preconditioning of the Euler and Navier-Stokes equations. PhD Thesis, University of Michigan; 1996.
[22] Jespersen D, Pulliam T, Buning P. Recent enhancements to OVERFLOW. AIAA Paper 97-0644; 1997.
[23] Merkle CL, Sullivan JY, Buelow PEO, Venkateswaran S. Computation of flows with arbitrary equations of state. AIAA J 1998;36:515-21.
[24] Sharov D, Nakahashi K. Low speed preconditioning and LU-SGS scheme for 3-D viscous flow computations on unstructured grids. AIAA Paper 98-
0614; 1998.
[25] Mulas M, Chibbaro S, Delussu G, Di Piazza I, Talice M. Efficient parallel computations of flows of arbitrary fluids for all regimes of Reynolds, Mach
and Grashof numbers. Int J Numer Meth Heat Fluid Flow 2002;12:637-57.
[26] Puoti V. Preconditioning method for low-speed flows. AIAA J 2003;41:817-30.
[27] Briley WR, Taylor LK, Whitfield DL. High-resolution viscous flow simulations at arbitrary Mach number. J Comput Phys 2003;184:79-105.
[28] Sockol PM. Multigrid solution of the Navier-Stokes equations at low speeds with large temperature variations. J Comput Phys 2003;192:570-92.