Sunteți pe pagina 1din 6

Proceedings of the 2007 American Control Conference WeB03.

3
Marriott Marquis Hotel at Times Square
New York City, USA, July 11-13, 2007

The 6-DOF Dynamic Model and Simulation of the Tri-Turbofan


Remote-Controlled Airship
Michael T. Frye, Stephen M. Gammon, and Chunjiang Qian

Abstract— This paper presents the results of a modeling wing aircraft, airships fell out of favor during the conclusion
and research program for the Tri-Turbofan remote-controlled of the Second World War. Only during the last decade have
airship. The purpose of this research is to develop a dynamic airships been rediscovered as noted by the extensive list
model for the purpose of guidance, navigation, and control tech-
niques of autonomous airships. The Tri-Turbofan Simulation of airship programs in [1]. However, the earlier work of
Environment is a first principle force-based model which makes Burgess [2], Munk [3], and Tuckerman [4] on aerostatic and
use of a 6-Degree of Freedom nonlinear equation of motion dynamic airship design is influential and still relevant today.
of an airship. Also, the Tri-Turbofan Simulation Environment Airships have several obvious advantages over both rotorcraft
calculates the linear longitudinal and lateral-directional models and fixed wing aircraft due to their larger payloads and
around the velocity trim points. A Graphical User Interface was
developed in MATLAB and an open loop nonlinear Simulink long loiter times when coupled with such advanced control
model was integrated into the graphical interface. Finally, techniques. In fact, an airship’s loiter time is only limited by
the Tri-Turbofan simulation was verified against flight test the endurance of its crew; remove the crew and an airship’s
data acquired using the VICON optical measurement system. endurance is limited simply to maintaining its helium supply.
This paper illustrates the use of the Tri-Turbofan Simulation Easy to transport and store, an autonomous airship mobile
Environment by integrating a simple velocity hold autopilot
into a linear longitudinal model of the airship. cell phone network could be quickly established in the
disaster area. Furthermore, airships are easier to maintain
I. I NTRODUCTION and have lower operational costs when compared to either
A particularly revealing issue during the aftermath of Hur- rotorcraft or aircraft and therefore are an affordable and
ricane Katrina was the difficulty of establishing communica- efficient alternative. With the airship’s ability to maintain
tion among survivors, first responders, and the coordinating a higher altitude, a rescuer’s cell phone will have better
agencies in the areas hardest hit by Katrina due to damage reception. With the use of airships as a mobile cell phone
on both cell phone and landline telephone infrastructures. network, rescuers in the area of an emergency can better
A rapid restoration of communication is obviously very maintain communications with the coordinating agency as
important since it will allow for, well as allowing for better damage assessment and overall
1) Quick assessment of the damage to the area situational awareness. This will also allow survivors to be
2) Prompt rescue by locating survivors, many of whom quickly identified and located.
own cell phones The University of Texas at San Antonio (UTSA) Con-
3) Coordinated disaster relief management among first trols, Computations, and Cybernetics (C3) Laboratory has
responders and agencies. been researching the modeling, control, and requirements of
autonomous airships operating within a communication net-
A novel solution to the problem of maintaining com-
work. The goal of this research is to investigate the feasibility
munication during the advent of extensive damage to the
of using airships in a mobile communication network. The C3
communication infrastructure; be it from a flood, earthquake,
Laboratory is using both a 10 and 20 [ft] airship for testing
hurricane, or terrorist attack, is the use of autonomous
and verification of control techniques. An integral part of this
airships performing the function of a temporary mobile cell
research is the development of an accurate airship simulation
phone network. The use of airships as a communication net-
model for guidance, navigation, and control (GNC) analysis
work is not a new idea. As noted recently in [1], the United
and verification. The intent of the C3 Laboratory research
States Marine Corps has deployed the Marine Airborne Re-
is to apply advanced control techniques to airship dynamics,
Transmission System (MARTS) which is an Aerostat-based
which due to the large moment of inertia are difficult to
communication relay, for over-the-horizon transmission of
control. Prior to any integration of the GNC techniques on
UHF and VHF signals. However, like most airship based
the 10 and 20 [ft] airships, a high fidelity model which
communication networks, the MARTS makes use of aerostats
accurately captures the dynamics is necessary for GNC
which are tethered and immobile.
testing and verification. As part of the first phase of the
While airships have a very long and storied history of use
model development and simulation, a small tri-motor airship
and its design parallels for some time the history of fixed
was used for developing an understanding of airship flight
The authors are with the Dept. of Electrical and Computer Engr., The dynamics and for collecting flight test data. This was done in
University of Texas at San Antonio, San Antonio, TX. M. Frye will be order to save on the cost of helium and due to the sheer size
with the Dept. of Engr. Management at the University of the Incarnate
Word, San Antonio, TX, starting from July 2007. M. Frye’s email is: of the larger airships. The airship used is the Tri-Turbofan
michael.frye@utsa.edu remote-controlled airship shown in Fig. 1.

1-4244-0989-6/07/$25.00 ©2007 IEEE. 816


WeB03.3

TABLE I
T RI -T URBOFAN G ENERAL I NFORMATION .

Gross Lift 0.4590 [lb]


Net Lift (w/Ballast + Reflector) 0.0644 [lb]
Total Empty Weight (Envelope w/fins + Carriage) 0.3065 [lb]
Available Payload (w/o Ballast + Reflector 0.1525 [lb]
Available Payload (w/ Ballast + Reflector) 0.0567 [lb]
Density of Helium 0.0654 [lb/f t3 ]
Density of Air 0.0765 [lb/f t3 ]

TABLE II
E NVELOPE G ENERAL D IMENSIONS .
Fig. 1. Tri-Turbofan Remote-Controlled Airship.
Helium Capacity 5.88 [f t3 ]
Length of Envelope 44 [in]
Max Diameter of Envelope 21 [in]
The purpose of this paper will be to present the results Height of Envelope 25.5 [in]
of the 6-Degree of Freedom (6-DOF) nonlinear simulation Weight of Envelope w/o Fins 0.1389 [lb]
model of the Tri-Turbofan airship. The simulation environ- Weight of Envelope w/Fins 0.1586 [lb]
ment is based on a first principle force-based model and Max Cross-Sectional Area 1385.4 [in2 ]
makes use of both translational and rotational velocities and
TABLE III
accelerations. The Tri-Turbofan Simulation Environment also
C ARRIAGE G ENERAL D IMENSIONS .
allows for the importing and comparison of flight test data.
Finally, the Tri-Turbofan model calculates the longitudinal Length of Carriage 7.5625 [in]
and lateral-direction linear models based on its initial veloc- Width of Carriage 4.6875 [in]
ity trim points. The purpose of this research is to develop a Height of Carriage 2.5 [in]
Weight of Carriage w/ battery 0.1479 [lb]
knowledge of airship dynamics through a small manageable
airship with simple controls in order to keep helium cost
down and flight testing efficient.
This paper is organized as follows; Section II presents more accurate mass moments of inertia, two separate projects
the dimensions and mass properties of the airship, Section are under way; 1) calculating the moments from a Computer-
III presents the equations of motion, Section IV presents the Aided-Design software such as Solid Works and 2) applying
MATLAB/Simulink model based on the equations of motion, the techniques proposed in [7]. Table IV lists the mass
Section V provides a comparison between flight test and moments of the Tri-Turbofan airship.
simulated results, Section VI illustrates an example of the
TABLE IV
Tri-Turbofan Simulation Environment using a velocity hold
M ASS P ROPERTIES .
PI controller. Finally Section VII concludes and summarizes
the paper. Center of Volume (22.1732,10.5237,12.8745) [in]
Thickness Ratio 2.1
II. G ENERAL D IMENSIONS AND M ASS P ROPERTIES Ix 0.00073 [lbf t2 ]
Iy 0.0017 [lbf t2 ]
The Tri-Turbofan is a remote-controlled airship with two Iz 0.0016 [lbf t2 ]
fans oriented along the longitudinal axis and a single fan Ix y 0.0031 [lbf t2 ]
oriented along the vertical axis of the airship carriage. Note Ix z 0.0038 [lbf t2 ]
Iy z 0.0018 [lbf t2 ]
that the fans do not have any thrust vectoring capability. The
Tri-Turbofan’s Envelope is 44 [in] in length, 25.5 [in] in
height, and has a 21 [in] maximum diameter. The airship
III. T RI -T URBOFAN A IRSHIP E QUATIONS OF M OTION
makes use of a cross configuration for the two vertical and
two horizontal stabilizers as shown in Fig. 1. The airship The Tri-Turbofan Simulation Environment is based on a
does not have any rudder or elevator control inputs and relies first principle force-based model and shares certain similar-
solely on propulsion for control. The Tri-Turbofan holds ities with traditional aircraft and rotorcraft dynamics with
approximately 6 [f t3 ] of helium. Tables I, II, and III provides the most significant difference being the apparent mass and
the general data for the Tri-Turbofan airship, its envelope, inertial effects of air on the airship hull. Due to the size
and its carriage. of an airship’s envelope volume, the displaced air’s mass
The moments of inertia and apparent moments of inertia and inertia effects on the dynamics of an airship can not
for the Tri-Turbofan airship were calculated analytically be ignored as is the case in aircraft design [8] and [2].
based on the geometry and dimensions of the airship using Equation (1) relates the airship response to the forces and
the techniques from [5]. The VICON optical motion mea- moments acting on the airship. The Tri-Turbofan Simulation
surement system was used to accurately measure the static Environment solves for the following states,
dimensions and is discussed in [6]. In an effort to calculate • u - longitudinal velocity

817
WeB03.3

• v - lateral velocity The nonlinear equations of motion from (2) can be restated
• w - vertical or normal velocity as, M ẋ = A(x) + Fa + Fb + Fg + Fp where M is the
• q - pitch rate Mass matrix containing the mass, inertia, apparent mass,
• p - roll rate and apparent inertia coefficients. A(x) is the aerodynamic
• r - yaw rate. equations of the airship. Finally, Fa + Fb + Fg + Fp are
The 6-DOF first principle force model described by the the aerodynamics forces on the airship hull, buoyancy force,
following equations of motion for an airship is based on [9] gravity force, and propulsion force, respectively.
and roughly describes ma = F , A. Mass Matrix
mx u̇ + (maz − Ẋq̇ )q̇ + · · · The Mass Matrix, as derived from (2) and using the
notation Ki and Ḱi , is defined as follows,
+ mz qw − my rv − maz pr − max (q 2 + r2 ) = X
my v̇ − (maz + Ẏṗ )ṗ + (may + Ẏṙ )ṙ + · · · M =
m − K1 0 0

+ mx ru − mz pw + max pq + maz qr = Y
 0 m − K2 0
mz ẇ − (max − Żq̇ )q̇ + · · · 
 0 0 m − K3
+ my pv − mx qu + max pr − maz (p2 + q 2 ) = Z


 0 −maz − Ḱ1 0
Jx ṗ − (maz + L̇v̇ )v̇ − Jxz (ṙ + pq) + · · ·  maz − Ḱ1 0 −max − Ḱ2
− (Jy − Jz )qr − maz (ru − pw) = L 0 max − Ḱ3 0
0 maz − K1 0

Jy q̇ + (maz − Ṁu̇ )u̇ − (max + Ṁẇ )ẇ + · · ·
−maz − K2 max − K2 0
+ (Jx − Jz )pr − Jxz (r2 − p2 ) + · · ·

0 0

−max − K3 
− max (pv − qu) + maz (qw − rv) = M Ix − Ḱ1 0 −Jxz


Jz ṙ − Jxz (ṗ − pr) + (max − Ṅv̇ )v̇ + · · ·

0 Iy − Ḱ2 0 
− (Jx − Jy )pq + max (ru − pw) = N (1) −Jxz 0 Iz − Ḱ3

where the apparent masses are defined by mx = m − The translational moments of inertia are K1 , K2 , K3 and
Ẋu̇ , my = m − Ẏv̇ , and mz = m − Żẇ . The apparent the rotational apparent moments of inertia are Ḱ1 , Ḱ2 , Ḱ3 .
moments of inertia are defined as Jx = Ix − L̇ṗ , Jy = The translational apparent moments of inertia calculations
Iy −Ṁq̇ , and Jz = Iz −Ṅṙ . Finally the apparent products are based on the geometric techniques from [5], [10], and
of inertia are Jxz = Ixz + Ṅṗ and Jxy = Jyz = 0. [11],
Rearranging (1) such that the rates are on the left hand 4pi α0
K1 = abck1 k1 = 2−α
side, the effects of the apparent masses and inertia are clearly 3 0

illustrated, 4pi β0
K2 = abck2 k2 = 2−β0
3
mx u̇ + (maz − Ẋq̇ )q̇ = 4pi γ0
K3 = abck1 k3 = 2−γ ,
− mz qw + my rv + maz pr + max (q 2 + r2 ) + · · · 3 0

+ Xa + Xb + Xg + X p where a, b, and c are the hull dimensions to the center of


gravity and are 22, 12.75, and 10.5 [in] respectively.
my v̇ − (maz + Ẏṗ )ṗ + (may + Ẏṙ )ṙ = The rotational apparent moments of inertia calculations
− mx ru + mz pw − max pq − maz qr + · · · are also based on [5], [10], and [11] and are defined as,
+ Ya + Yb + Yg + Yp 4pi b2 + c2 ´ 2 2
Ḱ1 = abc k1 k´1 = ( bb2 −c
+c2 )
2 γ0 −β0
mz ẇ − (max − Żq̇ )q̇ = 3 5
2
2( bb2 −c
2
+c2
)−(β0 −γ0 )

− my pv + mx qu − max pr + maz (p2 + q 2 ) + · · · 4pi c2 + a2 ´ 2 2


Ḱ2 = abc k2 k´2 = ( cc2 −a
+a2 )
2
2
α0 −γ0
2
+ Za + Zb + Zg + Zp (2) 3 5 2( cc2 −a
+a2
)−(γ0 −α0 )

Jx ṗ − (maz + L̇v̇ )v̇ − Jxz ṙ = 4pi a2 + b2 ´ 2 2


Ḱ3 = abc k3 k´3 = ( aa2 −b
+b2 )
2 β0 −α0
2 −b2 .
+ (Jy − Jz )qr + maz (ru − pw) − Jxz p + · · · 3 5 2( a
a2 +b2
)−(β0 −α0 )

+ La + Lb + Lg + Lp The variables α0 , β0 , and γ0 are solved by


Jy q̇ + (maz − Ṁu̇ )u̇ − (max + Ṁẇ )ẇ = 2abc
α0 = 1 (µ0 − E(µ0 ))
− (Jx − Jz )pr + Jxz (r2 − p2 ) + max (pv − qu) + · · · (a2 − b2 )(a2 − c2 ) 2
1
− maz (qw − rv) + Ma + Mb + Mg + Mp 2abc(a2 − c2 ) 2 b2 − c2 (a2 − b2 )c
β0 = 2 2 2 2
(E(µ0 ) − 2 2
u0 − 1 )
(a − b )(a − c ) a −c ab(a2 − c2 ) 2
Jz ṙ − Jxz ṗ + (max − Ṅv̇ )v̇ =
ac
1− 1 E(u0 )
− Jxz pr + (Jx − Jy )pq − max (ra − paw) + · · · b(a2 −c2 ) 2
γ0 =2 ,
+ Na + Nb + Ng + Np . 1 − ( cb )2

818
WeB03.3

where u0 and E(u0 ) are elliptic integrals and defined as were calculated as follows,
α
√ CD = CX1 cos2 (α)cos2 (β) + CX2 sin(2α) sin( )
a2 − c2
Z ϕ0
dθ 2
ϕ0 = sin−1 µ0 = β
CY = CY 1 cos( ) sin(2β) + CY 2 sin(2β) + CY 3 sin(β) sin(|β|)
p
a 0 1 − k 2 sin2 θ 2
r ϕ0 p α
a2 − b2 CL = CZ1 cos( ) sin(2α) + CZ2 sin(2α) + CZ3 sin(α) sin(|α|)
Z
k = sin−1 E(µ0 ) = 1 − k 2 sin2 θdθ. 2
a2 − c2 0 Cl = Cl2 sin(β) sin(|β|)
α
Cm = Cm1 cos( ) sin(2α) + Cm2 sin(2α) + Cm3 sin(α) sin(|α|)
2
B. Aerodynamic Equations β
Cn = Cn1 cos( ) sin(2β) + Cn2 sin(2β) + Cn3 sin(β) sin(|β|),
2
The 6-DOF nonlinear aerodynamic equations in A(x)
represents six states; u, v, w which are the translational pwhere the airship velocity is defined as Vt =
velocities and p, q, r which are the rotational velocities. (u2 + v 2 + w2 ). The angle of attack, α, is defined as
The aerodynamic equation matrix A(x) includes the cross- α = tan−1 ( w u ) and the angle of side-slip, β, is defined as
coupling and high order effects as shown below and are from
(2), β = sin − 1( Vvt ).
The aerodynamic coefficients (s.t. CX1 , CY 1 , etc.) were
derived based on the techniques developed by [12] and [13]
A(x) = and the equations from [14]. The techniques from [12] make
−mz qw + my rv + maz pr + max (q 2 + r2 )
 
use of describing the airship geometrically and based on the
 −mx ru + mz pw − max pq − maz qr 
hull geometry to estimate the aerodynamic coefficients. For
−my pv + mx qu − max pr + maz (p2 + q 2 )
 
the sake of brevity, the equations are not shown in this paper.
 
(Jy − Jz )qr + maz (ru − pw) − Jxz p  (3)
 


 −(Jx − Jz )pr + Jxz (r2 − p2 ) + · · ·

 Currently, we are investigating using a wind tunnel to ex-
 +max (pv − qu) − maz (qw − rv)  plicitly calculate the drag and lift coefficients. It is expected
−Jxz pr + (Jx − Jy )pq − max (ra − paw) that wind tunnel data will prove to be more accurate.

D. Buoyancy Equations
where the apparent mass from (3) are defined as mx =
m − K1 , my = m − K2 , and mz = m − K3 . Furthermore, The gravity and buoyancy forces, Fg and Fb , were com-
ax = −1 [in] and az = −2 [in] and are the distances of the bined together and written as follows using the gravity and
center of gravity relative to the center of volume. Finally, the buoyancy model from [9],
apparent moments of inertia are defined as, Jx = Ix − K1 ,  

(cos φ sin θ cos ψ + sin φ sin ψ)(W − B)

Jy = Iy − K2 , and Jz = Iz − K3 . Xb
 Yb   (cos φ sin θ sin ψ − sin φ sin ψ)(W − B) 
 
   (cos φ cos θ)(W − B) 
 Zb   
  =  (cos φ sin θ sin ψ − sin φ sin ψ)az W ,
 Lb   
C. Force Equations    (cos φ sin θ cos ψ + sin φ sin ψ)az W + · · · 
 Mb   
 −(cos φ cos θ)ax W 
The force equations summarize the net effect of the Nb
(cos φ sin θ sin ψ − sin φ sin ψ)ax W
aerodynamic forces and moments against the hull of the Tri-
Turbofan airship where Xa is the axial force, Ya is the side where W is the weight of the airship envelope and carriage.
force, Za is the vertical or normal force, La is the rolling B is the gross lift of the airship and is calculated by
moment, Ma is the pitching moment, and Na is the yawing B = V ρa . The buoyancy equations also require the airship
moment. They are defined as, attitudes φ, θ, ψ.

E. Propulsion Equations
1 2 23
 
2 CD ρVt V 2
 
Xa The forces due to propulsion are defined as,
1 2 3
Ya 2 CY ρVt V 2
       
  Xp Ts + Tp

1 2 3

Za  
2 CL ρVt V 2
 
 =   Yp   0
1 2 3

La  
2 Cl ρVt V 2
     
     Zp   Tz 
Ma   1 2 3 =
 ,
2 Cm ρVt V 2
 
 Lp   0
  
Na 1 2 3
2 Cn ρVt V
   
 Mp   (Ts + Tp )dz 
Np T p − T s dy
and are a function of individual aerodynamic coefficients,
where Ts represents the starboard propeller thrust, Tp repre-
air density ρ, airship velocity Vt , and the airship hull surface
sents the port propeller thrust, and Tz represents the vertical
area [9].
propeller thrust. The Tri-Turbofan airship does not have any
The aerodynamic coefficients CD , CY , CL , Cl , Cm , Cn
are calculated based on the geometry of the airship hull thrust vectoring capability and simply relies on the either
and configuration of horizontal and vertical stabilizers on forward or reverse thrust to maneuver. Table V lists the
the Tri-Turbofan. Using [14], the aerodynamics coefficients individual forward and reverse propeller forces.

819
WeB03.3

TABLE V
translational rates were calculated by taking the derivative of
P ROPULSION TABLE .
the position time history.
Control F orward Reverse Figs. 3, 4, and 5 compare the results of the Tri-Turbofan
Ts +0.0080 [lb] -0.0048 [lb] Airship model and flight test data while in forward flight
Tp +0.0080 [lb] -0.0048 [lb] with a delayed vertical step input. Fig. 3, which compares the
Tz +0.0099 [lb] -0.0026 [lb]
longitudinal position x and velocity u, reveals a divergence
in both x and u after 4 [sec]; however the model time history
trends similarly to the flight test data.
The variables dz and dy are the distances of the propellers
from the center of volume and are dz = 1.0729 [ft] and Longitudinal Comparison
dy = 0.4583 [ft]. 75

70
IV. N ONLINEAR S IMULATION Flight

x [in]
Model
The Tri-Turbofan Airship equations of motion were in- 65 Frwd

tegrated into MATLAB/Simulink for rapid simulation and 60


0 1 2 3 4 5 6 7
controller integration. The simulation was developed to allow
2
for the ability to
0
1) analyze flight test time histories

u [in/s]
−2
2) simulate open/close-loop nonlinear dynamics
−4
3) calculate the longitudinal/lateral-directional linear dy-
−6
namics 0 1 2 3 4 5 6 7
Time [sec]
4) integrate controllers for testing
5) compare flight test and simulated time histories. Fig. 3. Longitudinal Velocity Data Comparison.
Fig. 2 is the nonlinear Simulink model of the Tri-Turbofan
Airship. The primary block components shown in Fig. 2 are Fig. 4 compares the lateral position y and velocity v and
the equations previously discussed in Section III. reveals a large divergence at 3 [sec]. This divergence is due
to the apparent mass discrepancy in the airship model and is
[Xb,Yb,Zb,Lb,Mb,Nb]

1
[phi,theta,psi]
MATLAB
Function being investigated.
Bouyancy Model
frwd1
[Xp,Yp,Zp,Lp,Mp,Np]
1 MATLAB Apparent Mass & Interia
Function
frwd2 Engine Model M^(−1) Lateral Comparison
Constant −16
1
u
[x,y,z,phi,theta,psi] Flight
vert
v
−18 Model
M^−1 [u,v,w,p,q,r]
Matrix Roll
y [in]

w 1 1
[u,v,w,p,q,r] Multiply
Input
p
s s
−20
Integrator Integrator1
Matrix Product
q
[u_dot,v_dot,w_dot,p_dot,q_dot,r_dot]
−22
r
Airship Nonlinear Dynamics

MATLAB [Xa,Ya,Za,La,Ma,Na] −24


Function 0 1 2 3 4 5 6 7
AeroDynamic Model

2
Fig. 2. Nonlinear 6-DOF Tri-Turbofan Simulink Model.
v [in/s]

−2

V. T RI -T URBOFAN M ODEL V ERIFICATION −4

−6
0 1 2 3 4 5 6 7
The airship flight test collection of this project presented Time [sec]

a technical challenge due to the limited payload of the


Tri-Turbofan Airship. The airship has a maximum payload Fig. 4. Lateral Velocity Data Comparison.
of 0.1763 [lb] and due to this, avionics instrumentation
could not be installed. Airship data collection and aerostatic Fig. 5 compares the vertical position z and vertical velocity
modeling is addressed in detail in [6]. w. The comparison reveals a divergence at 1 [sec], however
The Gait Analysis and Innovative Technologies (GAIT) the model trends similarly with the flight test data.
Laboratory at the University of Texas Health Science Center
at San Antonio was the ideal solution to overcoming the VI. A IRSHIP V ELOCITY H OLD M ODE AUTOPILOT
airship weight constraint. The GAIT Laboratory operates an
eight camera VICON optical motion measurement system As an example of the Tri-Turbofan Simulation Environ-
that emits an infrared pulse which reflects off a series of ment, the nonlinear airship model was linearized and the
markers that were on the airship. The pulses illuminate one following A and B system matrices were calculated,
of 18 reflective spheres that are placed onto the airship. The   
−.0409 0 0 −9.8693 u
VICON camera system samples at a 120 [Hz] and optically 0 −36.8383 −0.8930 0  w 
A= −10.5885   q  ,

captures the X, Y, and Z coordinates of each of the 18 0 126.5885 −2.8317
reflective marker spheres on the Tri-Turbofan Airship. The 0 0 1 0 θ

820
WeB03.3

Vertical Comparison Velocity Hold Time History


36 3.5
Flight
Model
34 3
Up

z [in] 2.5
32

Velocity [in/sec]
30
0 1 2 3 4 5 6 7
1.5

4
1
3
w [in/s]

2 0.5

1
0
0

−1 −0.5
0 1 2 3 4 5 6 7 0 5 10 15
Time [sec] time [sec]

Fig. 5. Vertical Velocity Data Comparison. Fig. 7. Velocity Hold Time History.

where the states are u, w, q, θ and single GUI interface. Finally, flight test data was collected
  using the VICON optical motion measurement system which
0.016
 0.0099  £ ¤ recorded the position of the x, y, z coordinates of the Tri-
B =  0.017  δT .
 Turbofan Airship.
0 VIII. ACKNOWLEDGMENTS
Note that at this trim point, the airship has the follow- This work was supported in part by the U.S. National
ing eigenvalues, λ1 = −0.0409, λ2 = −33.1484, λ3 = Science Foundation under grant ECS-0554117 and the UTSA
−3.2608 + 1.0651i, λ4 = −3.2608 − 1.0651i. In open loop, Faculty Research Award. The UTSA C3 laboratory would
the first eigenvalue, λ1 , describes the heave mode of the like to gratefully acknowledge the use of the equipment and
airship, λ2 describes the pitching mode, and finally the λ3 personnel at the University of Texas Health Science Center
and λ4 are the pendulum mode of the airship. at San Antonio GAIT Laboratory during the airship flight
A simple proportional/integral velocity hold mode au- testing.
topilot was developed to hold the airship at a longitudinal R EFERENCES
velocity of u = 3. The Simulink model used is shown in
[1] S. A. Cambone, K. Krieg, P. Pace, and L. Wells, “Unmanned Aircraft
Fig. 6. Systems Roadmap 2005-2030,” Office of the Secretary of Defense,
United States Deparment of Defense, Washington, DC, August 2005.
[2] C. P. Burgess, Airship Design, The Ronald Press Company, New York;
1927.
A [3] M. M. Munk, “The Aerodynamic Forces On Airship Hulls,” NACA,
Dynamic Matrix

Matrix Ax
NACA-Report-184, 1924.
x 1 x 1
Multiply
Bu
s s
positions
[4] L. B. Tuckerman, “Notes On Aerodynamic Forces On Airship Hulls,”
Product Integrator Integrator3 To Workspace
x_dot
NACA, NACA-TN-129, 1923.
10

Proportional Gain B
[5] L. B. Tuckerman, “Inertia Factors of Ellipsoids For Use In Airship
rates
Velocity Hold Mode Control Matrix
To Workspace1
Design,” NACA, NACA-TN-210, 1925.
MATLAB
Function Matrix
[6] S. M. Gammon, M. T. Frye, and C. Qian, “The Development of the Tri-
Product2 Multiply
MATLAB Fcn
u
Product1
Turbofan Airship Model for Autonomous Flight Control Research,”
.75 1
3
Gain
s
Integral Gain
Proceeding of the 2006 AIAA Guidance, Navigation and Control
Vref
Conference, Workshop, and Exhibit, Keystone, Colorado, August 2006.
[7] M. P. Miller, “An Accurate Method of Measuring the Moments of
Inertia of Airplanes,” NACA, NACA-TN-351, 1930.
[8] S. B. V. Gomes and J. G. Ramos, “Airship Dynamic Modeling for
Autonomous Operation,” Proceeding of the 1998 IEEE International
Fig. 6. Velocity Hold Mode Simulink Model.
Conference on Robotics and Automation, Leuven, Beligium, May
1998, pp. 3462-3467.
Fig. 7 is the time history for the velocity hold autopilot. [9] G. A. Khoury (ed.) and J. David Gillett (ed.), Airship Technology,
Cambridge University Press, Cambridge, England; 1999.
VII. C ONCLUSION [10] M. M. Munk, “Some Tables of The Factor Of Apparent Additional
Mass,” NACA, NACA-TN-197, 1924.
This paper summarized the development of a first principle [11] H. Bateman, “The Inertia Coefficients of An Airship In a Frictionless
force-based model of the Tri-Turbofan Airship based on ge- Fluid,” NACA, NACA-Report-164, 1924.
ometric and flight test analysis. The Tri-Turbofan Simulation [12] S. P. Jones and J. D. DeLaurier, “Aerodynamic Estimation Techniques
for Aerostats and Airships,” Journal of Aircraft, vol. 20, No. 2, 1982,
Environment was designed to incorporate the mathematical pp. 120-126.
model of the equations of motion in a tool that would allow [13] R. Upson and W. Klikoff, “Application of Practical Hydrodynamics
for rapid prototyping of controller integration and analysis. to Airship Design,” NACA, NACA-Report-405, 1933.
[14] J. B. Mueller, M. A. Paluszek, and Y. Zhao, “Development of an
The Simulation Environment allows for flight test analysis, Aerodynamic Model and Control Law Design for a High Altitude
nonlinear simulation, linear simulation, controller integra- Airship,” Proceeding of the AIAA 3rd Unmanned Unlimited Technical
tion, and flight test/simulation data comparison through a Conference, Workshop, and Exhibit, Chicago, IL, September 2004.

821

S-ar putea să vă placă și