Você está na página 1de 26

2D Aircraft Trim

Holly Lewis
April 4, 2003












2
Abstract

The purpose of this document is to discuss a 2-dimensional trim model for an aircraft
along with the necessary mathematical methods to solve the problem, the computer
implementation of the method and the results.

The aircraft is climbing at a constant translational velocity, i.e. the aircraft is not
accelerating. The 2-dimensional simplification; restricting the motion of the aircraft to
longitudinal motions only; is made possible by assuming the sideslip is zero, the aircraft
is symmetric and the aircraft is climbing at an angle, . Given the aircraft geometry and
flight conditions, the aerodynamic forces and moments acting on the aircraft can be
computed using the Aero-F computer program (Ref. [2], Ref. [3] & Ref. [4]). It is
important to note that the output of the Aero-F program is defined in a reference frame
where the origin is not at the center of gravity; thus the Aero-F reference frame is clearly
defined below, in order for use in the development of the trim model. Then the equations
for the longitudinal forces and moments on the aircraft are developed, resulting in a
system of 3 equations with 3 unknowns. The system will be solved using the iterative
Newtons method for systems of equations, which is also presented below.

After the trim problem has been motivated and a method for solving it has been outlined,
the computer implementation of the problem is thoroughly discussed; including the
interactions of the Aero-F program with the trim program, the user interactions with the
input files and the output of the programs. Some of the nice features of the trim program
are as follows: it allows the user to specify a rigid or an elastic aircraft, it allows the user
to choose whether or not to compute the derivatives at every iteration, and the user can
specify how many unknowns to solve for. After the computer implementation is
presented, the results are discussed; showing clearly that both rigid and aeroelastic
computations yield similar results whether or not the derivative is recomputed.
3

ABSTRACT.................................................................................................................................................. 2
MODEL OF THE PHYSICAL PROBLEM.............................................................................................. 4
INTRODUCTION........................................................................................................................................... 4
COORDINATE SYSTEM OF THE AERODYNAMIC FORCES AND MOMENTS: USED IN AERO-F SOFTWARE ...... 4
LIFT-DRAG COORDINATE SYSTEM: USED IN TRIM SOFTWARE................................................................... 7
RIGID VS. ELASTIC AND AEROELASTICITY............................................................................................... 13
MATHEMATICAL METHOD................................................................................................................. 14
NEWTONS METHOD FOR SYSTEMS OF EQUATIONS.................................................................................. 14
COMPUTER IMPLEMENTATION........................................................................................................ 16
INTRODUCTION TO AERO-F, AERO-S AND TRIM SOFTWARE .................................................................... 17
BASIC FUNCTIONALITY FOR THE TRIM SOFTWARE COMPONENTS............................................................ 18
PREPARING THE TRIM INPUT FILE AND RUNNING THE TRIM EXECUTABLES............................................. 18
1 Unknown, 2 Unknowns or 3 Unknowns ........................................................................................... 19
Computing and Updating the Derivatives........................................................................................... 19
Rigid or Aeroelastic ............................................................................................................................ 20
Aero-F and Aero-S Input Files.................................................................................................................... 20
RIGID AND AEROELASTIC RESULTS FOR 3X3, 2X2 AND 1X1 SYSTEMS ................................ 20
COMPUTING DERIVATIVES AT FIRST ITERATION ONLY ............................................................................ 20
UPDATING THE DERIVATIVES AT EACH ITERATION.................................................................................. 22
CONCLUSIONS AND FUTURE IMPROVEMENTS............................................................................ 23
REFERENCES........................................................................................................................................... 25
APPENDIX A: PSEUDO CODE FOR NEWTONS METHOD FOR SYSTEMS OF EQUATIONS...... 26

4
Model of the Physical Problem
When solving problems in practice, the first step is to develop a model of the actual
physical circumstances, while paying attention to the parameters of interest and making
assumptions for simplification when possible. The problem of interest here is trimming
an aircraft in the longitudinal sense, using the thrust, angle of attack and control surface
deflection angle with a given flight condition.

Introduction
The focus of this section is the 2D trim model for the aircraft, which is climbing at a
constant translational velocity, i.e. it is not accelerating. Given the aircraft geometry and
flight conditions, the aerodynamic forces and moments acting on the aircraft are currently
computed by the Aero-F computer program (Ref. [2], Ref. [3] & Ref. [4]).

The 2-dimensional simplification; restricting the motion of the aircraft to longitudinal
motions only; is made possible by assuming the sideslip is zero, the aircraft is symmetric
and the aircraft is climbing at an angle, . Also, it is important to note that the output of
the Aero-F program is defined in a reference frame where the origin is not at the center of
gravity; thus the Aero-F reference frame is clearly defined below, in order for use in the
development of the trim model. Then the equations for the longitudinal forces and
moments on the aircraft are developed, resulting in a 3x3 system of equations.

Coordinate System of the Aerodynamic Forces and Moments: Used in Aero-F
Software
Given an aircrafts geometry, the Aero-F program computes the aerodynamic forces and
moments with respect to the mesh center, which will now be referred to as O. The mesh
is created according to the geometry of the aircraft and O is a parameter defined by the
user. The axis system used in the Aero-F program is a body fixed system with the origin
at O, the
b
X axis is along the fuselage in the direction of the tail, the
b
Y axis is along the
wings in the direction of the right wing and the
b
Z body axis is in the direction of the top
of the fuselage, see Figure 1 where the
b
Z body axis is coming out of the page.

5

Figure 1: Axes System for Aero-F Program

The 2-dimensional free-body diagram of the aerodynamic forces for the aircraft, at a
given angle of attack, , is shown in Figure 2. The aerodynamic forces in the body axis
are referred to as
b
X
F ,
b
Y
F ,
b
Z
F and the aerodynamic moments in the body axis are
referred to as
b
X O
M
,
,
b
Y O
M
,
,
b
Z O
M
,
, (i.e. the Aero-F output). However, it is easier to
work in another axis system; that is the Lift-Drag axis system, so, some rearrangement is
necessary. Also, the 2-dimensional simplification resulting in longitudinal motions only,
requires the knowledge and use of
b
X
F ,
b
Z
F , and
b
Y O
M
,
, the others will be considered at
a later date.

First, the origin of the Lift-Drag Axis system is at the aircrafts center of gravity and will
be referred to as CG. The axes in the Lift-Drag system are X , Y , and Z ; see Figure 2.
The X axis is in the direction of the aircrafts velocity vector and in the opposite
direction of the aircrafts drag vector; the Y axis is in the direction of the aircrafts lift
vector. Although the center of the Lift-Drag system does NOT coincide with that of the
program output, this does NOT affect the resultant forces in each direction, it only affects
the moments. Thus, the lift, L , and the drag, D, are found in terms of the Aero-F forces,
as given in equations (1) and (2) respectively.

6

Figure 2: Aerodynamic Forces in Body Axes and in Lift-Drag Axes

( ) ( ) sin cos
b b
X Z
F F L = (1)

( ) ( ) cos sin
b b
X Z
F F D + = (2)

The Aero-F forces are written in terms of the lift and drag in equations (3-4).

( ) ( ) sin cos L D F
b
X
= (3)

( ) ( ) cos sin L D F
b
Z
+ = (4)

In general the origin, O, is not at the same location as the center of gravity, i.e. the Aero-
F origin does not coincide with Lift-Drag origin. Since the center of gravity is the point
of interest, the moment about the Z axis in the Lift-Drag frame, must be computed at the
center of gravity, CG. This moment computation can be accomplished with a few
important parameters; given the total moment from the Aero-F program,
O
M , the
resultant force vector, F , (written in the body coordinate frame) and the position vector
from CG to O, r , (written in the body coordinate frame), see Figure 3 and equation (5).

7

Figure 3: Calculating the Moment in the Lift-Drag System

F r M M
O CG
+ = (5)

Equation (5) relates the total moment at O to the total moment at CG, but we are only
interested in the moment about the Z axis at CG. So, if the vector from the location of
the CG to O, written in the body coordinate system, is ( )
cg cg cg
z y x , , , then
substituting equation (6) into equation (5) yields equation (7).


b Z b Y b X
z F y F x F F
b b b
+ + = (6)

b b b b
Z cg X cg Y O Y CG Z CG
F x F z M M M + = =
, , ,
(7)

From now on, we will drop the subscripts of M in equation (7) and for simplicity denote
Z CG
M
,
by M.

Lift-Drag Coordinate System: Used in Trim Software
From the Aero-F program, the aerodynamic forces and moments have been computed,
but the forces of gravity and thrust are also acting on the aircraft and they must be
included in order to develop a complete trim model. The thrust force of the aircraft is
tilted at a fixed angle
T
from the
b
X axis, as shown in Figure 4.

8

Figure 4: Thrust Force with respect to the Lift-Drag Axes System

The previous section enables the lift and drag to be computed as a function of the angle
of attack, . However, the aircrafts position with respect to the Earth has not been
formulated when using angle of attack. Thus, the climb angle, , measured from the
horizontal axis,
H
X , is now introduced, see Figures 5-6.


Figure 5: Aircraft in Steady Straight-line Climb

The climb angle of the aircraft is important because the weight is tilted from the Y axis
due to the climb angle. So, taking into account the aerodynamic forces, the thrust and the
weight; the resultant forces acting on the aircraft, in the Lift-Drag axes system, are given
in equations (8-9).
9


Figure 6: Forces in the Body Axes, the Lift-Drag Axes and the Horizontal-Vertical Axes

( ) ( ) D W T F
T X
+ = sin cos (8)

( ) ( ) L W T F
T Y
+ + = cos sin (9)

The restriction to longitudinal motion and the choice of coordinate systems turns out to
be very convenient because the
b
Y axis and the Z axis are always in the same direction.
So, the moment, M , which has been computed at the CG, is already about the Z axis.
Also, since the origin of the Lift-Drag coordinate frame is the CG, the weight will not
create any moments. If the thrust acts at a location, ( )
T T T
z y x , , , written in the Lift-Drag
system with respect to the CG, then the total moment about the Z axis is given by
equation (10).

( ) ( ) M y T x T M
T T T T Z
+ + + = cos sin (10)

The description of the control surface deflections is the only detail that has been omitted
thus far. The lift, L , the drag, D and the moment, M , are all dependent on control
surface deflections. Control surfaces include: ailerons, rudders, flaps, elevators, slats and
spoilers. However, only the flaps and the elevators are considered to have an effect on
the longitudinal motion of the aircraft, for this model. The effects of the flaps and the
10
elevators are very similar, so they are combined into one variable denoted by, . So, lift,
drag and moment are functions of ; but they are also functions of the angle on attack,
, equation (11).

( ) , L , ( ) , D , ( ) , M (11)

Theoretically, the lift and the moment are linear functions of the angle of attack except in
the stall regime. Since trimming the aircraft will never fall in the stall regime, it is safe to
assume a linear relationship for these quantities. The lift versus the angle of attack,
computed using the Aero-F program, is shown in Figure 7 to be essentially linear for both
rigid and elastic aircraft, just as expected. Also, the moment versus the angle of attack,
computed using the Aero-F program, is shown in Figure 8 and again is essentially linear
for both rigid and elastic aircraft. On the other hand, theory indicates that the drag is a
parabolic function with respect to the angle of attack, which can be seen in the results of
the Aero-F program presented in Figure 9.

Lift vs. Angle of Attack
Control Surface Deflection = 0
3000
4000
5000
6000
7000
8000
9000
2.5 3.5 4.5 5.5 6.5 7.5 8.5
Angle of Attack (degrees)
L
i
f
t
Aeroelastic Rigid

Figure 7: Lift vs. Angle of Attack for both Rigid and Elastic Aircraft

11
Moment vs. Angle of Attack
Control Surface Deflection = 0
-500
0
500
1000
1500
2000
2500
3000
3500
4000
2.5 3.5 4.5 5.5 6.5 7.5 8.5
Angle of Attack (degrees)
M
o
m
e
n
t
Aeroelastic Rigid

Figure 8: Moment vs. Angle of Attack for both Rigid and Elastic Aircraft

Drag vs. Angle of Attack
Control Surface Deflection = 0
500
700
900
1100
1300
1500
1700
1900
2.5 3.5 4.5 5.5 6.5 7.5 8.5
Angle of Attack (degrees)
D
r
a
g
Aeroelastic Rigid

Figure 9: Drag vs. Angle of Attack for both Rigid and Elastic Aircraft

The variation of the lift, moment and drag with respect to the control surface deflection
are shown in Figures 10, 11, and 12 respectively. According to Figure 10, the lift is a
linear function of the control surface deflection and according to Figure 11, the moment
is also a linear function of the control surface deflection. However, Figure 12 shows that
the drag is NOT a linear function of the control surface deflection.

12
Lift vs. Control Surface Deflection Factor
Alpha = 3.8
o
3950
4050
4150
4250
4350
4450
4550
4650
4750
-1 -0.5 0 0.5 1
Control Surface Deflection Factor
L
i
f
t
Aeroelastic Rigid

Figure 10: Lift vs. Control Surface Deflection Angle for both Rigid and Elastic Aircraft

Moment vs. Control Surface Deflection Factor
Alpha = 3.8
o
700
1700
2700
3700
4700
5700
-1 -0.5 0 0.5 1
Control Surface Deflection Factor
M
o
m
e
n
t
Aeroelastic Rigid

Figure 11: Moment vs. Control Surface Deflection Angle for both Rigid and Elastic
Aircraft

13
Drag vs. Control Surface Deflection Factor
Alpha = 3.8
o
705
715
725
735
745
755
765
-1 -0.5 0 0.5 1
Control Surface Deflection Factor
D
r
a
g
Aeroelastic Rigid

Figure 12: Drag vs. Control Surface Deflection for both Rigid and Elastic Aircraft

Rigid Vs. Elastic and Aeroelasticity
When a material deforms under a load, as long as the deformation is not permanent, the
material is said to be elastic. On the other hand, if the material does not deform under
any load, then it is called rigid. Material properties of an object determine the elasticity
of that object. The materials used for aircraft are generally elastic, thus they exhibit
elastic deformation when subjected to a load. However, sometimes the load on an
aircraft is so small that the elastic behavior can be neglected and a rigid assumption is
suitable. Small load conditions occur at low mach numbers and low angles of attack,
where there is very little if any turbulence or flow separation.

The rigid assumption simplifies computation of the aerodynamic forces and moments
because the interactions of the fluid with the deforming geometry (a.k.a. aeroelasticity)
are not included. In other words, the geometry does not deform so the fluid does not
have to adjust to a deformed geometry. The rigid assumption is also useful to get a good
idea of the flow characteristics before the elastic effects have been included.

The lift, drag and moment on the aircraft depend on whether the aircraft is assumed to be
rigid or elastic. When the aircraft is elastic, it deforms and then the fluid must adjust to
the new geometry, thus resulting in different values for the lift, drag and moment; i.e. the
lift, drag and moment will be different assuming the aircraft is rigid, than they will be
assuming the aircraft is elastic. The concept of an elastic structure interacting with a
fluid, when the fluid and the structure share a physical boundary, is called aeroelasticity.
Empirical data of aircraft shows that aeroelasticity is indeed an observed phenomenon,
especially for large or high-speed aircraft. When the original geometry of the aircraft is
altered, it is clear that the aerodynamic forces and moments will also change, so in many
cases it is important to account for aeroelasticity but sometimes a rigid assumption is
permissible.
14

Mathematical Method
The next step, after modeling a problem, is determining a suitable mathematical method
to solve the problem. Trimming an aircraft is essentially finding the equilibrium point for
the forces and moments on an aircraft, which can be formulated as a root finding
problem; thus, techniques used to find roots will apply to the trim problem. However, it
is important to note that equation (8) is nonlinear with respect to the angle of attack, so a
nonlinear root finding method would be suitable; Newtons method for finding roots is
just that and it is the chosen method for this problem.

Newtons Method for Systems of Equations
Newtons method for systems of equations is just an extension of the method for
functions of one variable (Ref. [1]). The whole idea is based on a first order (linear)
approximation using Taylor series expansions; for example, if ( ) P is the function, and
0
is the initial approximation, then the first order Taylor approximation is:

( ) ( ) ( )
( )


d
dP
P P
0
0 0
+ (12)

Assuming equation (12) is equal to zero; it should be close to zero if we supplied a decent
initial approximation; then it is easy to solve for , equation (13).

( )
( ) [ ]


d dP
P
0
0
0
= (13)

If equation (13) does not yield correct results for , when we put it back into equation
(12), then we can repeat the process with our new , in other words we can iterate until
the solution converges. Iterating equation (13) is nicely written in mathematical form
below:

( )
( ) [ ]


d dP
P
k
k
k k
=
+1
(14)

If the solution does not converge but rather diverges after several iterations, then it is
necessary to supply a better initial approximation or in extreme cases it might be
necessary to use a different approximation method; maybe a second or third order
method. However, for the scope of this document, it is assumed that this method will
converge or that we will find a better initial approximation. The generalized form of
Newtons method is given in equation (15), where J, is the Jacobian matrix for system of
equations, P; for further explanation of Newtons method for systems consult Ref. [1].

15
( ) [ ] ( ) [ ]
k k k k k k
k
k
k
k
k
k
T P T a J
T T
, , , ,
1
1
1
1

+
+
+

(15)

Now, it is time to combine the aspects of the document to finish our formulation of the
trim model. Referring to equations (8-10), if the thrust, T , along with and are the
variables, then there are three equations with three unknowns. The concept of trimming
an aircraft requires equilibrium of the forces and moments that act on the aircraft about
the CG, i.e. it requires setting equations (8-10) equal to zero.

( ) ( ) 0 sin cos = + D W T
T
(16)

( ) ( ) 0 cos sin = + + L W T
T
(17)

( ) ( ) 0 cos sin = + + + M y T x T
T T T T
(18)

Now, we can write equations (16-18) in vector form and call it ( ) T P , , , equation (19).

( )
( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( )

+ + +
+ +
+
=

=




, cos sin
, cos sin
, sin cos
, ,
3
2
1
M y T x T
L W T
D W T
P
P
P
T P
T T T T
T
T
(19)


As we know from calculus, the Jacobian Matrix for a system of n equations with n
unknowns often replaces a simple derivative in one variable. The Jacobian Matrix for
equation (18) is given in equation (20).

( )
( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( )

=
T
T P T P T P
T
T P T P T P
T
T P T P T P
T J
, , , , , ,
, , , , , ,
, , , , , ,
, ,
3 3 3
2 2 2
1 1 1


(20)

It is important to note that Newtons method for systems gives quadratic convergence to
the solution of equation (19), assuming that equation (20) is computed exactly at each
iteration, and that it is NOT singular at any iteration. To compute the Jacobian, for a
particular iteration, simply add the iteration counter k to the variables in equations (19-
20). Then, substituting equation (19) into equation (20) results in the following matrix:

16
( )
( )
( ) ( )
( )
( )
( ) ( )
( )
( ) ( ) ( )
( ) ( )
( ) ( )

+ +

+ + + +
+

+ +
+

+
=
T k T T k T
k k
T k T T k T
k T
k k k k
k T
k T
k k k k
k T
k k k
y x
M M
y x T
L L
T
D D
T
T J




cos sin
, ,
sin cos
sin
, ,
cos
cos
, ,
sin

, ,


The partial derivatives above can be approximated using finite difference formulae, see
equations (21-22), and the Aero-F program. Setting thrust and as constants, call the
program with a particular and store the results; then call the program once again with
the same thrust and as previously, but with a different and store the results; then
call the program yet again with the same thrust and the same as originally, but with a
different and store the results. The finite difference formulae below are forward
differences; the user must call the Aero-F program at least three times to produce the
values below. Using the central difference formulae, (Ref. [1]), the user would have to
call the Aero-F program more than three times. However, the order of accuracy for
central differences is higher of than that of forward differences and thus can improve the
derivative approximations.

1 2
1 2

D D D
,
1 2
1 2

L L L
,
1 2
1 2

M M M
(21)

1 2
1 2

D D D
,
1 2
1 2

L L L
,
1 2
1 2

M M M
(22)

Since the lift, drag and moment are affected by elasticity, the derivatives of lift, drag and
moment will also be affected by elasticity. Thus, the derivatives will be different under
the rigid assumption than they will be under the elastic assumption. The derivatives are
essential to Newtons method and if they change so will the results of Newtons method.

Appendix A depicts the pseudo code for Newtons method for systems of equations.

Computer Implementation
The physical problem has been thoroughly defined, the mathematical method to solve the
problem has also been explained in detail and now it is time to actually solve the
problem. Solving the complex problems of the modern times generally involves
computer software of some type due the large number of computations that must be
accomplished in a short period of time: the trimming problem, when performed on
meshes with a large number of nodes, is one of the complex problems that will need a
computer to solve. Thus, this section of the document explains the computer software
and how the user interacts with it.

17
Introduction to Aero-F, Aero-S and Trim Software
The implementation of this trimming procedure involves more than simply Newtons
method, which was described in the previous section. The aerodynamic forces and
moments for an aircraft must be computed before trimming can be addressed; this is
accomplished with a piece of software called Aero-F. Then the trim software gets the
output of Aero-F and from it determines the lift, drag and moment in order to proceed
with Newtons method. Since Newtons method is an iterative process, Aero-F must
supply the corresponding forces and moments to the trim software at each iteration.

Aero-F is responsible for reading aircraft geometry data to create a mesh and reading the
fluid properties in order to determine the aerodynamic forces and moments that are acting
on the aircraft under certain flight conditions. The aerodynamic forces and moments are
computed with an iterative scheme in time that is described in Reference [2]; this
iteration will be referred to as the time iteration of the fluid.

Despite the wonderful capabilities of Aero-F, it is a fluid program and does not know
about the structural properties of the aircraft. So, there is another piece of software that is
used to account for the structural properties and structural dynamics called Aero-S, see
Reference [3]. Aero-S computes structural behavior using the finite element method, for
a particular mesh that is associated with certain material properties. Aero-F is designed
to communicate with Aero-S given the correct command from the Aero-F user input file
and the correct command from the Aero-S user input file, this is described in more detail
below.

The communication between the fluid and the structure can be done without the user
interacting at each time iteration of the fluid, or the data can be feed back and forth by the
user. Thus, aeroelastic computations can be accomplished when the communication is
automatic between the fluid and the structure. Also, aerodynamic computations around a
deformed rigid, structure, can be accomplished when the user feeds the data back and
forth between the fluid and the structure. There are other computations that can be done
with Aero-F and Aero-S but for trimming we are concerned with aeroelastic problems
and aerodynamic problems with a deformed rigid structure.

The Aero-S software currently does not allow for the user to specify an angle of
deflection for the control surfaces. Instead, the software is equipped to handle a moment
applied to the control surface that causes a deflection of that surface. When the original
moment; i.e. original means the surface is not opened; is multiplied by a factor, the
surface will open creating the new aircraft geometry. The applied moment on the control
surface is found in the Aero-S input file. The trim input file contains the factor which
will be multiplied by the original applied moment in the Aero-S input file when trimming
the aircraft. The factor on the moment is denoted by in all the equations above. The
relationship between the factor and the control surface deflection angle, , is given in
equation (23), where
0
N is the applied moment vector, and k is the stiffness coefficient
for the material. Equation (23) has reduced the stiffness of an element from a matrix to a
scalar equation due to the assumption that the material is isotropic. Equation (23) applies
18
only in the coordinate system of the aerodynamic forces and moments (described above),
if the angle is desired in another coordinate system then equation (23) must be modified.


k
2
0
N
= (23)

The trim software controls most of the interactions between Aero-S and Aero-F based on
the data and flags that are set in the trim input file. However, the input files for Aero-S
and Aero-F must be set up correctly for several reasons as well; in order that the trim
software be able to manipulate the template to create an input file and in order that the
input file include everything necessary for the computations. So, for user convenience
and necessity, there are templates for the Aero-S and Aero-F input files that are required
to be able to run the trim software. The input files are the focus of this discussion
because they are the primary means for the user to interact with the different pieces of
software.

Basic Functionality for the Trim Software Components
The trim software is composed of several smaller pieces of software in an attempt to
make it as modular as possible. The main code that runs the aeroelastic case for
trimming is called trim_ae, while that of the rigid case is called trim_rgd; either trim_ae
or trim_rgd is envoked when either e or r is provided respectively in the execution
command. The Newton solver for the trim software is called trimNew1, trimNew2, or
trimNew3 depending on how many unknowns are being solved for in the iteration.

The trim software automatically creates the fluid input files, using their corresponding
templates, with the procedure trim_alpha; this procedure searches for ALPHA and
replaces it with a floating point number for the angle of attack in degrees. The structure
input files are automatically transformed, from their corresponding templates, by
trim_delta. The procedure, trim_delta, searches for Delta(float original applied moment)
and replaces it with a floating point number equal to multiplied by the float original
applied moment; thus creating a moment on the control surface to open it. Also used to
generate the structure input files from the templates, under the rigid assumption only, are
fem_pp_input and fem_def_input. The structure input that creates a deformed mesh is
created by fem_def_input, and the structure input that sends the deformed mesh to the
fluid is created by fem_pp_input.


Preparing the Trim Input File and Running the Trim Executables
The trim input file is very short and simple; it controls the basic initial input for the trim
equations and some of the functionality of the trim software. The angles of attack used to
compute the derivatives in equation (21) are designated by a1 and a2 in the input file.
The control surface deflection factors for computing the derivatives in equation (22) are
designated by d1 and d2 in the input file. The initial solution guess used for Newtons
method is a1, d1, and t1, where t1 is the initial thrust approximation. The climb angle is
called gamma; the thrust incidence angle, in the aerodynamic forces and moments
coordinate system, is phiT; the position of the thrust, in the aerodynamic forces and
19
moments coordinate system, is posT; the position of the CG, in the aerodynamic forces
and moments reference frame, is posCG; and the weight of the aircraft is weight. The
maximum number of iterations allowed for Newtons method is controlled by N in the
input file. The Newton iteration can end either because the maximum number of
iterations has been reached or because the residual is small enough in comparison to the
initial residual, which is controlled by TOL in the input file, see Appendix A for actual
comparison technique.

1 Unknown, 2 Unknowns or 3 Unknowns
There are generally 3 equations, see equation (19), with three unknowns for the 2D
trimming of an aircraft. However, if one of the unknowns is set to a constant value, then
the 3x3 system of equations reduces to 2x2 system of equations. The trim procedures are
equipped to solve a 2x2 system; i.e. when the thrust is constant, then it is possible to
solve for the angle of attack and control surface deflection angle using the first two
equations in (19). Currently the 2x2 trim procedures can only eliminate the thrust, this is
done by fixing it and then solving for the corresponding angle of attack and control
surface deflection angle.

One might ask what about solving just one equation with two of the variables fixed.
Well, the trim procedures are also set-up to solve the first equation in (19) for the angle of
attack, given the thrust and control surface deflection angle are constants. The single
equation trim procedures currently allow for solving the first equation to determine the
angle of attack while the other variables are constant.

The number of unknowns and equations that will be solved is determined by the
executable that is used; i.e. the trim_3 executable solves the 3x3 system for all three
unknowns, the trim_2 executable solves the 2x2 system for the angle of attack and
control surface deflection angle, and finally the trim_1 executable solves the single
equation for the angle of attack.

Computing and Updating the Derivatives
The derivatives in equations (21-22) must be computed prior to Newtons method. After
the derivatives have been computed, they can remain constant for Newtons method or
they can be recomputed using the new solution at each iteration. If the initial solution
vector is significantly different then the solution vector at any iteration, then the non-
linearity of the problem could introduce a poor derivative into the Newton computations,
in which case it would be highly beneficial to recompute the derivative at that iteration.
The trim procedures allow the user to choose to compute the derivatives at each iteration
or to compute the derivatives only at the beginning and leave them constant for the rest of
the iteration. This is accomplished with a simple flag in the trim input file, where 1
indicates an updated derivative and anything else indicates a constant derivative. The
user can specify whether or not to recompute the derivatives using flagDer in the input
file; setting flagDer to 1 will recomputed derivatives at every iteration of Newtons
method, while setting it to anything else will compute the derivative only initially.
20

Rigid or Aeroelastic
As mentioned above, the rigid assumption might be practical for some purposes in order
to avoid the complications of an aeroelastic computation. Thus, the trim procedures are
designed to run for either a rigid or aeroelastic aircraft. The user must supply a flag when
running the executable to indicate rigid, ( -r), or aeroelastic, ( -e): for example, to run the
3x3 system for a rigid aircraft the user would execute with the following command at the
terminal: ./trim_3 r.

Besides the trim procedures, the Aero-F and Aero-S software must know whether it is an
aeroelastic or a rigid aircraft: this is accomplished with the correct commands in the fluid
input and structure input files. In order to avoid complication, templates are used for the
fluid input and the structure input files and they are provided with the trim software.

Aero-F and Aero-S Input Files
Generally the executables for Aero-S and Aero-F have already been generated and the
only user updating that is necessary belongs in the input files of Aero-S, (see Ref. [2]),
and/or Aero-F, (see Ref. [3]). However, the executables for the fluid and the structure are
not in the scope of this document, so if they need to be created the user must refer to
References [2] and [3].

Rigid and Aeroelastic Results for 3x3, 2x2 and 1x1 Systems
The trim software has been tested for rigid and aeroelastic scenarios with the 3x3, 2x2
and 1x1 systems, for cases of either constant derivatives or updating derivatives; all the
tests done to date are for the Langley Fighter. The rigid tests as compared with the
aeroelastic tests show very similar results as would be expected since the angle of attack
is fairly small and the Mach number is also fairly small, 0.3. The results also indicate
that recomputing the derivatives does not improve the outcome significantly and that the
cost of computation is not very much larger than for constant derivatives. However, it is
very possible that the solution could be improved by recomputing the derivatives in the
case that the initial approximation is very poor and the nonlinearity of the drag becomes a
non-negligible factor for the derivatives.

Computing Derivatives at First Iteration Only
Given the same initial input, the rigid and aeroelastic scenarios should only differ by a
small amount relative to one another; this can be seen in Figure 13, where the magnitude
of the residual is very nearly the same at every iteration.

21
Magnitude of Residual for 3x3 System
Initially [
,

, T ] = [2.78483,-0.8,3500]
Recomputing the Derivatives
Aero (4, 8.49719)
Aero (3, 62.518)
Aero (2, 618.595)
Aero (1, 9347.44)
Rigid (1, 9351.85)
Rigid (2, 602.199)
Rigid (3, 65.7536) Rigid (4, 8.72147)
-500
500
1500
2500
3500
4500
5500
6500
7500
8500
9500
1 2 3 4
Iteration of Newton's method
M
a
g
n
i
t
u
d
e

o
f

R
e
s
i
d
u
a
Aeroelastic Rigid

Figure 13: Residual with Constant Derivatives for 3x3 System

Referring to Tables 1-3, the number of iterations to get a similar solution for the rigid
case is the same as for the aeroelastic case except for 2x2 the aeroelastic case iterates one
time less than the rigid case. Since these results are almost the same for all possible ways
of running the trim software with the derivatives constant, it is safe to say that the rigid
and aeroelastic results are essentially the same with constant derivatives for 3x3, 2x2 and
1x1 systems.

Table 1: Trimming Aircraft for solution of , with , T, and Derivatives Constant
# Iterations Alpha Solution Magnitude of Residual
Case 1 Rigid 2 8.18847 0.318231
Aeroelastic 2 8.17704 0.275784
Case 2 Rigid 2 8014057 0.0957837
Aeroelastic 2 8.14055 0.115314
Case 3 Rigid 3 8.04383 0.829625
Aeroelastic 3 8.03338 0.800235
Case 4 Rigid 4 8.08071 0.414528
Aeroelastic 4 8.06982 0.355878
Constant Derivatives


Where the initial input to Newtons method, in Table 1, changes from case to case as
follows:
Case 1: Initial a = 7.78483o, d = 0.0872625, T = 1968.93 N
Case 2: Initial a = 7.78483o, d = 0.174525, T = 1968.93 N
Case 3: Initial a = 5.78483o, d = 0.261788, T = 2500 N
Case 4: Initial a = 1.78483o, d = 0. 174525, T = 2500 N
22








Table 2: Trimming Aircraft for solution of [, ] with T and Derivatives Constant
# Iterations Solution Vector Magnitude of Residual
Case 1 Rigid 3 [8.48022, -0.609806] 0.436681
Aeroelastic 3 [8.47062, -0.616328] 0.311359
Case 2 Rigid 4 [8.4769, -0.602007] 0.168832
Aeroelastic 4 [8.46793, -0.609978] 0.178997
Case 3 Rigid 4 [8.47612, -0.600078] 0.305162
Aeroelastic 3 [8.46649, -0.602899] 0.111275
Constant Derivatives


Where the initial input to Newtons method, in Table 2, changes from case to case as
follows:
Case 1: Initial a = 7.78483o, d = 0.0872625, T = 1968.93 N
Case 2: Initial a = 7.78483o, d = 0.174525, T = 1968.93 N
Case 3: Initial a = 5.78483o, d = 0.261788, T = 2500 N

Table 3: Trimming Aircraft for solution of [, , T] with Derivatives Constant
# Iterations Solution Vector Magnitude of Residual
Case 1 Rigid 3 [8.29348, -0.158712, 1968.89] 0.0973955
Aeroelastic 3 [8.28496, -0.166299, 1969.05] 0.599159
Case 2 Rigid 3 [8.29341, -0.158676, 1968.82] 0.141128
Aeroelastic 3 [8.28494, -0.166205, 1969.01] 0.247255
Case 3 Rigid 3 [8.29345, -0.158769, 1969.25] 0.389392
Aeroelastic 3 [8.28488, -0.166447, 1969.38] 0.894145
Case 4 Rigid 4 [8.30049, -0.161054, 1975.64] 8.72147
Aeroelastic 4 [8.29157, -0.168799, 1975.74] 8.49719
Constant Derivatives


Where the initial input to Newtons method, in Table 3, changes from case to case as
follows:
Case 1: Initial a = 7.78483o, d = -0.165699, T = 1968.93 N
Case 2: Initial a = 7.78483o, d = 0.1, T = 2500 N
Case 3: Initial a = 5.78483o, d = 0.5, T = 1000 N
Case 4: Initial a = 2.78483o, d = -0.8, T = 3500 N

Updating the Derivatives at Each Iteration
Referring to Tables 4-5, the number of iterations to get a similar solution for the rigid
case is very close to that for the aeroelastic case; the slight difference in the number of
iterations is not unexpected and very often can be attributed to a poor derivative
computation at some iterate of Newtons method. Since the results for the solution are
almost the same in Tables 4-5, with the updated derivatives, it is safe to say that the rigid
23
and aeroelastic results are essentially the same with updating derivatives, but sometimes
the cost of computation is slightly different.

Tables 1-5 are good evidence that the trim software works well for 3x3, 2x2 and 1x1
systems, with constant or updating derivatives. It can also be inferred from the tables that
the trim software handles the rigid and elastic assumptions similarly, and the results are
approximately the same.

Table 4: Solving for , with , and T constant, Updating Derivatives
# Iterations Alpha Solution Magnitude of Residual
Case 1 Rigid 2 8.18847 0.318231
Aeroelastic 2 8.17721 0.285886
Case 2 Rigid 2 8.15163 0.175986
Aeroelastic 2 8.14055 0.115314
Case 3 Rigid 3 8.04458 0.0195577
Aeroelastic 5 8.03379 0.0227707
Case 4 Rigid 4 8.0811 0.00296999
Aeroelastic 3 8.06934 0.949117
Updating Derivatives


Where the initial input to Newtons method, in Table 4, changes from case to case as
follows:
Case 1: Initial a = 7.78483o, d = 0.0872625, T = 1968.93 N
Case 2: Initial a = 7.78483o, d = 0.174525, T = 1968.93 N
Case 3: Initial a = 5.78483o, d = 0.261788, T = 2500 N
Case 4: Initial a = 1.78483o, d = 0. 174525, T = 2500 N

Table 5: Trimming Aircraft for solution of [, , T] with Updating Derivatives
# Iterations Solution Vector Magnitude of Residual
Case 1 Rigid 3 [8.29348, -0.158712, 1968.89] 0.0973955
Aeroelastic 3 [8.28483, -0.165962, 1968.93] 0.147438
Case 2 Rigid 3 [8.29341, -0.158676, 1968.82] 0.141128
Aeroelastic 3 [8.28486, -0.16607, 1968.93] 0.06118
Case 3 Rigid 3 [8.29344, -0.158706, 1968.88] 0.0305748
Aeroelastic 3 [8.28478, -0.165879, 1968.91] 0.35005
Case 4 Rigid 3 [8.29559, -0.158363, 1968.12] 3.99452
Aeroelastic 5 [8.28504, -0.166138, 1968.96] 0.327808
Updating Derivatives


Where the initial input to Newtons method, in Table 5, changes from case to case as
follows:
Case 1: Initial a = 7.78483o, d = -0.165699, T = 1968.93 N
Case 2: Initial a = 7.78483o, d = 0.1, T = 2500 N
Case 3: Initial a = 5.78483o, d = 0.5, T = 1000 N
Case 4: Initial a = 2.78483o, d = -0.8, T = 3500 N

Conclusions and Future Improvements
The trim program works well for both aeroelastic and rigid aircraft provided the initial
angle of attack is not in the stall regime, the initial thrust is in a suitable range for that
24
particular aircraft, and the deflection angle is not to large to cause flow separation or
mesh shearing; i.e. negative flow proerties. Tables 1-5 are good evidence that the trim
software works well for 3x3, 2x2 and 1x1 systems, with constant or updating derivatives.
It can also be inferred from the results that the trim software handles the rigid and elastic
assumptions similarly, and yields approximately the same solution.

The most obvious course of improvement for the trim code would be to develop and test
the trim procedures on several other aircraft meshes, besides the Langley Fighter, to
determine the applicability of the trim software. Also, the model of the physical problem
would be much more accurate if it was reformulated from longitudinal motions only, to
longitudinal and lateral-directions motions combined, i.e. 2D to 3D model. Also, a
correct model of the problem should account for the aeroelastic effects on the CG
location, thrust location and thrust incidence angle.

The mass characteristics and thrust characteristics of the aircraft are not included in the
Aero-F output that is used for the trim procedures. Thus, for the purposes of a complete
trim model, these characteristics and the resulting equations will be included in the Aero-
F output in the near future. Therefore, adding the capability to communicate the weight
and thrust properties of the aircraft from the input and output of Aero-F to the trim code
would improve the estimates used for weight and thrust.

The software could be much more useful if a method to detect when Aero-F encounters
errors could be added; thus avoiding erroneous results from Aero-F and preventing an
incorrect computation of the lift, drag or moment. Avoiding an erroneous computation
for the lift, drag and moment, in turn prevents the derivatives from being computed
incorrectly and allows Newtons method to proceed without diverging. Determining the
appropriate and allowable ranges for the deflection angle in order to avoid mesh shearing
and negative numerical flow properties (like density and pressure), would also help avoid
erroneous computations.

Currently, the number of unknowns and equations that will be solved is determined by
the executable that is used; i.e. the trim_3 executable solves the 3x3 system for all three
unknowns, the trim_2 executable solves the 2x2 system for the angle of attack and
control surface deflection angle, and finally the trim_1 executable solves the single
equation for the angle of attack. The 2x2 trim procedures must eliminate the thrust, and
the 1x1 procedures must eliminate the thrust and surface deflection. However, it would
be very beneficial if the trim code would allow for any of the variables to be fixed and
then solve for the remaining variables without a different executable; it is feasible to have
one executable that can solve for 1, 2 or 3 unknowns based on a flag from the user input
file, this is another upgrade to be addressed in the future.




25
References

[1] Burden, Richard L. and J. Douglas Faires. Numerical Analysis, 6
th
ed. Brooks/Cole
Publishing Company; Pacific Grove, CA: 1997.

[2] Farhat, Charbel and Bruno Koobus. Aero-F2D/F3D: AUsers Manual. University of
Colorado, Center for Aerospace Structures; Boulder, Colorado: 2003.

[3] Farhat, Charbel. RCfem and FEM. University of Colorado, Center for Aerospace
Structures; Boulder, Colorado: 2003.

[4] Geuzaine, Philippe. Aero-F Manual: The documentation for Aero-F, Version 1.0, ed
0.1. University of Colorado, Center for Aerospace Structures; Boulder, Colorado:
2003.

[5] Roskam, Jan. Airplane Flight Dynamics and Automatic Flight Controls: Part I.
Design, Analysis and Research Corporation; Lawrence, KS: 1998.






26
Appendix A: Pseudo code for Newtons method for systems of equations

Newtons method for Systems of Equations

To approximate the solution of the nonlinear system P(x) = P(,,T ) = 0 given an initial
approximation x
0
= (
0
,
0
,T
0
):

INPUT: initial approximation x = x
0
= (
0
,
0
,T
0
), and the tolerance TOL;
maximum number of iterations N.

OUTPUT: approximate solution x = (,,T ) or a message that the number of
iterations was exceeded before the tolerance requirement was meet.

Step 0 Compute and store ( )
0 0
x P R =

Step 1 Set k = 1, Set n = length(x) = # of elements in x.

Step 2 While (k N) do Steps 3-8.

Step 3 Call Aero-F code to calculate P(x) and J(x), where J(x)
i,j
is the partial
derivative of P
i
(x) w.r.t. x
j
for 1 i, j n, where x
1
= , x
2
= , and x
3
= T.

Step 4 Solve the n x n linear system J(x)y = -P(x).

Step 5 Set x = x + y, i.e. x
k+1
= x
k
+ y
k

Step 6 Compute ( )
k k
P R x = .

Step 7 If ||R
k
|| TOL * ||R
0
||, then OUTPUT (x);
(Procedure completed successfully.)
STOP

Step 8 Set k = k + 1.

Step 9 OUTPUT (Maximum number of iterations exceeded before tolerance met.)
(Procedure completed unsuccessfully.)
STOP

Você também pode gostar