Você está na página 1de 15

AIAA Guidance, Navigation, and Control Conference AIAA 2010-8338

2 - 5 August 2010, Toronto, Ontario Canada

Robustness Analysis on Pitch Control Law for an Executive


Transport Aircraft

Jong-Yeob Shin *

Gulfstream Aerospace Corporation,

Savannah, GA
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

ABSTRACT

This paper describes the results of robustness analysis of a pitch control law
designed at Gulfstream Aerospace Corporation for an executive transport aircraft.
In the robustness analysis framework, robust stability and performance are
calculated. In robust performance, worst-case gain and phase margins, worst-case
bandwidth (fastest and slowest), variations of handling qualities are calculated over
the predefined uncertain aerodynamic coefficient bounds. Handling quality
robustness analysis is performed using Low Order Equivalent System (LOES)
estimation techniques in the presence of aerodynamic uncertainties. Illustration is
presented that the nominal performance of the closed-loop aircraft meets the piloted
aircraft handling qualities requirements, and stability requirements. Using a
robustness analysis framework, it is also demonstrated that the closed-loop aircraft
can meet the requirements over the uncertain aerodynamic coefficients.

Nomenclature
AFC = Advanced Flight Controls program, a research and development program within Gulfstream
AoA = Angle of attack
ASE = Aero-Servo-Elastic
CAP = Control Anticipation Parameter
CFD = Computational Fluid Dynamics
C.G. = Center of Gravity
GM = Gain Margin
LFT = Linear Fractional Transformation
LOES = Low Order Equivalent System
HOS = High Order System
LQR = Linear Quadratic Regulator
PID = Proportion, Integral, and Derivative
PM = Phase Margin
REU = Remote Electronics Unit
Vmo = Maximum Operating Velocity

*
Flight dynamics and Control, AIAA senior member, jong.yeob.shin@gulfstream.com

1
American Institute of Aeronautics and Astronautics

Copyright 2010 by JY Shin. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
I. Introduction

Gulfstream is using the Advanced Flight Control Research and Development Program to develop
flight control technology, including a pitch axis flight control law discussed in Refs. [1-3]. In the control
design process, the key requirements were included to meet speed and maneuver stability, to meet key
handling qualities metrics, to retain conventional aircraft-like behavior, and to ensure ASE stability over
the flight envelope.
To design the pitch control law, linear models of the aircraft rigid body dynamics, the sensor
dynamics, the actuation, and the flexible-body modes were created for all flight conditions at a range of
mass and C.G. configurations. These were connected to form an analytical model for executing
performance and stability analyses using a multidisciplinary control law design tool developed in Ref. [1].
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

In the multidisciplinary control law design setup, ASE coupling requirements, stability and
performance (pole position and damping coefficients) requirements and handling qualities requirements
are considered at each flight condition for all weight and C.G configurations. Once the designed control
law has met all key requirements based on linear analysis results, nonlinear analysis was performed to
validate the control law before flight testing. One of the important roles of the nonlinear analysis is to
explore pilot interaction with the aircraft and control system. To provide a capability of testing pilot
interactions, the Conceptual Advanced Simulation Environment (CASE) simulator has been used. Using
CASE Lab test results, turn-around time for refining control law gains is shortened, as in Ref. [4].

After designing the control law, it is important to show if robustness exists in stability and
performance due to possible uncertainties including un-modeled dynamics and aerodynamic coefficient
variations. Robustness of the designed control law is also important to mitigate risk in testing the control
law in flight. There are several methods to do robustness analysis. Here, those are roughly classified as
two methods. One way is using a Monte Carlo method, which is widely used in the industry. In this
method, linear models are regenerated over all possible aerodynamic coefficient variation ranges and
linear analysis is performed for stability, handling qualities and ASE requirements. The method is
computationally expensive. Also, the computational cost increases as number of uncertain aerodynamic
coefficient parameters increase. An alternative way is to use a multivariable control technique [5-14].
Singular value analysis (Mu-analysis) [6] is one of them. It has been well developed in control theory [7-
11] and has been applied in Ref. [12]. In the technique, uncertain dynamic systems are represented in LFT
form and robustness is represented by robust stability margin [5]. Robust stability margin is represented
by a maximum singular value of the uncertain dynamic system to indicate the maximum size of uncertain
parameter bounds to keep the system stable. When an uncertain system is robustly stable, the performance
of the uncertain system is calculated in terms of the H norm of error signals which can be regulated
over the uncertain parameter space. Using a skewed-mu method [10, 12], the maximum degradation of
performance of the system can be also calculated over the uncertain parameters. This method is applied to
perform worst-case performance analysis. The method is used to calculate worst-case gain and phase
margins of the X-38 Crew return vehicle V132 in Ref. [12, 15].

In this paper, a singular value analysis technique is applied for calculating robustness of a designed
pitch control law in terms of worst-case stability margins, worst-case bandwidth variations, and handling
qualities variation. Handling qualities analysis covers short period frequency and damping ratio changes
using LOESs. Worst-case stability margin is calculated as the smallest gain and phase margin over
uncertain aerodynamic coefficient variations using the worst-case gain analysis in Ref. [12]. Using the
same technique, cross-over bandwidth of loop gains can be calculated. The lowest bandwidth system can
represent performance degradation from the nominal system with trade-offs in stability margin, as shown

2
American Institute of Aeronautics and Astronautics
in Ref. [3]. Handling qualities analysis is based on a LOES [1] model, which is obtained by fitting a HOS
transfer function of the closed-loop system over a specific frequency range. Handling qualities variations
of the uncertain closed-loop system are described here. Using the same technique, a handling qualities
analysis framework is demonstrated when the acceptable short period and damping ratio range are
defined. Flight testing for handling qualities can help refine the desirable short period frequency and
damping range for a particular configuration. This information can be included in robustness handling
qualities analysis. The worst-case performance analysis technique is applied to answer the question
Does the pitch control law provide a desirable range of handling qualities in the presence of uncertain
aerodynamic coefficients?
The structure of this paper is as follows. In Section 2, a well developed robustness analysis method is
briefly described. In Section 3, the robustness analysis problem of a pitch control law is described. In
Section 4, a robustness analysis framework is described using robust analysis toolbox (worst-case gain
analysis). In Section 5, analysis results are presented and in Section 6, the paper is concluded with
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

summary.

II. Overview of Robust Analysis

This section briefly describes robust analysis and worst-case analysis more fully developed in Refs.
[6, 12] in the presence of the pre-defined parameter variations. An uncertain dynamic system due to
parameter variation can be written in the form of LFT, as shown in Fig. 1.

M 11 M 12
M M 22
z 21 w

Figure 1. LFT form of the uncertain dynamic system

The transfer function from w to z is written as


Fu ( M , ) M 22 M 21 ( I M 11 ) 1 M 12 (1)
where matrix M represents system matrices and the block represents uncertain parameters. The
uncertainty block is defined as [12]
{diag ( 1r I r1 , , mr I rm , 1c I c1 ,, lc I cl )} (2)
where the parameters r and c represent real and complex uncertainties. The uncertainty block is also
norm bounded as one without loss of generality. In this case, robust stability of the uncertain system is
checked by calculating singular the value of system M 11 . When the singular value is less than one, the
system is robustly stable over the parameter varying range. Robust stability margin is also calculated
based on the singular values [6-8] as the inverse of the singular value.

Using the LFT form shown in Fig. 1, the worst-case gain analysis can be applied to calculate
maximum gain of the transfer function from w to z over the uncertain parameter range. The
optimization is written as
max ( Fu ( M ( jw), ) (3)
jw,

3
American Institute of Aeronautics and Astronautics
The detailed worst-case analysis method is already described in Robust Control Toolbox [6]. The
application of worst-case analysis is also demonstrated in Ref. [12].

III. Robustness Analysis Problem on Pitch Control Law

This section describes the robustness analysis problem of a pitch control law designed at Gulfstream.
The analysis considers closed-loop dynamics changes due to predefined aerodynamic coefficient
variations. A pitch control structure shown in Fig. 2 is a PID type of control law with a response model in
the feed forward path. Note that this structure of control law is not the same as a model following
structure. In order to define gains, the modern control technique known as LQR was used. Here, the LQR
method is only used to guide certain control gains. As such, some of the good properties of robustness
from LQR are degraded. In practical applications of LQR control, the robustness properties are usually
degraded due to sensor and actuator dynamics and communication time-delay. The technique is still
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

valuable to guide the selection of several gains at once.

In the gain design process in Refs. [1-3], the cost function in LQR is defined as the sum of two
penalties, namely: short period mode penalty and phugoid mode penalty. The short period mode penalty
is constructed as the third order function of natural frequency of open-loop short period, z-directional
acceleration and tracking errors of NzU [1]. Note that one desired parameter, the product of stick force
per g and knots per stick force, are also included in the penalty function. This structure results in three
target zeros in the LQR design setup [1-3]. The associated cost function penalties were set to achieve the
desired short period mode frequency and damping, and angle of attack command bandwidth. The
penalties on the elevator command were adjusted to set the corresponding desired control loop bandwidth.
The phugoid mode penalty is constructed as the first order function to help provide some phugoid mode
damping. The LQR algorithm minimizes the total criteria and computes gains on each state in the overall
synthesis model. The detailed process of defining control synthesis structure is described in Refs. [1-3].
Column input
Response model

Longitudinal
Elevator
Actuator Dynamics

Feed back V
control law: Sensor
AOA dynamics
PID type and
Filters q and
instrument
Theta correction
Nx terms
Nz

Figure 2. The overall controlled system structure for the longitudinal dynamics

Compared with many heritage medium transports, the introduction of fly-by-wire can add significantly
to the complexity of the control dynamics. Items that must be modeled include actuators and sensors, the
digital operations along with their processing delays and filters, and integration and filtering designed to
shape the response to pilot control. A typical model of the pitch attitude response to pilot control is 105th

4
American Institute of Aeronautics and Astronautics
order, compared with the customary fourth-order-plus-an-actuator model that represents the phugoid and
short period dynamics of a heritage aircraft.

It is important to determine whether the designed control system has robustness in stability and
performance due to the aerodynamic coefficient variations. In this paper, the following aerodynamic
coefficients are considered as uncertain parameters. The most uncertain aerodynamic parameter is
pitching moment coefficient due to pitch rate, Cmq , which is one of the rotary derivatives. This parameter
is important in longitudinal dynamics since it is the largest contributor to damping of short period motion,
as discussed in Ref. [10]. Its nominal value comes from a panel method approximation, and uncertainty is
treated as 40% of the nominal value. The longitudinal static stability derivative, Cm , has uncertainty of
10% of the nominal value. The parameter is related with short period natural frequency [16]. The lift
curve slope, CL , has uncertainty of 10% of the nominal value. It is important to ride quality and is also
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

related with short period damping [16] and the handling qualities CAP. The elevator effectiveness to lift
and pitch moment, CLde , and Cmde are important in control authority and closed-loop stability. In this
circumstance, there is some mismatch between wind tunnel data and computational predictions. The gaps
are modeled as an uncertain parameter, as a function of Mach number. Rotational moment of inertia in
the pitch axis, I yy , is also considered to have uncertainty of 20% of the nominal value. In this robustness
analysis, therefore, the uncertain parameters are Cmq , Cm , CL , CLde , Cmde , and I yy . Note that the
nominal value of each uncertain parameter is defined based on computational predictions or wind tunnel
test results.

Robustness analysis problems of the pitch control law with the uncertain parameters are as follows:
Is the closed-loop system robustly stable over the uncertain parameters?
What is the worst-case gain and phase margin for each broken loop when the closed-loop system
is robustly stable?
What are HQ variations for the specific aerodynamic coefficient values which lead to worst-case
gain and phase margins?
Does the pitch control law plus aircraft meet the desired handling qualities over the uncertain
parameters? The desired handling qualities are defined as acceptable short period frequency and
damping ratio.

IV. Robustness Analysis Framework

This section describes an analysis framework used for the problem. In order to apply the singular
value analysis method to this problem, the uncertain parameters must be described in LFT form. To
describe the uncertain longitudinal dynamics due to aerodynamic coefficient variations in the form of
LFT, the longitudinal equations of motion should be analytically linearized as a function of the uncertain
parameters. We take the detailed linear longitudinal model description from Ref. [10] and substitute the
uncertain parameters into the model. Using the Robust Control Toolbox, the longitudinal motion is easily
modeled in the form of LFT with the uncertainty block. Here the block is written as

diag ([ Cm q , Cm , Cm de , CL , CLde , I yy ]) .
The closed-loop system can be written in the form of LFT described in section 2 with the uncertain
block. For worst-case gain and phase margin analysis, disk stability margin is considered, as in Ref. [12,
15].

5
American Institute of Aeronautics and Astronautics
The disk stability margin is defined as the radius of the disk from -1 point to the loop gain in the Nyquist
plot (see Fig. 3)

L
-1
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

Cr

Figure 3. Illustration of disk stability margin

The radius Cr (shown in Fig. 3) is written as ||1 + L || where L is the loop gain for each signal loop.
For worst-case gain and phase margin, the optimization is formulated as
1
max || || (4)
,w 1 L( jw)
for each loop. For example, to calculate worst-case gain and phase margin on the pitch rate broken loop,
we add input (w) and output (z) on the pitch rate feed back signal shown in Fig. 4. The transfer function
from w to z represents 1/(1+L).

diag ([ Cm q , Cm , Cm de , CL , CL de , I yy ])

LFT
Elevator Longitudinal
Actuator
Model
z w
V
Feed back Sensor
control law: AOA dynamics
PID type and q and
Filters instrument
Theta correction
Nx terms
Nz

Figure 4. Worst-case gain and phase margin analysis framework


Note that the optimization is over the uncertain parameters and specific frequency ranges. The lower
frequency range about phugoid motion is excluded. For worst case gain and phase margin analysis, each
individual loop at sensor outputs plus elevator command is analyzed.

6
American Institute of Aeronautics and Astronautics
In order to calculate the worst-case cross-over bandwidth, the elevator loop gain is described in the
form of LFT as shown in Fig. 5. The worst-case cross-over bandwidth is calculated as the largest
deviation from the nominal elevator loop gain over the specific frequency range. It is formulated as
max || Fu ( L( jw), ) L( jw) || .
w,

diag ([ Cm q , Cm , Cm de , CL , CL de , I yy ])

Elevator loop gain


Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

Nominal elevator
loop gain

Figure 5. Worst-case cross-over bandwidth analysis framework

For handling qualities variation of the closed-loop system due to the uncertain dynamics, we check
handling qualities criteria such as CAP and bandwidth from column to attitude, as in Refs.[17,18] based
on the worst-case analysis results. The specific aerodynamic coefficient values are calculated for worst-
case gain and phase margin for each loop and for worst-case cross-over bandwidth. These values are
folded into the LFT closed-loop system. The high order transfer function from column to pitch attitude is
fitted with a LOES. It shows roughly the handling qualities variation of the uncertain closed-loop system
in the presence of uncertain aerodynamic coefficients.

For robust handling qualities analysis, the desirable handling quality range is predefined. In this paper,
CAP and short period damping ratio ( sp ) are considered as handling qualities criteria. The desirable
CAP range represents the desirable short period frequency range. Recall that the robust handling qualities
analysis problem in this paper is Does the pitch control law plus aircraft meet the HQ criteria (using
LOES) over the variations of the uncertain aerodynamic coefficients?. In order to answer the question,
the following steps were taken.

1) The bounds of column to pitch attitude frequency response variations are calculated over the uncertain
parameter ranges. The bounds are called the worst-case performance frequency response, The maximum
gain bound is calculated as
max || Fu (T ( jw), ) || (5)
,

where T ( j ) represents frequency response from column to pitch attitude and is the uncertain block.
The calculated uncertain block is called 1 . The minimum gain bound of possible frequency responses is
also calculated by using the worst-case gain algorithm [6, 12]. In order to apply the algorithm, the
acceptable maximum bound of frequency responses is needed, which is calculated as the upper bound of
bowtie from the LOES [19]. The LOES here is calculated from the desirable HQ range (CAP and short
period damping ratio range). Note that the bowtie bounds are defined as the envelope of maximum
unnoticeable added dynamics [20]. Once the acceptable maximum bound is calculated, the minimum gain
bound is calculated as

7
American Institute of Aeronautics and Astronautics
max || Fu (T ( j ), ) Tmax b ( j ) || (6)
,

where Tmax b ( j ) is calculated as the acceptable maximum bounds of LOES. The calculated uncertain
block is called 2 . Thus, all possible frequency responses should be between Fu (T ( jw), 1 ) and
Fu (T ( jw), 2 ) over the interested frequency range.
2) The LOESs of the calculated bounds of frequency responses are calculated. The criteria of LOES
fitting used in this paper has two parts: a) the LOES should match the HOS within bowtie and a low cost
function [19, 20] and b) the LOES should indicate the short period frequency damping ratio and CAP in
the desirable HQ range. The cost function [20] is defined as
n

{[G
20
cos t HOS ( i ) G LOES ( i )] 2 0.02[ HOS ( i ) LOES ( i )] 2 }
n i 1
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

where G HOS and HOS represent gain and phase of HOS, and G LOES and LOES represent gain and phase
of LOES. The number of discrete frequencies used in the match is represented by n . In this paper, the
maximum allowable cost value is set as 50.

3) When the CAPs and short period damping coefficients of the LOESs are in the desirable range for
the upper and lower bounds of frequency responses calculated in step 1. The robust HQ analysis stop. In
the case, we can say that the all possible closed-loop system over the uncertain aerodynamic coefficient
variations can meet the desirable HQ. However, when the criteria of LOES in step 2 are not satisfied,
the aerodynamic coefficient uncertain range is reduced until the conditions are satisfied.

For example, consider the case that the worst-case performance frequency responses with the
calculated block 1 do not allow the LOES and HOS to match within the bowtie bounds. The closed-
loop system over the uncertain aerodynamic coefficient variation can not meet the desirable HQ. In this
case, the uncertain parameter bounds in 1 must scaled down according to their sensitivity to the transfer
function. Using the redefined values, step 2 is repeated until the condition is satisfied.

V. Robustness Analysis Results

A. Robust stability

For robust stability of the closed-loop system, singular values of the LFT system are considered over
the flight envelope. Here, the analysis frequency range is specified as [0.2 100] rad/sec since dynamics
of the lower frequency range can be compensated for by pilot input. To check robustness over the flight
envelope, linear models are generated for each flap configuration with C.G and weight variations. The
maximum singular value of the tested cases is shown in Table 1.

Table 1. Robust stability over flight envelope


Flap 0 Flap 10 Flap 20 Flap 39
# of linear models 628 348 260 232
Maximum singular value 0.9 0.85 0.85 0.81

The worst-case robust stability is at the maximum operating airspeed in the cruise configuration. This
is caused by the magnitude of the uncertainty in elevator control effectiveness. The current range of the

8
American Institute of Aeronautics and Astronautics
parameter at high mach number around the Vmo condition is set at 45% to 105% of nominal values based
on wind tunnel and CFD data.

Note that the pitch control law is a PID type gain-scheduled control law. It would be of interest to
apply gain-scheduled analysis [11] for robust stability using the LMI approach [14]. However, the LMI
approach has not been considered here. It would also be of interest to compare these results in further
research.

B. Worst-case gain and phase margin

Worst-case gain and phase margins of each loop over the generated linear models in Table 1 are
calculated. All of the analysis results cannot be presented due to limited space. Here, one example is
presented: the pitch rate loop at one flight condition with flaps in the landing configuration. The nominal
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

pitch rate loop has 9.9 dB GM and 50 deg PM, which meet the industry standard stability requirements.
Using the worst-case gain and phase margin optimization described in section 4, the minimum disk size is
calculated as about 0.5. Also, the worst-case uncertain parameters are calculated. Using the worst-case
uncertain parameters, the pitch rate loop gain is plotted in Fig. 7.

Pitch Rate Loop Pitch rate Loop


40 4

30 3

20
2

10
Ga 1
in
(d 0
B) 0
-10

-1
-20 0.2 rad/sec
0.5 rad/sec
1 rad/sec -2
-30
2 rad/sec
5 rad/sec
-40 -3
-350 -300 -250 -200 -150 -100 -50 0 -6 -5 -4 -3 -2 -1 0 1
Phase (deg)

Figure 7. Pitch rate loop in a) Nichole chart and b) Nyquist plot.

In Fig. 7, the blue line represent worst-case pitch rate loop over the uncertain aerodynamic coefficient
variations. In the Nyquist plot on Fig. 7, the red circle represents the minimum disk stability (radius of
0.5), which is calculated from the maximum norm value. From the geometric calculation, GM and PM
are calculated as 6 dB and 28 deg using
4 (Cr 2 2) 2
GM 20 log(1 /(1 Cr )), PM 57.3 sin 1 , (7)
4
where Cr is 0.5.

Using the worst-case pitch rate loop with the calculated aerodynamic coefficient values, gain margin is
7 dB and phase margin is about 40 deg. In Fig. 7, red area in Nichols chart represents stability
requirements (6dB and 45 deg), which is used in the control design criteria. It is shown that the worst-
case gain and phase margins do not meet the nominal case requirements. However, 40 deg PM for worst-
case performance is acceptable. Note that the calculated gain and phase margins from disk stability

9
American Institute of Aeronautics and Astronautics
margin are always smaller than the worst-case gain and phase margin based on Nichols chart. It
represents that disk stability margins are conservative in stability margin calculation.

C. Worst-case cross-over frequency analysis

In this analysis, the largest deviation from the nominal case in terms of cross-over frequency is
calculated. In the previous section, the analysis framework is described. Using the analysis framework,
the specific aerodynamic coefficient values over the varying ranges are calculated using worst-case gain
analysis with the frequency range [1 3] rad/sec. For example, the elevator loop gain is shown in Fig 8.
The magenta line is the elevator loop gain with the worst-case aerodynamic coefficients for cross-over
frequency analysis. The green line represents the nominal elevator loop gain. In order to check the
analysis results, 200 samples over the uncertain parameters are selected randomly. Using the sample
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

aerodynamic coefficients, elevator loop gains are plotted in Fig. 8 as blue lines. It shows that the magenta
line represent the lowest cross-over frequency (bandwidth) over the uncertain parameter ranges. The red
line is the elevator loop gain with the calculated aerodynamic coefficients for the worst-case gain and
phase margin analysis. It makes sense that the case with the highest loop gain (highest crossover) would
have the minimum stability margin.

50

0
Mag (dB)

-50

-100
-2 -1 0 1 2
10 10 10 10 10
Freq (r/s)

Figure 8. Elevator loop gains with worst-case cross-over frequency and 200 samples over the parameter
range.

D. Handling quality robustness analysis results

Before presenting analysis results based on the framework of HQ robustness analysis described in the
previous section, HQ variations are presented due to the worst-case aerodynamic coefficients calculated
for the worst-case gain and phase margin for each broken loop and for worst-case cross-over frequency.
The closed-loop system is evaluated based on the calculated aerodynamic coefficients. Using LOES
fitting, CAP and short period damping are computed in Fig. 9.

10
American Institute of Aeronautics and Astronautics
MIL STD CAP requirements - Category C, Class II-L Short Period Damping
2
10

Short period frequency - rad/sec

1
10 2
EL
LE V

1 0
EL 10
LE V

SP
1
EL
LE V
3
L 2&
L EV E
10
0
Level 1
LEVEL 3 LEVEL 2
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

3
EL
LE V

-1 -1
10
-1 0 1 2
10
10 10 10 10 -2 -1 0 1
Nz/alpha - g's/rad 10 10 10 10
CAP

Figure 9. handling quality variations over the calculate uncertainties.

Note that the green symbol in Fig. 9 represents nominal case. The magenta, blue, black, and yellow
symbols represent the worst-case gain and phase margin of AoA broken loop, pitch rate broken loop,
elevator broken loop, and pitch attitude broken loop, respectively. The cyan symbol represents the worst-
case cross-over frequency. The blue symbol position is same as black symbol position in Fig 9.

When LOES is estimated from HOS, the HOS should be within the bowtie plot shown as the black
line in Fig. 10. The blue and the green lines represent HOS and LOES frequency responses, respectively.
L O E S F it t in g
6 0

5 0

4 0

3 0
gain, dB

2 0

1 0

-1 0
L O S
-2 0
-1 0 1
1 0 1 0 1 0
f re q u e n c y , ra d /s e c

1 5 0

1 0 0

5 0
phase, deg

-5 0

-1 0 0

-1 5 0

-2 0 0

-2 5 0
-1 0 1
1 0 1 0 1 0
f re q u e n c y , ra d /s e c

Figure 10. LOES for Nominal HOS of column to theta.

In Fig 10, the blue and the green lines represent HOS and LOES frequency responses, respectively. It
shows that LOES can represent the HOS dynamics. Note that, when HOS is not in the bowtie bounds,
LOES may not represent HOS dynamics. In Fig.9, the circle symbols represent that HOS for each case is
not in bowtie bounds.

11
American Institute of Aeronautics and Astronautics
In order to use HQ robustness analysis, the desirable CAP and short damping ratio range are defined in
Fig.11. In this paper, how to define the desirable range is out of the scope of this paper.

Short Period Damping

0
10
SP

Level 1
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

-1
10
-2 -1 0 1
10 10 10 10
CAP

Figure 11. The desirable CAP and short period damping coefficient ranges

The nominal closed-loop system meets the desirable HQ shown in Fig. 11 as blue symbol. Due to
uncertain aerodynamic coefficients, the closed-loop frequency response varies. The bounds of the
variations are calculated using HQ robustness analysis step 2.

40

20
agdB

0
M

-20

-40
-1 0 1
10 10 10
Freq rad/s

-50
e
dg

-100
h
Pae
s

-150

-200

-250
-1 0 1
10 10 10
Freq rad/s

Figure 12. 200 sampled closed-loop frequency responses over the uncertain aerodynamic coefficient
ranges.

In Fig.12, aerodynamic coefficient values are randomly selected over the uncertain bounds and using
the values, closed-loop frequency responses are plotted. The magenta line represents one extreme, as
calculated using Eq. (5). In the calculation, frequency range is defined as [2 - 8] rad/sec since the short
period frequency of the open-loop aircraft is within that range. The green line represents the other
extreme, as calculated using Eq. (6). The green line frequency response is fitted with a LOES and the
frequency response is within the bowtie as compared to the HOS. CAP and short period damping
coefficient calculated from the LOES are in the desirable HQ range. The magenta line frequency

12
American Institute of Aeronautics and Astronautics
response is fitted with a LOES and the frequency response is not in within the bowtie bounds. The
uncertain aerodynamic bounds are reduced until the fitting criteria is met.

The new re-defined uncertain aerodynamic coefficient bounds are written in Table 2.

Table 2. Uncertain Aerodynamic Coefficient Range


CL CLde Cm Cmde Cmq I yy
Redefined [0.9 - 1.1] [0.68 - 1.1] [1.1 - 0.98] [1.05 - 0.62] [1.6 - 0.96] [0.96 - 1.2]
Range * CL _ nom * CLde _ nom * Cm _ nom * Cmde _ nom * Cmq _ nom * I yy _ nom
Previous [0.9 - 1.1] [0.38 - 1.1] [1.1 - 0.90] [1.10 - 0.62] [1.6 - 0.60] [0.80 - 1.2]
Range * CL _ nom * CLde _ nom * Cm _ nom * Cmde _ nom * Cmq _ nom * I yy _ nom
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

In Table 2, the range is described as a value proportional to each nominal aerodynamic coefficient. It is
noted that the bounds of the pitching moment coefficient due to pitch rate Cmq is reduced to 4% from
40% range at upper value. Also, the pitch moment coefficient due to angle of attack is reduced to 2%
upper value, which is related with short period frequency of open-loop dynamics. Recall that the
redefined range is due to linear searching results and it leads to conservative results. Also, the criteria
used here for LOES fitting may be too restrictive. When the HOS is slightly outside the bowtie bounds,
engineering judgment is required. Another possible cause is that the LOES fitting method is not always
robust. In other words, it may be possible to fit a different LOES that matches the HOS, but the
automated search algorithm did not find it. When HOS is allowed to be the outside the bounds at 5% of
total fitting in magnitude and phase over the frequency range, the previous ranges of uncertain
aerodynamic coefficient parameters are acceptable and all LOES are in the desirable HQ range.

To verify the analysis results, 200 sampled closed-loop systems over the redefined aerodynamic
coefficient ranges are evaluated. The CAP and short period damping coefficients for all 200 LOESs are
plotted in Fig. 13.
Short Period Damping

0
10
SP

Level 1

-1
10
-2 -1 0 1
10 10 10 10
CAP
Figure 13. CAP and short period damping ratios from LOESs of 200 sampled closed-loop systems

13
American Institute of Aeronautics and Astronautics
It is noticed from Fig. 13 that all closed-loop systems over the redefined uncertain aerodynamic
coefficient range have met the desirable HQ range. Using this HQ robustness analysis method, we can
predict the closed-loop can meet robustly the desirable HQ range. Also, the uncertain parameter range can
be redefined to meet the desirable HQ range. Recall that a searching method over the uncertain parameter
space to redefine uncertain bounds is further research area.

VI. Conclusion

A robust control analysis method is applied in this paper to calculate the robustness of a pitch control law
designed at Gulfstream. Robustness is demonstrated in terms of gain and phase margin variations, cross-
over frequency variations, and handling qualities. The pitch control law design provides an executive
transport aircraft that is robustly stable over the uncertain aerodynamic variations. Uncertainties included
pitch control effectiveness as a function of Mach number and other key varying parameters. The closed-
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

loop system is analyzed over the entire flight envelope including cruise, approach, and landing
configurations. Handling qualities variations over the parameter range are calculated. It is shown that the
pitch control law and aircraft met the HQ requirements for a range of variation. A robustness handling
qualities analysis framework is also demonstrated with the desirable short period frequency and damping
ranges.

Acknowledgments

Author thanks to Dagfinn Gangsaas, John Hodgkinson, Clay Harden, and flight control law team
members at Gulfstream for helpful discussion and guidance during the design of the control laws and
linear analysis including handling qualities analysis.

VII. References

1) Gangsaas, D. Hodgkinson, J., Harden, M., Saseed, N., and Chen, K., Multi-disciplinary Control
Law Design and Flight Test Demonstration on a Business Jet, AIAA Guidance, Navigation and
Control Conference, Honolulu, Hawaii, 2008, AIAA-2008-6489.

2) Gangsaas, D., Bruce, K. R., Blight, J. D., Ly, U.-L, Applications of modern synthesis to aircraft
control: three case studies, IEEE Transactions on Automatic Control, 31, 1986, pp.995-1014.

3) Shin, J-Y., Harden, C., Saeed, N., Hartley, R., and Chen, K., Pitch Control Law Analysis and Flight
Test demonstration of an Executive Transport Aircraft, AIAA Guidance, Navigation, and Control
Conference, Chicago, IL, 2009, AIAA-2009-6126.

4) Harden, C., Saeed, N., and Grosse, G., Development of a Pilot-In-The-Loop Aircraft Simulation
Laboratory, AIAA Modeling and Simulation Technologies Conference and Exhibit, Keystone, CO,
2006.

5) Balas, G., and Hodgkinson, J, Control Design Methods for Good Flying Qualities, AIAA
Atmospheric Flight Mechanics Conference, Chicago, IL, 2009, AIAA-2009-6319.

6) Balas, G., Chiang, R., Packard, A., and Safonov, M., Robust Control ToolBox, the MATHWORKS,
2008.

14
American Institute of Aeronautics and Astronautics
7) Young, P., and Doyle, J., Computation of mu with Real and Complex Uncertainties in Proceedings
of the 29th IEEE Conference on Decision and Control, Vol2, (Honolulu, Hawaii), 1900, pp1230-1235.

8) Packard, A., The Complex Structure Singular Value, Automatica, Vol 29, No. 1, 1993, pp. 71-109.

9) Fan, M., Tits, A., and Doyle, J., Robustness in the Presence of Mixed Parametric Uncertainty and
Unmodeled Dynamics, IEEE Transactions on Automatic Control Vol. AC-36, 1991, pp. 25-38.

10) Packard, A., Balas, G., Liu, R., and Shin, J-Y., Results on Worst-Case Performance Assessment, in
Proceedings of America Control Conference, (Chicago, IL), 2000, pp.2425-2427.

11) Young, P., Newlin, M., and Doyle, J. , Practical Computation of the Mixed Mu problem, in
Proceedings of the America Control Conference, Chicago, IL, 1992, pp.2190-2194.
Downloaded by Indian Institute of Technology on October 1, 2012 | http://arc.aiaa.org | DOI: 10.2514/6.2010-8338

12) Shin, J-Y., Balas, G., and Packard, A., Worst-Case Analysis of the X-38 Crew Return Vehicle
Flight Control System, Journal of Guidance, Control and Dynamics, Vol. 24, No. 2, 2001, pp.261-
269.

13) Zhou, K., Doyler, J., and Glover, K., Robust and Optimal Control, Prentice Hall, 1996. +

14) Dullerud, G., and Paganini, F.:A Course in Robust Control Theory, A convex Approach, Springer,
2000. Ch.1.

15) Shin, J-Y., Worst-Case Analysis and Linear Parameter-Varying Gain-Scheduled Control of
Aerospace Systems, Ph.D. Thesis, University of Minnesota, 2000.

16) Nelson, F.R., Korener, W., and Truderl, R.E., Dynamics of the Airframe, Vol II, Fundamentals of
Design of Piloted Aircraft Flight Control systems, BuAer Report AE-61-4, 1953.

17) Duke, E., Antoniewicz, R., and Krambeer, K.: Derivation and Definition of a Linear Aircraft
Model, NASA Reference Publication 1207, Ames Research Center, 1988.

18) Hodgkinson, J., Wood, J.R., and Hoh, R.H.; An Alternate Method of Specifying Bandwidth for
Flying Qualities. AIAA Atmospheric Flight Mechanics Conference, Seattle, WA, August 1983.

19) Hodgkinson, J., LaManna, W., and Heyde, J.: Handling Qualities of Aircraft with Stability and
Control Augmentation Systems - A Fundamental Approach. J.R.Ae.S., February 1976.

20) MIL-STD-1797B, Flying Qualites of Piloted Aircraft, Department of Defense Interface Standard,
MIL-STD-1797B , 15 Feb., 2006.

15
American Institute of Aeronautics and Astronautics

Você também pode gostar