Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Design, Development and Manufacture of a New VTOL,

Canard UAV
Diogo Beirão Valente Henriques Tomás
diogo.beirao.tomas@tecnico.ulisboa.pt

Instituto Superior Técnico, Universidade de Lisboa, Portugal


January 2021

Abstract
In this project, a vertical take-off and landing (VTOL) unmanned aerial vehicle (UAV) with a canard
configuration was designed.
The design started by selecting the wing and canard geometry that satisfies the project requirements. To
do so, a lifting line theory program was developed and used to determine the aerodynamic performance
of the different lifting surfaces geometry.
The remaining control surfaces were designed based on empirical data and the fuselage geometry was
scaled down from a predetermined shape.
The structural design was made having in mind the availability of off-the-shelf items and their compatibility
with the model. A static load simulation was made, showing that the flexural stress limit wasn’t exceeded
in any structural component.
A total of 4 configurations were designed which only differ in the propulsion system distribution. Therefore
the structures that support the VTOL systems will also differ, resulting in 4 different outer mold lines
(OML). These were analyzed using computational fluid dynamics (CFD) and their aerodynamic efficiency
assessed. Penalties arising from exposure of VTOL motors and structural components to the flow
represent up to 13% decrease in L/D.
With CFD data, aerodynamic derivatives were calculated and with an initial estimate of the inertia, the
flight dynamics model was built. All dynamic modes are stable with the exception of the spiral, which
becomes less unstable as the cruise speed increases.
Configuration 1 was manufactured. The inertia was determined using the bifilar pendulum method. To
assess the structure, static load and hover tests were conducted, yielding satisfactory results.
Keywords:UAV, aircraft design, canard configuration, aerodynamic sizing, lifting line theory

1. Introduction With this in mind, in this project a proof of con-


1.1. Motivation, Project Overview and Requirements cept for a military application, whose purpose is
Technological advances in propulsion and control to detect magnetic anomalies that occur under the
increased the demand for aircraft that possess sea, will be fully designed. For this reason, the
VTOL capability while maintaining the cruise effi- prototype will be as similar as possible to the full
ciency of fixed wing aircraft. Investment on these scale aircraft, respecting several requirements. It
aircraft have propelled their development, render- has to be a canard configuration where the canard
ing them viable for multiple applications, from mili- stalls before the wing, with the wing producing lift
tary to urban air mobility. equivalent to 80% of the UAV’s weight and the ca-
In this context, achieving peak cruise performance nard the other 20%. For future reference this will
became essential to increase the endurance and be addressed as 80-20 lift distribution. The max-
mission range, which is obtained by employing the imum take-off weight (MTOW) must be 7kg, the
advantageous aerodynamic characteristics of a ca- UAV should cover 400 meters in 20 seconds or
nard configuration. This type of configuration is more and its static margin should be between 7
best suited for VTOL aircraft, as it eliminates its and 20%
main drawback, the need of longer runways. Four different proof of concept prototypes will be
To assess the design’s airworthiness as well its produced prior to the development of the full scale
capabilities and shortcomings, UAVs are utilized, aircraft. These configurations only differ in propul-
since they are less expensive to develop and have sive system, with each iteration being closer to the
less risk associated with them. intended for the full scale. Configuration 1 has

1
4 equidistant VTOL motors from the CG and a Prandtl’s theory models the wing as a bound vortex
pusher motor. Configuration 2 has 4 VTOL motors filament of strength Γ placed at the quarter-chord
with a ratio for the front rotors distance to CG to position, the lifting line. This vortex is shed as a
the rear rotors distance to CG of 4/1 and a pusher. vortex sheet at the trailing edge.
Configuration 3 has the same distribution as con- The lifting line theory can be represented by a sin-
figuration 2, but the front motors have the ability to gle equation, Equation 1, where s is the semi-span,
retract inside the fuselage, tilting the rear rotors to θ is the mapping angle, c is the chord, Clα is the
perform forward flight. Configuration 4 has a triro- lift coefficient slope, α∞ is the free stream angle
tor configuration with the same distribution as con- of attack and θt is the twist. All these values refer
figuration 2 and 3, with two motors in the front and to the airfoil at each section and therefore change
one on the back, having the same retracting mech- at each station, if the airfoil varies along the span.
anism and the motor on the rear has the ability to This equation’s only unknown are the coefficients
tilt, serving has a pusher during cruise flight. In this of An .
work, configuration 1 is fully designed and manu- ∞
factured while the others will only be designed and X Clα c
An sin (nθ)(sin θ + n )=
compared by means of computational methods. 8s
n=1 (1)
1.2. Literature Review Clα c
sin θ(α∞ + θt − α0 )
One of the first canard configurations dates back 8s
to 1903, with the first powered aircraft, the Wright
Flyer. The Wright brothers’ reasoning behind Knowing the An coefficients, the aerodynamic
the choice of a canard configuration was intuitive characteristics of the wing, namely the lift and in-
rather than supported by analytical/empirical data, duced drag coefficients, can be computed as fol-
since they wanted to avoid a control loss accident lowing
like the one that occurred in 1986, in an aft tail CL = πARA1 , (2)
configuration glider. The visibility of the control ∞
X
surface by the pilot was another factor that led CDi = πAR nAn 2 , (3)
the Wright brothers to the selection of a canard n=1
configuration. where AR is the aspect ratio of the surface [5]
Despite being controlled, the aircraft was unstable.
For an aircraft to be stable its center of gravity (CG) 3. Prototype Design
needs to lie in front of its neutral point (NP). In a 3.1. Airfoil Selection
canard configuration the NP is in front of the wing The airfoil selection was made with the aid of
aerodynamic center (a.c) while in a conventional XFLR5. In this software several NACA 4-digit se-
configuration the NP lies aft of the wing a.c. In ries airfoils were analysed at Re = 2.379 × 105 ,
the Wright Flyer most of the weight (biplane cell, which was obtained from the desired cruise speed
engine and pilot) was located in the biplane cell, and an initial estimate of the chord. From this study
placing the CG aft of the NP [1]. the curves of CL vs AoA and CL /CD vs AoA were
Several researches were conducted to determine obtained, the latter, which is presented in Figure 1,
which configuration (conventional or canard) was was used to select the desired airfoil.
superior, either aerodynamically, weight or stability From Figure 1, the airfoil NACA 4412 stands out
wise. These studies [2] [3] [4] concluded that the
canard advantages are comprised by better stall
control and better L/D when properly designed.
Regarding the disadvantages, canard configu-
rations require longer runways for take-off and
canard sizing is more critical than tail sizing since
it can highly affect the overall performance of the
aircraft.
Figure 1: CL /CD vs AoA curves for different airfoils
2. Theoretical Background
The theory selected for preliminary design is the as having the highest CL /CD among all the anal-
lifting line theory. Prandtl’s lifting line theory is used ysed airfoils, making it a good candidate for the air-
as preliminary calculation method to estimate the craft airfoil. However, the aircraft mission enters
characteristics of a finite wing. This theory is valid into play for the airfoil selection, since dash is also
for unswept wings with high aspect ratio under in- important. To trim the aircraft during dash, a low
compressible, inviscid and steady flow. CL is necessary, making the L/D low during this

2
flight condition. The higher the camber, the lower mined using
the L/D will be at higher speed, so to have a bet-
ter compromise between cruise and dash perfor- 1 2 Sw c¯w CMac w + Sc cc CMac c
Fcg = ρV + 0.2, (5)
mance the airfoil NACA 2412 was chosen. Both 2 cr dc−w W
wing and canard have the same airfoil.
where W represents the aircraft weight and dc−w
3.2. Lifting Surfaces Design represents the distance between wing and canard.
The lifting line theory was the selected preliminary Equation 5 was derived from
design method used to size the wing and canard.
−Mc −Mw +0.8W Fcg dc−w −0.2W dc−w (1−Fcg ) = 0,
To apply it, a MATLAB® program was developed,
(6)
which takes as inputs the air properties, the cruise
which is the moment balance equation, taking into
speed, the angle of attack (AoA) and the wing and
account a 80-20 lift distribution. Note that these
canard’s geometric characteristics. The wing char-
two equations assume the pitching moment to be
acteristics taken into account are: taper ratio (λ),
negative. Similarly to the CG, the neutral point po-
with a linear chord distribution along the span, twist
sition relative to the wing’s a.c., denoted Fnp was
(θt ) at the root and at the tip, assuming a linear
also calculated with
distribution along the span, aspect ratio (ARw ) and
area (Sw ). The canard was defined to be rect- CLα c SSWc
angular, so the only characteristics taken into ac- Fnp = . (7)
CLα c SSWc + CLα w
count are the aspect ratio (ARc ) and the ratio be-
tween wing and canard area (Fc-w ). To comply with
Having the CG and NP, the static margin (Kn ) is
the requirements this ratio has to be smaller than
obtained from
0.25. The distance between wing and canard aero-
dynamic centers (a.cs.) was also taken into ac- dc−w
Kn = (Fcg − Fnp ) . (8)
count. To determine the wing/canard combination c¯w
that best suits the project’s needs, a parametric
study was conducted varying the parameters men- To determine the trim AoA and canard incidence
tioned above. (ic ), the values of CL corresponding to the 80-20
The program can be subdivided into three subse- lift distribution were calculated for each surface.
quent parts. In the first part, the geometric prop- With these values and the data obtained in the first
erties of the wing and canard, such as chord dis- part of the program, an interpolation was made re-
tribution and span, are calculated based on the sulting in the AoA and ic for trimmed flight. With
input variables. Afterwards, the position of each these angles another interpolation was made to
spanwise station is computed, on which the 2D obtain the induced and parasitic drag of both sur-
aerodynamic properties are obtained from interpo- faces. Then, an approximation of the fuselage
lated airfoil data. With these values Equation 1 is drag, based on the distance between wing and ca-
solved for each surface, obtaining the CL and the nard, was made using data from a previous project
CDi from equations 2 and 3, respectively. These [6] and all the drag components were added, ob-
values are stored with respect to the AoA. To es- taining the total drag. Lastly the L/D was obtained
timate the parasitic drag of each surface, the CL by dividing the aircraft weight by the total drag.
at each section is calculated, based on the Kutta- In the third, and last, part of the program the stall
Joukowski theorem and Equation 4, to obtain the speed (Vs ) and the AoA at which stall occurs are
profile drag at each station by interpolating the CD determined. To do so, starting from the previously
vs CL data from the airfoil. Then the profile drag isobtained trim angles and cruise speed, the speed
integrated over the span. The CMac of the full sur- is gradually reduced. Since the canard stalls first,
face, for both wing and canard, was also calculated the criterion for stall was based on the CL at each
by integration. canard section and the CLmax of the airfoil. Stall
was considered to occur whenever the CL at a
X∞ given section was greater than or equal to CLmax . If
Γ(y) = Γ(θ) = 4sV∞ An sin (nθ) (4) stall is detected the angles at which it occurs were
n=1 stored, as well as the speed of the current iteration,
being this the stall speed. At each new speed the
The second part of the program computes the trim, forces and moments balance, presented in Equa-
using the lift and pitching moment contributions tion 9, is solved to obtained the wing and canard’s
from each surface. Using the previously obtained CL at the new trim condition. With these values
data, the CLα of the wing and canard is deter- the trim angles are computed with the process de-
mined. Afterwards, the position of the CG relative scribed previously, which will be used as the start-
to the wing’s a.c. position, denoted Fcg , is deter- ing point for the next iteration. This loop runs until

3
stall occurs. Figure 2 is a schematic representation
of the program’s algorithm.
(
LW + Lc = W
−Mc − Mw − dc−w ((1 − Fcg )Lc − Fcg LW ) = 0
(9)
Figure 4: L/D vs Sw from program results

Figure 5: L/D vs Fcg from program results


Figure 2: Program’s algorithm schematic

After analysing all configurations, plots of L/D vs


Vs , L/D vs Sw , L/D vs Fcg and L/D vs Kn , repre-
sented by Figures 3 to 6 respectively, were made
and studied. From Figure 3, it can be seen that
higher values of L/D are achieved for configura-
tions whose stall speed is in the range of 11 to 13
Figure 6: L/D vs Kn from program results
m/s. To achieve lower stall speed the wing area
has to be higher, however as seen in Figure 4, the
higher L/D are achieved with lower wing areas. L/D with low stall speed; have the smallest wing
Since the lift produced in cruise is the same for all span possible, for ease of manufacture and trans-
configurations, having a higher area will result in port and have a big enough distance (>= 1 m) be-
higher parasitic drag, thus having lower L/D. In tween wing and canard to mitigate any interference
Figure 5, five different regions can be highlighted, the latter might cause on the wing, since this pro-
each corresponding to a different value of wing and gram doesn’t account for the wing-canard interac-
canard distance. Note that the only difference be- tion.
tween each cluster of points is the distance be- The selected configuration has a cruise L/D of
tween the two surfaces. A larger distance will lead 19.23, a stall speed of 13 m/s, a static margin of
to a higher Fcg , refer to Equation 5, as well as lower 19% and the distance between surfaces is 1 m,
L/D, since the increase in distance will result in an leading to a Fcg of 0.1783. The selected configura-
increase in the fuselage length, therefore increas- tion has a wing with an area of 0.5 m2 , 2.4 m span
ing the drag and decreasing the L/D. In Figure 6, (bw ) and 0.208 m mean chord (c̄w ). Its taper ratio
the static margin range requirement is highlighted. is 0.8 and the twist at the tip is -2◦ . To achieve trim
These results are dependent on the ones shown in cruise, the wing incidence has to be 5◦ . While
in the previous figure, as displayed by Equation 8, the obtained canard has an area of 0.08 m2 , 1.02
thus an increase in the distance between wing and m span and a mean chord of 0.0784 m. In order to
canard will lead to an increase in the static margin, achieve trim, the canard incidence has to be 7◦ .
since the NP relative position is not affected by this
parameter. 3.3. Other Control Surfaces Design
The selection criteria used to obtain the configura- The remaining control surfaces were designed
from empirical data, resulting in an aileron whose
hinge line was located at 30% chord from the trail-
ing edge, the aileron’s span was 10% of the wing
span and it was located 1.5% of the wing span from
the wing tip.
The vertical tail was designed based on the vol-
ume coefficient (cV T ), of 0.012, obtained in a pre-
Figure 3: L/D vs Vs from program results vious work [6]. To avoid having a tall vertical tail,
the required total tail area was divided between two
tion that best suits the mission were a static margin surfaces, one on top of the fuselage,referred to as
between the required range (7%-20%); having high upper tail, and one below, referred to as lower tail.

4
The distance between the CG and the tail a.c., de- has one spar located at its quarter chord, as well
noted LV T , is 0.4 m. The total tail area (SV T ) is as the vertical tail. The wing has two spars, one
0.0364 m2 , with the upper tail having 60% of the placed at the quarter chord and another one at
total area. Both tails have a taper ratio of 0.4 and 61% of the chord at the root section and 71% of the
the upper tail has a height of 0.3 m and the lower chord at the tip section, to provide torsional rigidity
tail has a height of 0.2 m. Both have a root chord of to the wing.
0.104 m and tip chord of 0.0416 m. The airfoil se- To account for the manufacturing process, 3D print-
lected for both tails is the symmetric NACA 0012. ing, the wing structure was designed to ensure
The rudder design followed the same method as the highest structural strength and the least weight
the aileron one. The rudder hinge line was located possible.
at 20% chord from the tail trailing edge, the rud- To ensure rigidity of the aircraft, 3D printed pieces
der span was 80% of the tail height and it was lo- were designed whose purpose is to connect the
cated at 3% of the tail height from the tail tip. A spars. One connecting the canard spar and two of
rudder was made for both upper and lower tail. In the fuselage spars, another connecting both wing
the CFD analysis two situations were studied, one spars and all the fuselage spars and a last one con-
where both rudders where deflected and another necting the tail spar and two of the fuselage spars.
where only the upper one was deflected. This was The image of the structural model is presented in a
to determine if one rudder was enough to ensure section below.
sufficient authority. The results of this study are To accommodate the VTOL system of configura-
also presented in the CFD section below. tion 1, two booms connected to the canard and
wing were used, placed along the fuselage mid
3.4. Fuselage Design plane. These booms are also carbon fiber tubes
The fuselage was sized from an available model available at the supplier. Four additional connec-
of the full scale aircraft fuselage. To obtain the tors were made (two in each side of the aircraft) to
fuselage scaling factor, a SOLIDWORKS ® model connect the booms to the wing and canard.
was created were the wing, canard and vertical tail A rough design of the motor mounts was made and
were placed. The nose was defined to be at the placed where the motors for configuration 1 would
origin of the model and the CG was defined to be be. With this, the outer mold line (OML) of the air-
1.2 m aft of the nose. With everything placed and craft is finished and ready to use for CFD simula-
with the right distance and angle relations, the full tions. Figure 7 shows the OML of this configura-
scale fuselage was introduced in the same model, tion.
making the nose coincide with the origin. Then,
the scaling factor was changed until the vertical tail
was at the end of the fuselage. The resulting scal-
ing factor is 50%.
To confirm if the fuselage was big enough to ac-
commodate all the required avionics’ components,
the shape of these components was approximated
by cylinders and parallelepipeds, with the overall
dimensions of said components, and then intro-
duced inside the fuselage. A scaling factor of 50%
proved to be enough to fit all the required com-
ponents inside of the fuselage, yielding a fuselage
with 1.75 m in length (Lf use ).
Figure 7: Configuration 1 OML
3.5. Structural Design
Having performed the aerodynamic design, the The only differences, structure wise, between con-
structural design followed. The design was made figuration 1 and 2 are the VTOL system supports.
having in mind time of manufacture, availability of For this configuration, the booms will come out of
off the shelf items and structural rigidity. The de- the fuselage and connect to the front motors. The
sign for all configurations is made, presenting firstly rear motors will be placed in a support located be-
the structures that all have in common and later low the wing and more inboard, in relation to the
discussing the differences between each configu- rear motors of configuration 1. Since the front mo-
ration. tors will be in the front of the aircraft, a redesign
The main load carrying structures were chosen to of the front motor mounts was made, with the goal
be off the shelf carbon fiber tubes, acting as spars. of reducing drag. The OML of this configuration is
The fuselage has four spars that run from the nose presented in Figure 8
to the tail, placed in a square shape. The canard In configuration 3, during cruise, The OML will be

5
Mechanical® . Only configuration 1 was analysed.
The model is presented in Figure 11, where each
line corresponds to a carbon tube and the other
pieces are the plastic connectors. All the tubes
were modeled as beams and the contact points
were modeled as deformable and fixed, except for
the canard contact points which allowed rotations.

Figure 8: Configuration 2 OML Figure 11: Structural model used in structural analysis

After applying the loads, equivalent to a 4.5g pull


similar to the previous one, but without the front up maneuver, the structure is analysed obtaining
motors and their booms. Configuration 3 OML is the vertical displacement and the maximum stress.
presented in Figure 9. The maximum stress occurs at the wing spars with
a value of 320 M P a, which is lower than the max-
imum compressive stress (517 M P a) indicated by
the manufacturer. The maximum stress detected
at the 3D printed connectors was 7 M P a, which is
below the specified flexural strength for this mate-
rial (PETG).
This software also provides an estimate of the
inertia matrix of the system, which is given by
Ixx = 0.62512, Iyy = 0.71090, Izz = 1.3323 and
Ixz = 0.013, in kg.m2 . These values were used in
a first approach for the flight dynamics model.
On the same environment, a modal analysis was
conducted to determine the natural frequencies of
the aircraft structure. The results obtained for the
Figure 9: Configuration 3 OML
first 6 mode range from 6.43 to 13.29 Hz.
As for the forth configuration, during cruise, all the 3.7. CFD Analysis
VTOL systems will be retracted or tilted, making The CFD analyses were made in ANSYS CFX® us-
the airframe, i.e. canard, wing, tail and fuselage, ing a SST turbulence model in a steady state en-
its OML, presented in Figure 10. vironment with air at 25◦ C and 1 atm. A domain
was created around the aircraft with dimensions
of 5bw x10Lf use x16c¯w . The simulations were made
at cruise speed and with inlet boundary conditions
at the front face of the domain (where the aircraft
nose points to) and at the side faces of the domain
(4 faces), the rear face of the domain has an open-
ing boundary condition and the aircraft has a wall
boundary condition. For the inlet boundary the ve-
locity vector is specified, the opening boundary has
a relative static pressure of 0 and the wall bound-
ary has a no slip condition.
Before analysing all configurations a mesh conver-
gence study was performed to determine the num-
Figure 10: Configuration 4 OML ber of elements necessary to obtain satisfyingly ac-
curate results. From this study, it was decided that
a mesh with approximately 5million elements pro-
3.6. Structural Analysis duced satisfactory results.
To perform the structural analysis, a beam model A batch of analysis were made where the AoA var-
of the aircraft structure was created in ANSYS ied from -7 to 7◦ and the sideslip angle (β) from

6
0 to 5◦ . Note that the aircraft OML was made so
that the wing and canard are at their incidence an-
gles during cruise, therefore at 0◦ AoA the aircraft
should be trimmed. To achieve these different an-
gles, the velocity vector was computed based on
the current AoA and β and used as the inlet bound-
ary condition. These analyses were performed for
all configurations.
An analysis was also made were only the wing and
canard were included. This served as a compari-
Figure 13: L/D vs CL for different configurations
son term to validate the results obtained from the
lifting line theory program.
Lastly, four more simulations were made where the
The first comparison to be made is between the
control surfaces were deflected. All of these were
results from the lifting line theory program and its
made at 0◦ AoA: one where the canard incidence
equivalent from CFD. It can be seen that CFD pre-
was decreased 3◦ ; another where the ailerons
dicts a L/D 16 % higher than the lifting line the-
were deflected -15◦ ; one where both rudders were
ory (22.33 vs 19.23). This discrepancy is due to
deflected 15◦ ; and a last one where the upper rud-
the wing-canard flow interaction, which was not ac-
der was deflected 15◦ .
counted for in the lifting line theory program. In a
The aerodynamic derivatives with respect to the
canard configuration, the canard tip vortices cause
control surfaces, obtained in CFD, were compared
an upwash effect on the wing, which will increase
with empirical data from Cessna 172 [7], to assess
the lift produced by the wing, resulting in higher lo-
the control effectiveness of the surfaces as well as
cal AoA and overall L/D. This will also effect the
the need for a second rudder. The rates produced
trim at cruise, since the speed needed to achieve
by the deflection of the control surfaces were also
the same lift will be lower than the one used in lift-
determined, concluding that one rudder was suffi-
ing line theory. As it can be seen from Figure 12,
cient and that the remaining control surfaces pro-
trim (CM = 0) happens at CL between 0.72 and
duced satisfactory rates.
0.91, depending on the configuration, which results
The results obtained from the other simulations are
in a range of cruise velocities between 15.64 and
presented in Figure 12, where the relation between
17.65, which is also summarized in Table 1.
CM and CL for all configurations is presented, in
Configuration 1 and 2 have lower L/D since the
Figure 13, where the relation between L/D and
VTOL booms are exposed to the air flow during
CL is presented and in Table 1 where the trim L/D
cruise. Configuration 2 has the lowest L/D be-
is presented, and its variation with respect to con-
cause its booms are not parallel to the flow. In an
figuration 4, for all configurations as well as for
attempt to increase this value, a symmetric airfoil
the case with only canard and wing, alongside the
was applied to the booms. With this change, the
cruise speed at which trim is achieved (Vtrim ).
L/D increased 13.5% however, the booms with the
airfoil behaved like an additional canard, destabiliz-
ing the aircraft. Therefore, the airfoil booms were
rejected. The difference between configuration 3
and 4 is due to the quadrotor support located un-
der the wing which, in configuration 4, will be tilted
during cruise. This reasoning is the same behind
the different curves seen in Figure 13.
From this results is clear that configuration 1 and
Figure 12: CM vs CL for different configurations 2 would have significant range penalty, when com-
pared to configuration 3 and 4, although this range
difference would be reduced if the weight of the re-
tract and tilt mechanisms were accounted for.
Configuration L/D ∆L/D [%] Vtrim [m/s]
Regarding Figure 12, the pitching moment for all
Wing+canard 22.33 — 17.47 configurations is very similar, except for configura-
1 16.44 -13.93 17.26 tion 2. This can be explained by the location and
2 13.91 -27.17 15.64 orientation of the booms in this configuration.
3 18.53 -2.98 17.65 3.8. Flight Dynamics Model
4 19.1 0 17.06 To build the flight dynamics model (FDM), the
Table 1: Trim L/D for different configurations aerodynamic derivatives were determined from the

7
CFD results previously obtained. as for the rapid prototyping ability it provides. De-
With this information a MATLAB® program and a spite these advantages, 3d printing has its short-
Simulink® model were developed. The program comings, since the printer might malfunction during
works as initialization for the Simulink® model, as the printing process and additional post-print man-
it computes the reference state, i.e the trim of the ufacture needs to be done to make the part airwor-
aircraft for a certain flight condition, based on the thy.
derivatives, flight condition and wind condition. Only configuration 1 was manufactured.
To determine the aircraft trim and stability charac- After fully assembling the aircraft it was noted that
teristics at different cruise velocities, for all configu- the fuselage had low torsional rigidity, due to the
rations, in a range between 10 and 35 m/s, a new lack of rigidity of the plastic connectors for the fuse-
MATLAB® program was developed. This program lage spars.
uses the same algorithm as the previous one to
compute the trim conditions, which will be calcu- 4.2. Mass and Inertia
lated for each velocity analysed. After obtaining After manufacturing, assembling and placing all
the trim state, the dynamics matrix for both longi- the internal components (avionics, servos and har-
tudinal and lateral motions is built. Afterwards, the nessing) in the aircraft, its CG position, mass and
poles, damping and natural frequencies of the sys- inertia were determined.
tem are calculated with the aid of the MATLAB® To ensure that the CG was in the desired position,
function damp. These results allowed to conclude the aircraft was hung from two points in the booms
that all the modes are stable with the exception of that were aligned with the desired CG position. The
the spiral, which gets less unstable as the speed CG is in the right position if the aircraft remains hor-
increases. To determine if the structural vibration izontal while hung. In this case, the aircraft rotated
modes would be coupled with the aerodynamic upwards (tail heavy), so 100 g of lead were added
modes, the natural frequencies and damping coef- into the nose cone.
ficient were compared to the ones obtained in the To determine the mass, a scale was placed under
structural modal analysis, leading to the conclusion each landing gear and the value read in each scale
that the aerodynamic modes are not likely coupled was added, resulting in 7.118 kg, without the lead.
with the structural ones. To determine the inertia the BFP method was used.
In the Simulink® model, configuration 1 response A ceiling crane with wires was available at Cfar
to perturbations in the control surfaces was stud- (Center for aerospace research). To make use of
ied. The results show that, for a canard deflection, this device, and to ensure that the aircraft’s struc-
the aircraft returns to its reference state, therefore ture wasn’t damaged, a clamp was designed.
it’s deemed that the aircraft is longitudinally sta- To conduct BFP in all 3 axes, 5 clamps were placed
ble, as expected from the previous results. Two on the aircraft so that the CG would lie between the
distinct behaviours were detected. In the initial wires. The clamps were fixed and all the tests had
seconds, higher oscillation was seen in z velocity, the 5 clamps. Using a VectorNav sensor placed
which quickly dampens, making it possible to state at the CG, the aircraft was disturbed to oscillate
that initial motion is dominated by the short period around each axis. For each axis 3 measurements
mode. Afterwards, higher oscillations in x velocity were made. The inertia was computed from Equa-
were seen, which have low damping, characteristic tion 10, where T is the period of the oscillation, W
of the phugoid mode. is the weight of the test object, A is the distance
Looking at the response to the aileron deflection, it between wires and l is the length of the wires.
was seen that in the beginning of the motion there
T 2 W A2
were oscillations, characteristic of the dutch roll. I= , (10)
16π 2 l
After this, the oscillations are damped and an expo-
nential behaviour, characteristic of the spiral, can Afterwards, the parallel axis theorem was used to
be seen for the roll and yaw angles. This values remove the clamp contribution to the inertia, as-
diverge since the spiral mode is unstable. suming each clamp to be a point mass.
The same analysis can be done when the rudder An average was calculated between the inertia ob-
is deflected. Similarly, the initial motion was char- tained for each measurement, resulting in Ixx =
acterize by the presence of dutch roll. The spi- 1.2981, Iyy = 0.3274 and Izz = 2.5212, all in kg.m2 .
ral mode instability was also seen here, after the Ixz was assumed to be negligible when compared
damping of the dutch roll mode. to the other inertia.
4. Manufacturing & Experimental Testing 4.3. FDM update
4.1. Manufacturing With these results the FDM was updated to deter-
3D printing was chosen as the manufacturing mine if the aircraft was stable as well as to compare
method due to the project’s time constraints as well the frequencies of each mode with the structural

8
natural frequencies. No further conclusions regarding the work devel-
It was seen that with the new inertia the poles vari- oped could be drawn from the hover test.
ation with velocity remains the same for all modes, In Figure 15 a picture of the aircraft performing
except for the short period, which stops being os- hover is presented.
cillatory at higher speeds. The same behaviour
can be identified when studying the influence of low
static margins in the dynamic modes.
Regarding the natural frequencies obtained, these
were still lower than the structural ones, so it’s ex-
pected that coupling will not occur.

4.4. Static Load Tests


A wing static load test was conducted with a 3g
load to assess, qualitatively, the wing structure, as-
suming an elliptic lift distribution.
The load at each point of interest was computed Figure 15: Hover test
and sand bags, whose weight corresponded to the
calculated loads, were placed along the wing span,
which was flipped upside down. Figure 14 shows 5. Conclusions
the wing loaded with the sand bags as well as its During this research an UAV was sized, manufac-
deflection. tured and tested.
The process started by sizing the aircraft’s lifting
surfaces from Prandtl’s lifting line theory. To com-
ply with the requirements, the aircraft has a ca-
nard configuration where the wing produces 80%
of the lift and the canard produces the remaining
20%. The airfoil was selected for both surfaces
based on their performance at the desired Re and
a MATLAB® program, based on the lifting line the-
ory, was developed to perform a parametric study
on the wing and canard geometric characteristics,
while assessing the trim and stall conditions. From
this study, the wing and canard were selected tak-
Figure 14: Wing static load test ing into account the static margin requirement (7-
20%), the highest possible L/D while maintaining
From this test it was concluded that the wing can the lowest possible stall speed as well as the low-
support the loads of the operational limit (3g) and est possible wing area.
that the deflection seen is close to the one com- Having sized the wing and canard, the remaining
puted for a cantilever beam. control surfaces followed. These were sized based
on data collected empirically. The fuselage was
4.5. Hover Test also sized taking into account the internal compo-
The aircraft is now ready to flight test. In order to nent dimensions.
fly an UAV outdoors in Canada, the aircraft needs After completing the aerodynamic design, the
to be commissioned and registered to the authori- structural design followed. In this process, the
ties. This is a lengthy process which wasn’t able to availability of off-the-shelf items was taken into ac-
be completed within this work’ time frame. There- count, as well as their compatibility with the air-
fore, indoor hover tests were conducted to eval- craft’s airframe and internal components. The
uate the VTOL system and the controller perfor- wing structure was designed with the manufacture
mances. This test also allowed to evaluate the air- method in mind, 3D printing. Considerations were
craft’s structure under the loads generated in VTOL also made for the different configurations VTOL
flight mode. system supports. At this point, the OML of all con-
The hover tests were successful, proving that the figurations was available, making possible com-
structure is able to support the VTOL loads. Dur- putational simulations in both aerodynamics and
ing yaw maneuvers, high torsion in the front section structural domains.
of the aircraft, where the canards are located, was The structural analysis for configuration 1 was
detected. This was due to the low torsional rigidity made in ANSYS® Mechanical, where the aircraft’s
of the fuselage, which can be fixed by adding con- spars and connectors were designed. The spars
necting parts to the fuselage spars. were treated as beams. The structure was loaded

9
with the equivalent forces and moments produced and its design validated, the project can move on
at a 4.5g pull up maneuver and the deflection and to the remaining configurations, having configura-
stress were obtained. A modal analysis was also tion 4 has a end goal.
conducted to assess the aircraft’s natural frequen- Along the way, the manufacture process of the lift-
cies. ing surfaces can be improved by the use of foam
Afterwards, CFD analysis were conducted in machining (CNC) and the application of carbon
ANSYS® CFX for all configurations. These anal- fiber layers to their surface, for improved structural
ysis were made at different angles of attack and stiffness and weight reduction.
side-slip, also including the deflection of the con-
trol surfaces. With these results it was concluded References
that configuration 4 had the highest L/D, followed [1] F. E. C. Culick. The Wright Brothers: First aero-
by configuration 3 and then 1. Configuration 2 had nautical engineers and test pilots. AIAA Jour-
the lowest L/D. This study also provided the aero- nal, 41(6):985–1006, June 2003.
dynamic coefficients needed to develop the FDM [2] Joseph P. Landfield and Dario Rajkovict. Ca-
and assess the aircraft behaviour. nard/tail comparison for an advanced variable-
The determination of the aircraft dynamic be- sweep-wing fighter. Journal of Aircraft,
haviour was made using MATLAB® and Simulink® . 23(6):449–454, June 1986.
In a first instance, the dynamic matrix was built for
both longitudinal and lateral motion, and the poles, [3] Warren F. Phillips. Mechanics of Flight. John
natural frequencies and damping of each mode Wiley & Sons, 2nd edition, 2010. ISBN 978-0-
were obtained for all configurations. This analysis 470-53975-0.
was made for a range of cruise speeds between
10 and 35 m/s. The results obtained allowed to [4] Bruce P. Selberg and Kamran Rokhsaz. Aero-
conclude that all the modes were stable with the dynamic tradeoff study of conventional, canard
exception of the spiral, which got less unstable as and trisurface aircraft systems. Journal of Air-
the speed increased. Next, a Simulink® model was craft, 23(10):768–774, October 1986.
created to assess the dynamics of configuration 1. [5] John Anderson. Fundamentals of Aerodynam-
This model corroborated the results previously ob- ics. McGraw-Hill, 6th edition, 2017. ISBN 978-
tained. 1-259-12991-9.
The manufacture of the aircraft followed, were the
spars were made of carbon fiber tubes and the air- [6] Bruno Miguel Wong Luı́s. Design and devel-
frame was 3D printed. Low torsional rigidity was opment of a magnetic anomaly detection un-
detected on the fuselage due to the lack of imple- manned aerial vehicle. Master’s thesis, Instituto
mentation of 3D printed connectors. Superior Tecnico, 2019.
With the aircraft built, experimental tests were con-
ducted to determine the inertia, which was used to [7] Jan Roskam. Airplane Design Part VI: Prelim-
update the FDM. The modes behaviour remain the inary Calculation of Aerodynamic, Thrust and
same, with the exception of the short period, which Power Characteristics. DARcorporation, 1nd
became non oscillatory at higher speeds. edition, 1987. ISBN 978-1884885525.
A wing static load test at 3g load was conducted
to assess, qualitatively, the structure behaviour,
which was within expected.
Lastly, hover tests were conducted where the per-
formance of the overall aircraft structure was as-
sessed. The lack of torsional rigidity was visible
when performing yaw maneuvers, nonetheless the
aircraft’s structure was able to withstand the loads
produced in VTOL flight mode.

5.1. Future Work


For future work, a reinforcement of the fuselage
structure can be performed, mitigating the torsional
problems encountered.
There are plans to fly configuration 1 as soon as
possible, with different flight tests in mind, such as
detecting sunken ships with the magnetic sensor
(Qu-spin). Once configuration 1 has been flown

10

You might also like