L§AMPLED-DATA FREQUENCY RESPONSE SYSTEM ...
-
Upload
khangminh22 -
Category
Documents
-
view
2 -
download
0
Transcript of L§AMPLED-DATA FREQUENCY RESPONSE SYSTEM ...
L§AMPLED-DATA FREQUENCY RESPONSE
SYSTEM IDENTIFICATION FOR
LARGE SPACE STRUCTURESI
A Thesis Presented to
the Faculty of the College of
Engineering and Technology
Ohio University
In Partial Fulfillment
of the Requirements of the Degree
Master of Science
by
Thomas T. Hammon~
June, 1988
ACKNOWLEDGMENTS
First, I would like to extend my gratitude to the
Faculty and Staff of the Department of Electrical and
Computer Engineering. Their tireless efforts have aided me
immensely in my graduate studies at Ohio University.
Special thanks is due to Dr. Jerrel Mitchell, my
advisor, for suggesting the thesis topic, guiding my
research, and editing the thesis drafts. I appreciate the
input of my thesis committee, Dr. G.V.S. Raju, Dr. O.E.
Adams, Jr., and Dr. R. Dennis Irwin, to the final draft of
this thesis. Also, I would like to thank Bill Maggard for
his cooperation in the experimental stage of this work.
Finally, I would like to express my appreciation to my
family whose love and encouragement made this endeavor
possible. My mother, Harriet, and late father, Claude,
taught me the value of education and demonstrated the
perseverance required to obtain it; and my fiancee, Judy,
steadfastly supported my efforts with her love and patience.
i
TABLE OF CONTENTS
ACKNOWLEDGMENTS
LIST OF FIGURES
LIST OF TABLES
Chapter 1. INTRODUCTION
1.1 Control of Large Space structures
1.2 Design-To-Performance
1.3 One-Controller-At-a-Time
1.4 Summary
Page
i
iv
viii
1
1
3
4
8
2.1 Overview
2.5 Stochastic Analysis
2.6 Summary
2.2 Sampled-data Systems
2.3 Fourier Transforms
2.4 Deterministic Analysis
3.6 Summary
3.1 Signal Properties
3.2 unit Pulse
3.3 Increasing Magnitude
3.4 Increasing Duration
3.5 Shaping Pulses
Chapter 2.
Chapter 3.
SYSTEM IDENTIFICATION
EXCITATION ENHANCEMENT
9
9
11
14
18
22
23
24
24
26
30
32
37
44
ii
Chapter 4. EXPERIMENT DESIGN ...••...........•..•...... 46
4 • 1 Obj ective 46
4 • 2 Physical Model 4 7
4.3 Mathematical Models 49
4.4 Digital Simulation 52
4 • 5 Computations 60
4 • 6 Digital Results 62
Chapter 5. EXPERIMENT IMPLEMENTATION 71
5.1 Hybrid Experiment 71
5 • 2 Analog Computer 73
5.3 Data Acquisition/Control Unit 75
5 • 4 Digital Computer 76
5. 5 Hybrid Results 77
Chapter 6. CONCLUSION 107
6.1 Observations
6.2 Applications
107
108
RE~ERENCES 110
Appendix A. DIGITAL SIMULATION SOURCE CODE 113
Appendix B. HYBRID SIMULATION SOURCE CODE 155
iii
LIST OF FIGURES
Page
1.1 Multiple-input, mUltiple-output system 5
2.1 Typical sampled-data control system 12
2.2 Sampled-data control system modeled with ideal
sampler and zero-order-hold 12
2.3 Sampled-data system transfer function 15
2.4 Sampled-data system frequency response function 15
2.5 System with noise modeled at output 20
2.6 Unity negative feedback configuration 20
3.1 z-plane mapping of frequencies up to the half-
sampl ing frequency 27
3.2 unit digital pulse's characteristics 28
3.3 Increased magnitude pUlse's characteristics 31
3.4 Increased duration pulse's characteristics 36
3.5 Binomial triangle shaped pulse's characteristics 40
3.6 Bipolar triangle shaped signal's characteristics 43
3.7 Complementary binomial pair's spectral
characteristic 45
4 • 1 Three disk physical model 48
4.2 Three disk system sampled-data block diagram 50
4.3 Three disk system sampled-data frequency response
function 55
4.4 Three disk system continuous root locus 56
iv
4.5 Three disk system continuous root locus of 0.1 Hz.
mode 57
4.6 Three disk system sampled-data root locus for
unity gain 59
4.7 Three disk system sampled-data frequency response
block diagram 61
4.8 Frequency response function obtained using unit
digital pulse 64
4.9 Frequency response function obtained using
increased magnitude pulse ......................... 65
4.10 Frequency response function obtained using
increased duration pulse .......................... 66
4.11 Frequency response function obtained using
binomial triangle shaped pulse 68
4.12 Frequency response function obtained using bipolar
triangle shaped pulse ........•.................... 69
4.13 Frequency response function obtained using unity
white noise 70
5.1 Hybrid experiment hardware configuration 72
5.2 Analog computer simulation functional block
diagram 7 4
5.3 Spectral characteristics of unit digital pUlse 79
5.4 Frequency response function using unit digital
pulse ....•........................................ 80
5.5 Spectral characteristics of increased magnitude
pulse 81
v
5.6 Frequency response function using increased
magnitude pulse 82
5.7 Spectral characteristics of increased duration
pulse 83
5.8 Frequency response function using increased
duration pulse 84
5.9 Spectral characteristics of binomial triangle
shaped pulse 85
5.10 Frequency response function using binomial
triangle shaped pUlse 86
5.11 Spectral characteristics of bipolar triangle
shaped pulse 88
5.12 Frequency response function using bipolar triangle
shaped pulse 89
5.13 Frequency response function using unity white
noi se •••••••••••••••••••••••••••••••••••••••••••.• 9 a
5.14 Frequency response function using single increa.sed
magnitude record .................................. 92
5.15 Frequency response function using four increased
magnitUde records ................................. 93
5.16 Frequency response function using sixteen
increased magnitude records ....................... 94
5.17 Frequency response function using single binomial
triangle record 95
5.18 Frequency response function using four binomial
triang1e records 96
5.19 Frequency response function using sixteen binomial
vi
triangle records •.........•.•..••......••......... 97
5.20 Frequency response function using single bipolar
triangle record 99
5.21 Frequency response function using four bipolar
triangle records 100
5.22 Frequency response function using sixteen bipolar
triangle records .............•...........•........ 101
5.23 Frequency response function using sixteen spliced
triangle records 102
5.24 Frequency response function using single unity
white noise record 103
5.25 Frequency response function using four unity white
noise records ......•.............................. 104
5.26 Frequency response function using sixteen unity
white noise records 105
vii
1
CHAPTER 1.
INTRODUCTION
1.1 Control of Large Space Structures (LSSs)
As scientists and engineers attempt to utilize the
unique properties of the Earth-orbit environment, it is
clear that larger structures must be built in space to house
these imaginative projects. Currently, there are several
projects being planned that will require a physical
structure many times larger than any previously placed in
orbit. Two notable examples are the National Aeronautics
and Space Administration's Space station and the Department
of Defense's strategic Defense Initiative's Space-Based
Lasers. It is the challenge of control systems engineers to
develop effective methods for controlling the mechanical
dynamics of these huge structures.
The construction of LSSs in orbit dictates the use of a
minimal amount of light-weight materials. Typically,
implies the interconnection of long, thin, flexible beams to
form the framework. This construction technique produces
structures that are very flexible and exhibit very little
mechanical damping. Hence, the unconstrained LSS will
vibrate for an extremely long time when disturbed.
Additionally, due to the complexity of the LSS, these
vibrations, or flexible modes, will occur at several
2
frequencies with significant magnitudes. Effective suppres
sion of these undesirable modes is essential for any LSS
control scheme.
The process of assembling LSSs will span several months
or years. During this time, the testing and use of modules
already completed requires the implementation of structural
control systems while the structure is evolving. The
implication is that the dynamics of the LSS will change
slowly over time. Similar effects will be produced by
reorienting subsystems such as solar panels, docking with
transportation vehicles, expanding and contracting due to
thermal effects, etc. Therefore, the control scheme must be
adaptable to these slowly changing dynamics.
Finally, much of the information on which the control
system is based must be extracted from the structure itself.
Ground-testing of the true structure is not possible, and
even the investigation of scaled models is of questionable
value due to gravitational effects. The method of
finite-element analysis may yield some useful information,
but models obtained in this fashion generally lack
sufficient fidelity when applied to extremely high order
systems. So the successful control scheme must be capable
of incorporating empirical information gleaned from the true
dynamics of the actual LSS.
3
1.2 Design-To-Performance (DTP)
A control system design strategy that has been tailored
to meet the stringent requirements of LSS control is the
Design-To-Performance procedure [1]. The key elements of
DTP are the following:
* Determining LSS characteristics on-orbit.
* Designing an appropriate digital controller.
The design process entails gathering structural response
data for the LSS; extracting a high-fidelity frequency
domain system model from the data; specifying a high-perfor
mance control law based on the model; and implementing the
control law with the digital controller onboard the LSS.
This process may be repeated as necessary to fine-tune the
design or to cope with changes in the dynamics of the LSS.
The DTP concept offers some important advantages in the
LSS application. other control methods often require
sensors and actuators dedicated t~ gathering data to
identify the dynamics of the LSS. The DTP method uses only
those actuators and sensors necessary to control the LSS.
Another attractive feature is the use of frequency domain
models. Methods that require state variable models are
faced with the formidable task of developing these models
for a system that may be of order 100 or more. Furthermore,
4
design procedures for frequency domain models can be used
which produce much simpler control laws than state-space
methods. A design procedure that is particularly well-suit
ed for the DTP philosophy is considered next.
1.3 One-Controller-At-a-Time (l-CAT)
One-Controller-At-a-Time is a design procedure for
mUltiple-input, mUltiple-output (MIMO) feedback control
systems [2]. A linear MIMO system may be viewed as an array
of single-input, single-output (5150) systems. For a MIMO
system with n inputs and m outputs, there are mxn 8150
systems in the model. Each MIMO output is formed by summing
the outputs of n 5I50 systems, each with a different MIMO
system input. The input-output relationships of this model
are conveniently represented with a vector equation using an
mxn coefficient matrix of 8150 system transfer functions.
This representation of a MIMO system is illustrated in
Figure 1.1, where
VI
V 2
V m
H i l H I 2
H 21 H 22
H m I · H m 2
H In
H 2n
H mn Un
6
(1. 1)
The frequency domain description of this system is formed by
replacing the transfer functions with the corresponding
frequency response functions. In practice, these may be
obtained by exciting the system inputs one at a time,
measuring each output due to the current input, and
computing the frequency response function from the current
input to each output.
Now, the addition of control loops to this system will
couple the loops: the control loops cannot be designed
independently. However, they can be designed sequentially,
or one-at-a-time. This is the underlying theme of I-CAT.
The procedure is to close one loop, to evaluate the effects
of this closure on the next loop to be closed, and to close
the next loop based on the effects of closing the previous
loop. This process is repeated until all required loops are
closed. with prudent selection of the sequence of loop
closures, high-performance control may be achieved with this
7
step-by-step method. Used with frequency domain design
techniques, I-CAT produces low-order compensators for
high-order systems.
Another design enhancement that is compatible with the
methods outlined above is that of Modal Suppression by Phase
Stabilization (MSPS) [1]. This technique is especially
attractive for suppressing low-frequency flexible modes
encountered in LSS control. "A mode is said to be phase
stabilized if its gain, in the compensated open loop
frequency response, is greater than 1 (0 dB) and no
encirclement of the -1+jO point on the polar frequency
response occurs due to this mode. A mode is perfectly phase
stabilized if the phase of the loop frequency response is
o degrees at its peak frequency" [1J. Modes which are phase
stabilized are stabilized throughout the structure: MSPS is
a global dynamic effect, not merely a local loop phenomenon.
Phase stabilization occurs inherently in certain
hardware configurations. In particUlar, the collocation of
a force or torque actuator with the corresponding rate
sensor forms a "complementary pair" that produces modal
phase stabilization [1]. Thus, neglecting actuator and
sensor dynamics, modes excited and measured in this pair's
loop will be phase stabilized by closure of the loop; and,
as noted above, this stabilization is effective throughout
the structure. Obviously, this is an extremely nice
property to exploit whenever it is possible.
1.4 Summary
As detailed in this introductory chapter, the DTP
strategy encompasses all the salient features desirable in
LSS control. However, its effectiveness is rooted in the
fidelity of the empirical frequency domain model elicited
from the LSS. It is the purpose of this thesis to examine
some techniques of enhancing the sampled-data frequency
response functions that are characteristic of LSS dynamics.
First, in Chapters Two and Three, the theoretical basis
of the experimental methods used later is examined. Next,
an experiment is designed and implemented to explore the
practical application of these methods. Finally, the
experimental findings are analyzed, and suggestions are
offered for the application and further investigation of
these methods.
8
9
CHAPTER 2.
SYSTEM IDENTIFICATION
2.1 Overview
The design of control systems requires sufficient
knowledge of the mathematical relationships that describe
the physical system that is to be controlled. Before a
system is constructed, attempts may be made at predicting or
"modeling" the system's behavior. For simple systems, the
mathematical model may be constructed by applying physical
principles known to govern the system's behavior. More
complex systems may be viewed as a combination of simpler
subsystems, and the same techniques applied. However, as
the order of the system grows, its behavior becomes more
difficult to satisfactorily predict in this manner.
After a system has been constructed, the analysis of its
input and output signals may reveal a wealth of information
concerning the dynamics of the system. "System identifica
tion" is the process of deriving a mathematical model to
describe the system's dynamic behavior from observations of
its input and output signals. As one would expect,
incorporating the system modeling information into the
system identification process often yields a more suitable
system model.
10
There are many questions to be answered before a
suitable system identification method can be prescribed.
What is to be the mathematical structure of the model (e.g.
differential or difference equations, transfer functions,
frequency response functions, etc.)? Is the system linear
or nonlinear; stationary or nonstationary; discrete,
sampled-data, or continuous; single-input or multiple-input;
etc.? What input signals are desirable and permissible to
be employed to excite the system? Proper consideration must
be given to all these questions and more, in order to
successfully identify a high-fidelity system model.
For the LSS control design procedures outlined in the
introduction, sampled-data frequency response functions are
required. It is assumed that the LSS has linear
characteristics, and excitation signals will be limited to
levels that will not evoke nonlinear behavior of the
actuators, sensors, or structure. Also, the system is
assumed to be stationary, at least during the identification
process. What input signals are desirable? This is the
question to be addressed in the sequel. First though, a
closer examination of sampled-data frequency response system
identification is necessary.
11
2.2 Sampled-data Systems
Historically, frequency response methods of system
identification have enjoyed much success. For continuous
systems, the general approach is to excite the system with
an input signal whose spectral characteristics are known.
The system's output signal due to this excitation is
observed, and its spectral characteristics are determined.
Now, the system's frequency response function may be
estimated from the input and output spectra. The same
general procedure may be extended to sampled-data systems.
A typical SISO sampled-data control system is shown in
Figure 2.1. The control loop is composed of the continuous
system that is to be controlled, an analog-to-digital (A/D)
converter, the digital computer/controller, and a digi
tal-to-analog (D/A) converter. The A/D converter samples
the system's continuous output at regular intervals of T
seconds and provides the digital computer with a number
representative of the system's output at the last sampling
time. This sampling function may be represented diagrammat
ically by an ideal sampler whose input is the system's
output, and whose output is equal to the system's output,
but only exists at the sampling instants. The D/A converter
accepts a number from the computer that represents the input
signal to be applied to the system'and applies the specified
12
Reference SaMpledInput u(t) v(t) Output
DigitalContinuous
--+ COMPuter/ H D/A H -+ AID -r-+-Controller
SysteM
~
Figure~ Typical sampled-data control system.
ReferenceInput
Digital
COMPuter/Controller
u(kT)
ZOH
u(t)
Conti nuousSysteM
SiMpl edOutput
v(kT)
Figure~ Sampled-data control system modeled with idealsampler and zero-order-hold.
13
input signal to the system until the next number in the
sequence arrives. Usually, D/A converters are used that
apply a constant input signal between numbers received from
the computer. This manner of constructing a continuous
signal from a digital sequence is described by the
zero-order-hold (ZOH) function. An equivalent representa-
tion of the sampled-data system is shown in Figure 2.2.
Now, the system to be identified is the sampled-data
system from the input of the ZOH to the output of the
sampler. So the input signal is u(kT), and the output signal
is v(kT), where k is the sample index. The relationship
between the input and output signals of this model may be
described by the discrete-time system transfer function. A
sampled-data signal is transformed to the discrete domain by
using the z-transform. The z-transform is defined as
OQ
F(z) = I f(kT)z-k ,k-O
where the relationship between the z-transform complex
(2 . 1)
variable z and the Laplace transform complex variable s is
z = eST. (2.2)
So the transfer function of the system may be expressed in
terms of the ratio of the input and output signals in the
z-domain as
H(z) = V(z).U(z)
Thus, the sampled-data system to be identified may be
represented as shown in Figure 2.3.
14
(2.3)
Finally, the relationship between the complex variable z
and the real frequency w is
(2.4)
Then, the sampled-data frequency response function for the
system is found to be
This relationship is depicted in Figure 2.4.
(2.5)
The direct computation of the input and output spectra
from empirical data is commonly accomplished using the
Discrete Fourier Transform (DFT) and its efficient
implementation with the Fast Fourier Transform (FFT). These
topics are reviewed in the next section.
2.3 Fourier Transforms
The spectra for signals in continuous control systems
are determined by applying the Fourier Transform. Similar-
ly, the spectra of sampled signals may be obtained by
U(z) ,.. H(z) U(z)
15
Figure~ Sampled-data system transfer function.
J~TH(e ) ".
Figure~ Sampled-data system frequency response function.
16
applying the Discrete Fourier Transform; and since signals
are sampled for a finite time period, or for a finite number
of samples, the finite DFT is used. The finite DFT is
defined as
N-l
F(e i GO T )=I: /CkT)e- i w k T•
k-O
(2 • 6)
Note that the spectra of sampled signals are periodic with a
period equal to the sampling frequency of
2rrw s = - ·
T
In order to implement DFTs efficiently, computer
algorithms termed Fast Fourier Transforms have been
(2.7)
developed. The effectiveness of FFTs has been a real boon
to digital spectral analysis. However, there are two
problems which arise in deriving the spectra of sampled-data
signals with FFTs. These are 1) aliasing and 2) leakage.
Both phenomena must be considered when computing sampled-da-
ta frequency spectra and are explained next.
Aliasing occurs when the sampled signal contains
frequency components that are higher than one-half the
sampling frequency. Instead of appearing in the spectrum at
the proper frequencies, these components are "folded back"
about the one-half sample frequency and added to the
17
spectral components of these lower frequencies. The obvious
solution is to sample the signal at a frequency at least
twice that of the highest component frequency. However, the
spectral content of the signal is not always well-known, and
the noise present in the signal is likely to have frequency
content that can cause aliasing. Thus, choosing a
conservatively high sampling frequency is advisable, if one
has this option.
Leakage occurs if the sampled signal is aperiodic, or if
the sampling window is not precisely an integral mUltiple of
the signal's period. The finite nature of the sampled
signal may be viewed as mUltiplying an infinitely long
signal with a unity-magnitude rectangular window of the same
length as the sampled signal. The frequency domain
representation of this rectangular window contains signifi
cant "side-lobes" over frequencies adjacent to the desired
frequency. Therefore, the resultant spectrum obtained by
combining these signals in the frequency domain has coarser
resolution than would be expected, due to the spreading of
spectral information over adjacent frequencies. The usual
precaution for leakage is to use a window that is shaped to
make the time signal appear to be periodic in the window:
this suppresses the side-lobes in the frequency domain.
18
Typical time windows taper the signal to zero at each end of
the time window. One window particularly popular for system
identification is the Hanning window, which is
2nth(t)= O.5-0.5cos-, for O~t~Tr' and
T r
h(t) = 0, elsewhere,
where T r is the sampled record duration.
(2.8)
NOTE: For the remainder of this thesis in the interest
of notational simplification, (!) will be used to represent
the dependence of a signal's representation or of a function
on the complex variable exp(jwT) with the radian frequency w
replaced by the equivalent expression 2nf. This simplifica-
tion is summarized as
(2.9)
2.4 Deterministic Analysis
Input signals used for exciting systems during the
identification process may be divided into the classes of
deterministic signals and stochastic, or random, signals.
The analysis of systems differs for these two classes of
signals, and each is considered separately in this and the
succeeding section.
19
For the ideal sampled-data system outlined previously,
the frequency response function is simply
H(/) = V(/).U(!)
(2 • 10)
However, real systems must be modeled by adding a noise
signal to the measured output signal. This is necessary
because real systems and measurements differ somewhat from
the assumed ideal systems. The real system may have slight
nonlinearities, a small degree of nonstationarity, minor
disturbances, inherent sensor noise, etc. All variations
due to these nonideal characteristics are modeled as a noise
signal spectrum Nef) added to the ideal output signal
spectrum Vef) to form the measured output signal spectrum
yef). The system block diagram now is illustrated by Figure
2.5, where
Y(f}=V(f)+N(f)· ( 2 • 11)
Now, for a deterministic input signal, an estimated
frequency response function, Ref), may be calculated from
the input signal and the measured output signal by noting
that
f! _Y(/)(I) - U(!)'
(2 . 12 )
U(f)
H(f)U<f)
H(f)
20
Figure~ System with noise modeled at output.
N<f)
X(f) U(f)
H(f)
U<f)+
~< f)
Figure~ Unity negative feedback configuration.
R(f)=V(f)+N(f) andVe!) ,
fI(f)=H(f)+N(f),Ve!)
21
(2.13)
(2 • 14 )
Note the important fact that the fidelity of the estimate is
improved by increasing the relative amount of the ideal
output signal in the measured output signal, or in other
words, by increasing the signal-to-noise ratio at the
system's output. This observation is the basis of enhancing
frequency response functions by prudent selection of
deterministic input signals as detailed in Chapter Three.
Additionally, if the noise may be characterized as zero-mean
and ergodic, then the random error of an averaged magnitude
estimate, obtained by averaging nd estimates, is reduced by
a factor of l/~ [3]. This fact offers another tool for
empirical frequency response function enhancement, and its
practical effectiveness is demonstrated in the experimental
section of this thesis.
A final system configuration that needs to be examined
is that of identifying a frequency response function that is
embedded in a feedback loop. For simplicity, unity negative
feedback is assumed as depicted in Figure 2.6.
22
with knowledge of the input spectrum X(f) and the measured
output spectrum Y(f), the control error signal spectrum U(/)
is
U(·f) = X(f) - Y(f)·
In view of this relationship, the frequency response
(2 • 15)
function may be estimated using the formula developed for
the open-loop configuration, i.e. Equation 2.12.
2.5 Stochastic Analysis
The empirical analysis of systems excited by stochastic
input signals is considered in this section. Often, the
input signal is actually pseudo-random, but its statistics
are exemplary of a stochastic signal. The spectral
input-output relationships of this analysis are in terms of
one-sided spectral density functions. In particular, for
the open-loop system with output noise which is uncorrelated
with the input signal, the optimal frequency response
function estimate, in the least squares sense, is found to
be
F! ( /) = ~ u.y ( /) ,
G uuCf)
where Guy(f) is the input-output cross-spectral density
function estimate, and GuuC!) is the input auto-spectral
(2.16)
23
density function estimate [4J. In this case, the random
error of the magnitude estimate is reduced by a factor of
1/~2nd through averaging nd independent data records [3J.
For the unity negative feedback configuration, the
following relationship yields the frequency response
function estimate of the open-loop transfer function [3J
2.6 Summary
(2.17)
In this chapter, the theoretical foundations for the
methods of analysis used in the remainder of this thesis
have been reviewed. The process of system identification by
sampled-data frequency response function determination was
outlined, and the differences encountered in deterministic
and stochastic excitation methods were considered. In the
ensuing chapter, the spectral properties of several
deterministic signals are investigated to develop a
rationale for improving frequency response function
estimates using these signals.
24
CHAPTER 3.
EXCITATION ENHANCEMENT
3.1 Signal Properties
In this chapter some properties which enhance a
deterministic signal's efficacy as the excitation signal for
frequency response system identification are considered.
Then some signals are examined that offer the opportunity of
tailoring the choice of input excitation to meet constraints
of the specific system being identified. Also, comments are
included concerning the applicability of each signal.
Recall from the development of deterministic analysis in
section 2.4 that
H(!) = Y(!) andU(!) ,
Y(f)=V(f)+N(f)·
This leads to the important results that
H(!)=V(!)+N(!) andV(/) ,
H(!)=H(!)+N(!).U(f)
(3 • 1)
(3.2)
(3.3)
(3.4)
25
Equation 3.4 clearly characterizes the effect of the input
signal spectrum on the estimated frequency response function
magnitude: increasing the magnitude of the input spectrum
over frequencies of interest improves the estimate HCf).
The signals considered for use in the sequel are finite
duration sequences that are non-zero for only a portion of
the identification time window. The start of the
identification window is coincident with the first non-zero
element of the sequence, and the input sequence will be
permitted to assume non-zero values for M elements, or
sample periods. Thus, the signals of interest may be
expressed by their z-transforms as
M-I
U(z) = I u(kT)z-k.k-O
Furthermore, by invoking the relationship z=exp(jwT) the
spectrum of the signal is shown to be
M-l
U(e i GlJ T) = I u(kT)e- i GlJ kT
•k-O
(3 .5)
(3 .6)
Note that Equation 3.6 is the finite DFT of the signal as
introduced in section 2.3. The foregoing relationships will
be used to investigate the signal sequences proposed in this
chapter.
26
A closer look at the relationship z=exp(jwT) will provide
some needed insight into the signals presented. In the
z-plane this relationship describes a unit circle centered
at z=o. Real frequencies of interest are restricted to the
range of zero to the half-sampling rate, or O~W ~ws/2. This
describes the half of the unit circle that begins at z=l, or
w=o, and proceeds in the counter-clockwise direction to
z=-l, or w=w s/2. This relationship helps to reconcile the
two signal representations above, and it is graphically
depicted in Figure 3.1.
3.2 unit Pulse
An obvious choice for launching this signal investiga
tion is the unit digital pulse. The unit digital pulse is
defined by
u(kT)= 1, for k=O, and
u( kT) = 0, otherwise.
(3.7)
From Equation 3.6, the spectrum of the unit digital pulse is
found to be
U( e}WT) = u(O) = 1 . (3.8)
The unit digital pulse's characteristics are graphically
displayed in Figure 3.2. A shorthand notation will
occasionally be used in the sequel to denote the peak
Uni t Digi tal Pulse (1 xl)
Sequence
28
10
8
6
4OJ'0 2::::l+J
t: 0t::n~ -2:2
-4
-6
-8
• - = - - - - =- - - - - - -
•Input
-10o 2 4 6 8 10 12 14
40
20
o
-20
-40
-60
-80
-100
Sample Index k
Uni t Digi tal Pulse (1 xl)
Spectrum
111111
Input
1/16 1/8 1/4 1/2
Figure .l..:..L..
Fraction of Sampling Frequency
Unit digital pulse's characteristics.
29
magnitude and duration in sampling periods of an input
signal. The form of this notation is (peak magnitude x
duration), which for the unit pulse is (lx1) as shown in
Figure 3.2.
The most convenient result of Equation 3.8 coupled with
Equation 3.1 shows that the frequency response function
estimate is simply the measured output spectrum
Hef) = yet)· (3.9)
Two advantages of unit digital pulse excitation are the
simplicity of computation and uniformity of magnitude across
the spectrum of interest. Additionally, this signal and
estimated frequency response function provide convenient
references for comparison with other excitation methods.
The unity amplitude of this pulse is assumed to be
sufficiently small so that no structural response
constraints are violated, e.g. actuator saturation,
non-linear structural behavior, etc. However, the unity
spectral magnitude does not enhance the estimated frequency
response function by suppressing the noise component as
suggested by Equation 3.4.
30
3.3 Increasing Magnitude
Suppose the magnitude of an input signal is increased by
the factor a. That is
ua(kT) = au(kT). (3 . 10)
since the DFT is a linear transformation, the effect of this
factor on the signal's spectrum is
V a(/) = aV(!).
So application of Equation 3.4 yields
H(!)=H(!)+ N(!).aUe/)
(3 • 11)
(3 • 12)
In the particular case of u(kT) being the unit digital pulse
H(!) = H(!)+ N(!).a
(3 • 13 )
This equation shows the effect of increasing the amplitude
of the unit pulse excitation by a factor of a. The
resulting input signal is noted in the shorthand introduced
as (aXl) , and its characteristics are shown in Figure 3.3
for the signal (lOxl). Clearly, the estimated frequency
response function is improved by increasing the magnitude of
the input.
Increased Magnitude Pulse (lOx 1)
Sequence
31
10
8
6
4
2
o
-2
-4
-6
-8
-
- - - - - - - - -- - - - - - - -
•Input
-10o 2 4 6 8 10 12 14
Sample Index k
Increased Magni tude Pulse (lOx 1)
Spectrum
40
20
co 0'0
ca,) -20"0:;:j
1-1
c: -400)
<0
2 -60
-80
-100
1/16 1/8 1/4 1/2
Input
Figure~
Fraction of Sampling Frequency
Increased magnitude pulse's characteristics.
32
This section demonstrates the possibility of enhancing
the system identification results by increasing the
excitation magnitude. However, as alluded to in the
previous section, there are limits on the signal's
magnitude; and hence, there are limits on the improvement
possible in this manner. On the other hand, the improvement
is still uniform across the spectrum, which is quite
desirable.
3.4 Increasing Duration
In this section the effects of increasing the duration
of a constant amplitude pulse are considered. A constant
amplitude pulse that extends over M sample periods has the
z-transform
U-l'\ -k
Uu(z)= L az ·k-O
The expansion of the summation yields
U u(z)= a[l +Z-l + .•• +Z-CU-2)+Z-CU-I)].
(3 • 14 )
(3 . 15 )
Now multiplying the summation by the unity factor of ZU-l/ZM-l
gives
(3.16)
33
The effect of extending the duration of this pulse at
low frequencies may be shown by evaluating Equation 3.16 for
z=l, which corresponds to w=o. This indicates that UMCl)=aM
, so the low-frequency magnitude spectrum is increased by a
factor equal to the number of sampling periods over which
the pulse extends. Hence, increasing the pulse duration
increases the input signal's magnitude spectrum in this
region and enhances the estimated frequency response
function in the manner explained previously.
Further consideration of Equation 3.16 is necessary to
determine the effects at higher frequencies. This represen
tation of the signal points out that nulls may occur in the
signal's spectrum if roots of the numerator lie on the unit
semi-circle of interest. Note that the roots of the
denominator are all at z=o, and these do not affect the real
frequency spectrum. The roots of the numerator may be
investigated by simplifying the expression by mUltiplying it
by (z-l) as suggested in [5], which adds a root at z=l that
may be discarded later. This analysis proceeds as follows:
zA/-l =0, and
(3.17)
(3 • 18)
(3.19)
34
Now, the frequencies of the spectral nulls may be determined
by sUbstituting the expression z=exp(jwT) and solving for
frequencies that are in the range 0 s w ~ w .r: as
(. T)Me 1 W = 1, or
e}wMT = 1 .
Equation 3.21 is satisfied for all w that satisfy
wi'vlT=2nrr , n an integer, or
2nllW=--.
u r
(3.20)
(3 • 21)
(3.22)
(3.23)
By sUbstitution of the relationship T=2nlw s , the equation
becomes
nw sW=--.
M
(3.24)
Application of the constraints on w leads to constraints on
n of
w ~ 0 =? n ~ 0, and
ui , Mw~-~n~-.
2 2
(3.25)
(3.26)
35
Note that n = 0 =+ w = 0 =+ z = 1 and corresponds to the root added
earlier,and discarded now. Therefore, the admissible
numerator roots correspond to the values of n
Mn= 1,2, ... ,-, for M even, and
2
M-1n= 1,2, ... ,--, for M odd.
2
(3.27)
The values of n listed in Equation 3.27 in conjunction
with Equation 3.24 identify frequencies at which magnitude
nulls appear in the signal's spectrum. These nulls prohibit
identification of the frequency response function at these
frequencies. However, with proper precautions, signals of
constant amplitude and extending over more than one sampling
period may be used in certain circumstances. An example of
this type of signal is illustrated in Figure 3.4 for a pulse
of (lxlO).
First, observe that the lowest frequency at which a null
occurs is for ri = 1, or w = w sl M • So for frequencies
sufficiently below this null, the increased duration pulse
may be used successfully. Also, note that for signals of
length M and M+l, the nulls are distinct from each other.
This suggests that using this pair of signals and averaging
the results may successfully identify modes that would
otherwise be masked by the nulls of one signal or the other
[5]. However, for pulses extending over several sampling
Increased Duration Pulse (1 x 10)
Sequence
36
10
8
6
4
2
o
-2
-4
-6
-8
• • .1 • .1 • .1 • .1 • - - -. - - -
•Input
-10o 2 4 6 8 10 12 14
Sample Index k
Increased Duration Pulse (1 x 10)
Spectrum
40
20
co 0"C
C
Q) -20"C::l
~ -40t::0><'0
:2 -60
-80
-100
Input
1/16 1/8 1/4 1/2
Figure~
Fraction of Sampling Frequency
Increased duration pUlse's characteristics.
37
periods, the nulls cause much difficulty as the frequency
approaches the half-sampling rate. Of course, the same
considerations of structural limitations must be considered
as were in the increased magnitude case of section 3.3.
Finally, note that a combination of increasing magnitude and
duration may better fit specific system constraints.
3.5 Shaping Pulses
The drawbacks associated with the nulls encountered in
the previous section may be overcome to a large extent by
shaping the input signal. The strategy is to avoid nulls
below the half-sampling rate. Consider a signal with the
z-transform representation [5]
(Z+l)M-lUu(z)= M-l •
Z
Examination of Equation 3.28 shows that nulls on the
(3.28)
semi-circle of interest occur only at z = - 1 , or w = w 5/2,
which is the half-sampling rate. This will avoid much of
the problem associated with using a protracted constant
amplitude pulse. Further analysis of Equation 3.28 will
reveal the input sequence that has this z-transform.
First, expansion of the numerator and expression of the
fraction as a series yield the following:
M- 1 b M-2 b 1Z + M-2Z + ... + lZ+UM(z)= M-l ' and
Z
- 1Uu(Z)= 1+b M - 2 z +
b -(U-2) -(M- 1)+ lZ +Z ·
38
(3.29)
(3.30)
So the sequence desired is shown in the expression of the
z-transform
(3 • 31)
Note that this sequence is the ordered coefficients of the
binomial expansion of (z+ 1) using (M - 1) factors in the
expansion. Therefore, this signal will be referred to as a
binomial weighted triangular pulse, or simply a binomial
triangle pulse.
The increased spectral magnitude at low frequencies
offered by this signal may be ascertained from Equation 3.28
evaluated at z=l, or w=o, which gives
(3.32)
This result indicates that each sampling period extension of
this set of signals yields a doubling of the low-frequency
magnitude. Thus the magnitude increases exponentially with
increments in M, in contrast to linear increments in the
case of extending a constant amplitude pulse. However, due
to the increasing peak magnitude of the binomial triangle,
39
these effects may be restricted to a few sampling periods.
If it is desirable to use the binomial excitation over a
longer time, the magnitude of the signal may be scaled so
that system constraints are observed. Of course, this will
reduce the spectral magnitude as well, but the nulls are
still located at the half-sampling rate. This method
provides the designer with some freedom in tailoring the
excitation to the system at hand. A binomial triangle pulse
with peak magnitude normalized to unity and spanning ten
sampling periods (lxlO) is shown in Figure 3.5.
The problem with spectral nulls was diminished
considerably by the introduction of the binomial triangle
signal. However, the null at the half-sampling rate may
still mask system modes at frequencies in this region. This
leads to the consideration of a signal which is
complementary to the binomial triangle signal [6]. Note
that the nulls of the following expression all occur at z=l,
or w = 0/
(z-l)M-lUM(z)= M-l •
Z
(3.33)
The sequence which has this z-transform may be identified in
a like manner to the binomial triangle analysis as
Binomial Triangle Shaped Pulse (1 x 10)
Sequence
40
10
8
6
4
2
o
-2
-4
-6
-8
•• • II • .1- - • ;;;- - - ~-~
•Input
-10o 2 4 6 8 10 12 14
Sample Index k
Binomial Triangle Shaped Pulse (1 x 10)
Spectrum
40
20
o
-20
-40
-60
-80
-100
~-.\\\\
1IIITllfTII
Input
Figure~
1/16 1/8 1/4 1/2
Fraction of Sampling Frequency
Binomial triangle shaped pulse's characteristics.
U ( M-l ( M-2M Z)= Z + -1)b M _ 2 z + ...
41
(3.34)
(3.35)
(3.36)
So in this case, the sequence is seen to be the ordered
coefficients of the binomial expansion of (z- 1) using (1\-1- 1)
factors in the expansion. Note that the elements of this
sequence alternate signs. Therefore, this signal will be
referred to as a bipolar binomial weighted triangular
signal, or simply a bipolar triangle. Additionally, the
cryptic notation used will include a minus sign with the
peak magnitude to distinguish it from the binomial triangle,
i.e. (-lx10) indicates a bipolar triangle with peak
magnitude normalized to unity and spanning ten sample
periods.
The increase in spectral magnitude offered at the
half"-sampling rate is determined by evaluating Equation 3.33
at z=-l, or w=w s / 2 , which yields
(3.37)
This result verifies that the bipolar triangle is
complementary to the binomial triangle. This suggests using
42
these signals as a pair to span a wider range of frequencies
if necessary. A rationale for combining the results
obtained with these two signals is developed next. A
bipolar triangle signal is presented in Figure 3.6 that
corresponds to the binomial triangle signal in Figure 3.5.
since the spectral magnitude of the binomial signal
decreases as frequency increases, and the spectral magnitude
of the bipolar signal increases as frequency increases, it
is reasonable to find the frequency at which these spectra
are equal, and to use the binomial results for lower
frequencies and the bipolar results for higher frequencies.
The boundary frequency may be found by identifying the
frequency on the semi-circle in the z-plane at which the
z-transforms of the two signals have the same magnitude. By
equating the magnitudes of Equations 3.28 and 3.33 the
expression obtained is
(3.38)
In terms of frequency, this expression becomes
(3.39)
The solution of Equation 3.39 for the range of admissible
frequencies yields the boundary frequency of
Bipolar Triangle Shaped Pulse (- 1 x 10)
Sequence
43
10
8
6
4Q)
'0 2='~
0Ct:nco -2:2
-4
-6
-8
•• .1 ••- - - - -- - • • - - - -•
•Input
-10o 2 4 6 8 10 12 14
Sample Index k
Bipolar Triangle Shaped Pulse (- 1 x 10)
Spectrum
40
20
a::l 0'0
....Q) -20"0:::3~
c: -40t:nC'O
:2 -60
-80
-100
-:/
//
//
Input
1/16 1/8 1/4 1/2
Figure bh
Fraction of Sampling Frequency
Bipolar triangle shaped signal's characteristics.
W sW=-.
4
44
(3.40)
Thus, for a given sampling frequency, the system may be
identified by using the binomial excitation signal results
for frequencies below the quarter-sampling frequency, and
using the bipolar excitation signal results for frequencies
between the quarter-sampling frequency and the half-sampling
frequency. Of course, this doubles the identification
effort, but it is a method for consideration in particular
systems. The effective excitation spectrum using this
strategy is obtained by splicing the appropriate spectral
sections of Figures 3.5 and 3.6, and the resulting spectrum
is shown in Figure 3.7.
3.6 Summary
In this chapter several methods of altering determinis-
tic signals to enhance their effectiveness in system
identification applications have been considered. These
methods include altering a signal's magnitude, duration,
shape, and using complementary signal pairs. The method or
combination of methods that will provide the best results
depends primarily upon the system constraints and the
frequency range of interest. The practical aspects of using
these methods are highlighted in the experiment which
follows.
45Shaped Complementary Pulses (1 x 10)
Spliced Spectra
40
20
o
-20
-40
-60
-80
-100
-:~
<;
Input
1/16 1/8 1/4 1/2
Fraction of Sampling Frequency
Figure~ Complementary binomial pair's spectral characteristic.
46
CHAPTER 4.
EXPERIMENT DESIGN
4.1 Objective
A prime objective of this thesis is the experimental
investigation of the effectiveness of the methods of
estimated frequency response function enhancement outlined
in the previous chapter. Since the identification of LSS
systems is the ultimate goal, the experimental model should
display LSS characteristics, i.e. lightly damped flexible
modes. In addition, some of the concepts associated with
DTP, I-CAT, and MSPS, which were introduced in Chapter One,
will be demonstrated in the hypothetical system analysis.
For example, a simple control loop will be employed to
stabilize the system to permit application of an excitation
signal, which is a fundamental step in the DTP procedure.
This control loop will incorporate the property of MSPS in
order to ensure global system stability. Furthermore, the
system selected for the experiment may be viewed as one of
the single-input, mUltiple-output subsystems in Figure 1.1.
Thus, the estimated frequency response function is exemplary
of an element in the matrix in Equation 1.1, which is the
foundation of the 1-CAT methodology.
47
In order to add credence to the simulation, the
continuous structural dynamics will be modeled on an analog
computer, and the stabilizing compensation will be provided
by a digital controller to form a hybrid simulation. The
use of the analog computer introduces limits on system
dynamics that may be viewed as structural constraints; and
these must be carefully observed, as the admonishments of
Chapter Three indicate. In particular, excitation signals
must be scaled so that the system behavior remains linear,
i.e. that the analog computer amplifiers do not saturate.
So it is clear that the ensuing experiment embodies many of
the practicalities encountered in an actual system
identification endeavor.
4.2 Physical Model
As an aid in visualizing the physical concepts
discussed, a hypothetical structure is adopted, which is the
same configuration as that used by Maggard [7J. This
structure consists of three rigid disks joined at their
centers by two massless rods as depicted in Figure 4.1. The
motion of this structure is restricted to rotation about the
axis through the centers of the disks. However, the two
rods exhibit torsional spring constants of k} and k 2 about
this axis, so that the system dynamics include a rigid-body
mode and two flexible modes. In the interest of
49
computational simplification, the mass polar moments of
inertia of the disks are all deemed to be unity. Thus, the
specification of the two torsional spring constants will
determine the frequencies of the two flexible modes. Also,
since LSS modes are very lightly damped, no damping is
modeled in the structure.
A torque actuator and a rotational velocity sensor are
collocated on the first disk. This complementary pair will
be used to form the MSPS control loop to stabilize the
structure during the identification process. Additionally,
a rotational velocity sensor is placed on the third disk,
and the sampled-data frequency response function to be
identified is from the digital torque input to the Disk-3
sampled velocity. A block diagram of the sampled-data
system is presented in Figure 4.2.
4.3 Mathematical Models
The mathematical models presented in this section are in
the form of transfer functions. Since the goal is to
implement these transfer functions efficiently on an analog
computer, they are stated in forms which facilitate
programming the analog computer. The transfer functions
derived here may be transformed to differential equations
that lead to simple analog computer implementations.
... H3
( s)" T
.,.(kT)
II- ZOH Hi (s)...T
50
Figure 4.2. Three disk system sampled-dat~ block diagram.
51
Application of the laws governing rotational motion
yields the differential equations describing the behavior of
the structure. Manipulation of these equations and Laplace
transformation leads to the pertinent transfer functions
(4. 1)
s4+(k 1 +2k 2)S2+k 1k 2
s[s4+2(k 1 +k 2)S2+3k 1k 2 ] '
(4.2)
The determination of the flexible modes' frequencies is
now considered. An analysis of the dynamics of a version of
the proposed Space station indicates that significant
flexible modes will span the frequency range of 0.1 to 1.5
Hz. [1]. So these two extreme frequencies are chosen to be
the flexible modes of this model. These frequencies are
characteristic of the physical model adopted when k 1=O.2636
and k 2 = 44.35 N-m/rad. Finally, the expression of the transfer
functions in terms of the modal components yields
0.3333s
0.3348s 1.488x 10- 3s
-----+S2 + 0.3948 S2 + 88 .83
and(4.3)
0.3333 0.6667s---+-----
s s2+0.3948
5.194XIO- 6s
S2 + 88 .83
(4.4)
In general' terms, Equation 4.3 may be written as
52
(4.5)
where B3 R is the rigid body component of the transfer
function, and B3 L and B3 H are the low- and high-frequency
flexible modes' components, respectively.
Note that the transfer function of the first disk may be
expressed as a linear combination of the modal components of
the third disk transfer function. This permits implementa-
tion of the third disk transfer function on the analog
computer and formation of the first disk transfer function
by using the existing third disk modal components. This
relationship may be expressed as
(4 • 6)
4.4 Digital simulation
As a precursor to the hybrid simulation suggested in
this chapter, a digital simulation of the experiment was
conducted. Several benefits accrue from this tack. First,
the analog computer realization may be modeled, and the
range of values assumed by program variables under various
experimental conditions may be readily ascertained. This
information is necessary to "magnitude-scale" the analog
computer variables and also provides a basis for
53
establishing the "structural constraints." Also, the
effects of changing parameters of the identification
process, e.g. sampling period or time record window, may be
quickly evaluated. In addition,· the results of the digital
simulation identification process provide reference data to
validate the implementation of the hybrid experiment.
The digital simulation was created using the MatrixX
control system analysis and design software on an IBM PC-XT
microcomputer. The MatrixX programs written to effect the
investigation are presented in Appendix A. As a first step,
the CALCFRSP program was developed to display the continuous
and sampled-data frequency response functions corresponding
to the Disk-3 velocity transfer function. Next, the RTLOCUS
program was instituted to demonstrate the stabilizing
effects of closing the Disk-l loop in either the continuous
or sampled-data domains. Finally, several subprograms were
written and combined to provide an interactive system
identification program titled SYSTEMID.
In order to generate sampled-data results, a sampling
period must be chosen. According to the sampling theorem,
the sampling rate must be at least twice the highest
frequency of interest in the sampled signal [4J. Since
there is a mode of 1.5 Hz. in the structure, a minimum
sampling rate of 3 Hz. is required. However, to provide
54
some tolerance for modal uncertainty and to abate the
aliasing of noise, as described in Chapter Two, a somewhat
higher sampling rate is necessary. On the other hand, a
higher rate increases the number of data points collected
for a fixed time record window, and thereby complicates the
data collection and analysis. In view of the foregoing, a
sampling rate of 5 Hz. was selected for the experiment.
Using this rate, the CALCFRSP program produces the ideal
Disk-3 velocity sampled-data frequency response function
shown in Figure 4.3.
The RTLOCUS program provides some insight into the MSPS
stabilizing effects due to closure of the Disk-l loop.
First, for the continuous system, Figure 4.4 shows the root
locus. Note that the single pole at the origin will migrate
to the left along the real axis with increasing gain. The
four poles on the imaginary axis move toward the adjacent
zeros on this axis. Closer examination of a typical path
taken by these poles is provided by the magnified view of
Figure 4.5. This shows that the entire locus for the 0.1
Hz. mode is confined to the left-half plane for all
closed-loop gains. Thus, the global effect of MSPS inherent
in the complementary actuator-sensor pair configuration is
graphically confirmed.
4e
20
e
-20
-40
-60
-80
-100
55
~
...~
- ~\~
c-~.... ""~~ -
---... -.~
~
"~ ~- ~,
~ '"N }~~~~~
"-
--
I
-~~~
....
• 1
FREQUENCY IN HZ
1
Figure~ Three disk system sampled-data frequencyresponse function.
1~
12
9
6
-3
-9
-12
-1~
-15 -10 -5 e
REAL
5 10
56
15
Figure~ Three disk system continuous root locus.
58
Now, the feedback compensator gain used to stabilize the
modes is selected by investigation of the sampled-data root
locus. Some care must be taken in choosing this gain in the
sampled-data case. If the gain is too large, MSPS can be
lost in the sampled-data system, and the system becomes
unstable. Also, careless gain choices may unduly suppress
the closed-loop time responses. On the other hand, gains
that are too low may not provide sufficient modal damping
for limiting responses to desirable excitation signals.
Another consideration for deterministic identification is
that it is desirable to have the responses due to the pulse
inputs to settle within the time record window. This
obviates the need for using a tapering window to suppress
side-lobe leakage due to the input responses in the Fourier
analysis. In consideration of these constraints, a feedback
gain of unity was deemed to be sufficient, and the resulting
sampled-data root locus generated by the RTLOCUS program is
shown in Figure 4.6. Note that all the roots for this gain
are within the unit circle, so system stability is assured
in the unity negative feedback configuration, which was
analyzed in Chapter Two.
The system identification program SYSTEMID was designed
to permit the user to have much flexibility in the
identification process. The program provides the capability
to specify any of the deterministic signals discussed in
60
Chapter Three and stochasti~ identification is possible
using pseudo-random white noise. Also, the user may alter
the sampling rate and the time record window. However, due
to memory constraints, the number of data samples are
limited to 256 for each record, and averaging of mUltiple
records is not implemented. Additionally, no noise is added
to the output "measurements", so the results are idealized.
4.5 Computations
At this point, a few comments are in order concerning
the computation of the Disk-3 velocity open-loop frequency
response function estimate. The analysis methods of Chapter
Two are sufficient to find the oisk-l response function
estimate, if this is desired. with reference to the system
block diagram shown in Figure 4.7, the oisk-3 response
function may be estimated by noting the following
relationships for deterministic excitations:
T(f)=X(f)-8 1(f), and
(4.7)
(4.8)
(4.9)
N1
(£)
1+X(£) ,.(£) 81
( £)
HI (f) t
Figure~ Three disk system sampled-data frequencyresponse block diagram.
61
62
Equation 4.9 shows the computations necessary for
determining the estimate in terms of known or measurable
signals' spectra.
A parallel analysis for stochastic excitations with the
output noises uncorrelated to the input signal yields the
following computations:
(4 • 10)
( 4 • 11)
( 4 • 12 )
In this case, Equation 4.12 shows the necessary computations
in terms of spectral density estimates.
4.6 Digital Results
In this section, the results of the digital simulation
identification process are presented for each of the
excitation signals introduced. Here, only the magnitudes of
the frequency response functions are computed and displayed.
However, this lays the groundwork for a more extensive
investigation in the hybrid experiment.
63
The first results presented in Figure 4.8 are due to
unit digital pulse excitation. The flexible modes are
identified clearly at the expected frequencies in accordance
with Figure 4.3. The noticeable spreading of the 0.1 Hz.
peak occurs because the spectral resolution at this
frequency is less than desirable. Since the digital
experiment is limited to 256 data samples per record at a
sampling period of 0.2 seconds, the record time window is
51.2 seconds. This implies a spectral resolution of about
0.02 Hz. which is a significant amount at the 0.1 Hz.
frequency. The corresponding analysis for the hybrid
experiment shows that the resolution will be increased
fourfold and will not be nearly so objectionable.
Next, the results for an increased magnitude pulse are
shown in Figure 4.9. Since no noise was added to the
measurements for the digital simulation, this response is
essentially the same as the response for the unit digital
pulse. Later in the hybrid experiment, the advantages of
this signal in the presence of noise will be demonstrated.
The response function for the increased duration pulse
is displayed in Figure 4.10. As anticipated, satisfactory
results are achieved at low frequencies; but at frequencies
approaching the half-sample frequency, the spectral nulls of
the excitation introduce aberrations into the identified
30
20
10
o
-10
-20
-30
-40
-50
-60
-70
-80
64
l-
•--...
i
-- ~
~l-
I-
I~
~------..If
\~. ~I- ~ r--~•I- \l-
•I-
- \l-
I-
l-
I-
~..~
'-
I- \~
~\I-
~
.. \...~
I- '",.. [\.... )\.. ~....
" ~.... ......,/
\....l-
I-
I- \'- ~~..
• 1
FREQUENCY IN HZ
1
Figure~ Frequency response function obtained usingunit digital pulse.
30
20
10
o
-10
-20
-30
-40
-50
-60
-70
-80
65
l-
I-
l-
I-f
I- ~
\I-
~
/I-
t--~II
\, ~"""" ~ ~....... I---..
- \I-~
l-
I- \l-
I-
l-
I-
~l-
•I-
- \I-
~\l-
I-
I- 1\I-
~
I- \""""
r\, }\l-
I- ~I- i"\~
I-1---""";
\l-
I-
l-
I- '\I- ~I--
• 1
FREQUENCY IN HZ
1
Figure~ Frequency response function obtained usingincreased magnitude pulse.
:sa
20
10
-10
-20
-30
-40
-50
-60
-70
-80
66
I-
~
l-
I-~
- I- \l-
I....
~f
\~ ~~ ....-. :...--- ~
- \---- \I-..-I-
1\---~ \...
\-..- \- \-~
.. \-J
.. \ t.. -,~ \ V
\J..--- \J
- V~..• 1
FREQUENCY IN HZ
1
Figure 4.10. Frequency response function obtained usingincreased duration pulse.
67
frequency response function. If modes need to be identified
at these frequencies, e.g. the 1.5 Hz. mode, a different
excitation signal is indicated.
For the binomial triangle excitation results presented
in Figure 4.11, the low frequency fidelity is again
apparent. Also, the peak at 1.5 Hz. may be attributed to
the response function since the excitation spectrum does not
have any nulls below 2.5 Hz. However, this mode is on the
borderline for confident identification.
In Figure 4.12 for the bipolar triangle excitation, the
low frequency results are erroneous, but the high frequency
mode is clearly defined. Again, this suggests using these
shaped pulses in tandem for broad-spectrum identification
applications. This is another point that is developed in
the hybrid experiment.
Finally, Figure 4.13 shows the results of the digital
experiment for a single record stochastic identification.
Due to the small number of samples in the record and not
averaging several results, the response function is not
reliable for the identification of either mode. However,
there is considerable evidence of the 1.5 Hz. mode. A more
realistic application of stochastic identification is
provided in the following chapter.
30
29
10
e
-19
! -20
-30
-40
-50
-70
68
~
~
lIIIO..II
I- 1• \I- jI-
~f
\- ,---- ",..
.",..;
- .......
~ \ r-:I---lIIIO \I- J~
~
~
\ j"""-~
.. '\
I- '\~
~
I-
~ ~l-
I-
~ \~ i\..I- " JI- r-,
~
.... '"'J..
...-
• 1
FREQUENCY IN HZ
1
Figure 4.11. Frequency response function obtained usingbinomial triangle shaped pulse.
30
20
10
e
-10
! -20
-30
-40
-60
-70
70
l-
t-l-
II- r\~ v \=/ \v
~ ~lI-
~
I-
~ 1\ ~ \ J\I-
~ ~~~
l- V'~
I-
~ ~-l- Vy f~
I-
~~
I- ~I-
~~
~l-I
I- IJ ~I Jl-
I- ~... , ~ ~il I ~i- ~r ~~l-
I-
~
I-
"'""l-
I-
• 1
FREQUENCY IN HZ
1
Figure 4.13. Frequency response function obtained usingunity white noise.
60
49
20
o
-20
-40
-60
-80
-100
-120
69
--- -~'\
I- \..~..
~ \.. \~ ,..\....
...
.- \~
~..\
0- ~~
-..~
\~I-
-,\-..
l-
I-..
• 1
FREQUENCY IN HZ
1
Figure 4.12. Frequency response function obtained usingbipolar triangle shaped pulse.
71
CHAPTER 5.
EXPERIMENT IMPLEMENTATION
5.1 Hybrid Experiment
The hybrid experiment hardware configuration is shown in
Figure 5.1. The continuous system dynamics are modeled on
the EAI Analog Computer Model TR-20. The digital controller
is a Hewlett-Packard (HP) HP-3852A Data Acquisition/Control
unit (DACU). The D/A converter is an HP-44727A Digital to
Analog Voltage Converter, and the A/D converter is an
HP-44702A High Speed Voltmeter. Both the converters are
option modules installed in the HP-3852A DACU. Finally, the
digital computer is an HP Series 9000 Model 300 computer
which is linked to the DACU via the HPIB parallel bus
(IEEE-488) .
The hybrid experiment system identification is accom
plished in two steps. First, the real-time simulation is
executed to elicit time responses due to the desired
excitation signal. During this part of the process, the
DACU drives the system and logs the necessary data, while
the digital computer generates a video display "strip chart"
of the first twenty seconds of the simulation. After the
required number of time records are collected and stored on
diskette, the second part of the identification process,
Digital COMputer
II
IIII
x(kT) I HPIBI
I
I
I
I
81(t),I,
,.(kT) 1'(t) 83(t)
Digital Controller "'- D/A .. Analog' COMPuterp:-
(HP-44?2?A)
(HP-3852A) (EA I TR-20)
81(kT) ,
83(kT)
AID
(HP-44?02A)
Figure 5.1. Hybrid experiment hardware configuration.
72
73
frequency analysis and response function computation, may be
executed. The hybrid experiment incorporates all the input
signals that the digital simulation accommodated. In
addition, averaging of results for several records is
possible in this implementation. Also, there is sufficient
noise inherent in the sampled-data system to demonstrate the
fidelity differences in response function estimates. The
function of each of the major hardware components and its
associated program(s) are summarized next.
5.2 Analog Computer
The oisk-3 and Disk-l velocities' transfer functions are
programmed on the TR-20 analog computer as mentioned in
Chapter Four. An investigation of the analog computer
realization was conducted with the digital simulation to
find the range for each differential equation variable. A
pseudo-random gaussian sequence of zero mean and unity
variance was used as the excitation signal. Based on the
results of this examination, a magnitude-scaled program was
developed for the analog computer. The program is
represented in functional block form in Figure 5.2 using
integrators, summers/inverters, and potentiometers [8]. The
variables shown in this figure are normalized "x" variables
corresponding to the SUbscript variable. Also, the
nUmbering of the various components in this figure
X'T
74
SUJ"'IMer/Inverter
LEGENDn : ta9' nUMberg : gaIn
oPotentioMeter Integ-rator
Figure~ Analog computer simulation functional blockdiagram.
75
corresponds to the actual hardware tag numbering. The
potentiometers provide magnitude-scaling of the differential
equations' variables, feedback gains, and scaling for linear
combinations of modal components. The resultant values
determined for the potentiometers are recorded in Table 5.1.
Analog Computer Potentiometer settings
Pot. Setting Pot. Setting Pot. Setting
1 none 7 0.398 13 1.000
2 0.100 8 0.001 14 1.000
3 1.000 9 1.000 15 0.200
4 0.079 10 0.224 16 0.200
5 0.888 11 0.500 17 0.002
6 0.200 12 0.500 18 0.007
Table 5.1. Analog computer potentiometer settings.
5.3 Data Acquisition/Control unit
The DACU is loaded with a data acquisition/control
subroutine and the specific excitation sequence by the
digital computer. During the execution of the subroutine,
the DACU performs the following functions:
76
* Reads and stores the system outputs via the A/D.
* Computes the closed-loop error signal and sends it
to the system via the D/A.
* Sends the first twenty seconds of data to the
digital computer via the HPIB for strip chart
display.
since the DACU program is downloaded from the digital
computer, it is part of the SIMULATION program in the hybrid
experiment software listed in Appendix B.
5.4 Digital Computer
There are three programs listed in Appendix B that are
used by the digital computer. The first is SIMULATION,
which is an interactive program to drive the time data
acquisition part of the identification process. The second
is SIUDI (System Identification Using Deterministic Inputs),
which does the frequency domain analysis, frequency response
function computations, and averaging for deterministic
excitations. The third is SIUSI (System Identification
Using Stochastic Inputs), which performs the corresponding
functions for stochastic excitation analysis.
The SIMULATION program performs the following major
functions:
77
* Prompts user to specify the excitation signal and
number of data records to add to the storage
archive.
* Downloads subroutine and excitation signal to DACU.
* Displays strip chart of time data.
* Retrieves time data records from DACU and stores
them in the archive.
The hardware and software outlined above were used to
evaluate the effectiveness of the signals discussed in
Chapter Three for identifying this hybrid system. The
results of this investigation are considered next.
5.5 Hybrid Results
The results of the hybrid experiment presented in this
section are divided into two parts. First, the signal set
used in the digital experiment is evaluated in this case
with the addition of phase characteristics. Then, a group
of signals is selected whose characteristics are desirable
for this specific system. An extensive identification
process is conducted with these signals which includes
averaging multiple results. Throughout this section,
graphical analysis is again employed to judge the relative
effectiveness of the various' methods considered.
78
First, Figure 5.3 displays the unit digital pulse's
characteristics, and Figure 5.4 presents the identified
frequency response function. Both the magnitude and phase
functions are very noisy and contain little information.
The 1.5 Hz. mode's magnitude peak is discernible, but on the
whole this response function is very unsatisfactory.
Next, the increased magnitude pulse data is given in
Figures 5.5 and 5.6 respectively. There is remarkable
improvement in the results. The low frequency magnitude and
phase are close to the ideal values, and the high frequency
magnitude peak is clearly identifiable. The slight shift in
frequency of the low frequency mode from the target value of
0.1 Hz. is most likely due to slight inaccuracies inherent
in setting the analog computer potentiometers.
The increased duration pulse results are provided in
Figures 5.7 and 5.8. The low frequency fidelity compares
favorably with that achieved in the increased magnitude
case, but the spectral nulls of the input interfere with
high frequency identification.
Now, the binomial triangle shaped pulse excitation is
considered in Figures 5.9 and 5.10. This signal performs
better than the unit digital pulse at low frequencies, but
it has no merit at high frequencies. The increased
magnitude and increased duration pulses show superior
Unit Digital Pulse (txt)Exc;tation Slgnal Spectrum
79
!························~················_·· ..t···········..·..·..··....·....··1··..···..··· ..··········~· ..·..·· ......·..····_·! ..···..·..···_··....·········..·j················..··--f··..···········..···....1····
: i j + ,;, t ;
. ; ! : : . } :·······..·······..·.. ···1'··· ..···..··········· ..1..· ·..··..·..···..·..··· 1..····..················~..·····..·····..····..· ·~ ···..·····..·····..···· I· .. ·· ..·..·· ..··········i···..· ; .
r + r · f ~ ~ ;·······..···············1'···..···············..·1···········.. ···..······..·..···]'······················1·..·····.. ··············~·· ..·····..··················..··i.. ·················....·1..··.... ·· ..············1'···
+ r f I ~ ! r j························r·············..···..···1·--··········..········_·..·1'····..···_··..····..··1· ·..· ··..······r..··..·..···..·····..······ 1' ·..·..-1"··· ..·····..··· ··..1..··
• • + • .. • ~:::: r::':I ! ~ J ~ I . .·....···..··..·..··..···r·········_·········..r ..······....········......l···..·····....···_··_~· .. ···..···..·..······.. ~..·····..····..·..··..···········r..········..····......r ....·..·········..rt r t • ; • ~ :
-100"------......---------------------'t£IS)
40
20
'300-0
c -20.-0
-0 -40:J+J.-CC"J -60..,I:
-80
Frequency in Hertz ..-Unit Digital Pulse ( Lx L)
Excitation Signal Spectrum180
135
90
(J)11) 45~s,0')ll) e(J
c.--45
d)V)
~
~ -90a..
-135
; i
..........._._ L..._ .t.. _._ _.: _.._ : J_ _..: .1.. Lttl : 1 ~ I :
···· ············ ..·..i..····..·_··· ..·..·..·· ·..·..· ··..· j i..·· ···· .. ···.· ..·.·· ..····.i··.··· ..····· .. ··· ..·..· j .: : : ~ : . ~ :
• + I + ~ • :. : ; : : : : ~...... ·,········· .. ··_···~······· ..················1·········· _~ ~ ~ _ ~ _~ _ ~._.
+ • j. + t. . ~ ~ ~ ~
.... ~.;.
t t t t t t ~ t························r······ ·············~······· · · · · · · ·__··········_··1························4!'························f····_·__·····················~····· ~ ~ .
····· ..· ··········;····..····..·..··..·····f············ ! .;. ···i··..· ····· ..·.._·········~ ..···..·····..···..·····~··· ; .
.- t .;. ~ ~
. : .. . . . . :··· ·.. ··.. ··········;·· ··.. ·····..····..·t············ j ; ·i· ..······· ..····..··..···· ..·· .. ·j························~ j ••..
t t :.; : f t ~ ~~
("\JI"'\J-~-180 '------------------------------...1lJ'")~I~
Frequency In Hertz
Figure~ Spectral characteristics of unit digitalpulse.
Disk-3IJn; t
VelocityDi g; ta 1
Frequency Response tunc.Pulse (txt) Excitation
80
(,\J
i+
~ ~ . . : . i :............. " : ••••••••••••• II" + ~ ..........•.......•.....~•..•.••.••••..•.•....... ~ ..•••.....•..._....•.•........••~..•..•.•......._ ~ ~...•
; . i . : :: 1 1:
··············.. ········~············ ..····.. ··..t..··.. ···..·····················:·········..···..······..r·········· !..t r
..............._ + "1 -·················'j'····..··_-····-f-·--···..··..···1 __ - +·'"j 1 I j
r . r t ...................................................; _ _ ····..·..··..·..·..···,··..· ·.····..···..········.··i····.·· __ .
~ I ~ ~ ~ ~ ~ f. : : ; i : :
r t r r • r····..··..········..····i·..·..··· ·· ···..j""..··· ····· i..···..·················!"··..··..··..·····..··r··..·..··············..··..··..I····..·..·..·· ····t·..··..········..·..··..: .
. . . t
-100 '----.......----------------------........-----........,\[)
~
40
20
aco-0
c -20
w'"C --to:J+>
C0") -6e'"z
-80
Frequency in Hertz
Disk-3IJn i t
Velocity Frequecny Response Func.Ili q t t a l Pu l s e CLx l ) Ex c t t at t o n
.: ......,:
..;.. ...
ti : :c:_.1~. ii:: :i: ii::.. ~;~~ H~~~
: ~;~ i~ ~ ~ ~ rg~~i~!~i ~1~5~iii ii g 1 ~ i :;;iiiii~f i~i~ai~
..· ·..··..·..···..··i..· ····..··..~.1--..::··..·_··..··•·•·· ~ + _ ~.._ ····++~··~·..·~·+·~~·f~~~+·H~~~i·~~• : ..t ~~ : ~~~.;~ : ~ ;:~ii·~~
... ~. .... .... ~ :'1 ' i:: ":f l:-18e I----......-.---""----------------------~
I.J")
~
180
1"iC'oJ.J
90
v.tt1)
~511)
~
mOJ
0I~
c
-451)(/J
It1J: -90a.
-135
Frequency In Hertz
Figure 5.4. Frequency response function using unit digitalpulse.
Magnitude 10 Pulse (10xl)Excitation Signal Spectrum
81
I:'\J
........................! '! ··········.._···············~························1········_·· ..-········i···· ···························~··············,···_····t························1····
; ~ : : : : : :
w::;:;; ~.'::!:.' ~ ~ ~ ~. . r r I . . r························l..···········..·········i ···..······..· ···· ;..··..···..··..···..····r···..······..·..········:·· · ····· ..·········.. ··~········ ..··· ·· ·I· ~ .
r r + r : • t 1···········..·..······-r···················....r··..··········..·····......·····r······....··..·..·..···T···············..·······i..·······..················· ..·..1'..·······......······..7···....·······..····....1'..·
~ 1 l ! ~ ~ i ~
r r 1 t r · r r··..·····.._·..····..··r-···..·········..·····:··············_·_········_)·········_······_··:..·..··_···········_·f········..······················r·······..········..·_·T· _ _ : .
r f r r r · t ;..·········· ········r····..······..··· ··r·..·..· ···..· - _: 1' ·..·..······ 1'..··..·..·······..· 1··· ········ · 1·_·
• r r r r t • •-10e "-----------------------------'
U'1~0'.)
40
20
0CQ"'0
c -20.-Il)
"0 -49:3+>.-c0')
-60,~
~
-S0
Frequency in Hertz
Magnitude 10 Pulse (10xl)Excitation Signal Spectrum
t • • t + • + •......·· ·..···· ·].. ···· · ·f··.. ·· .. ·· .. ·· ..····· ·····..1..·· ·..·· ··..·+····· ··_ .._···.. i · ····..•..·l..···· .. ········_·· t·········· ·.. ·····.. ~· ..·
• t • t • • • ~
........................I. J 1... : _ : L J L..! f ~ t • t • •
····..· ·········..··~·······..·········..·..·t······························ ..~··..··· ··..······..·t ·········..· ! ! ·..······· .. ·..·..·r ····· .. ·· .. ······· ..i.. ··I ' · ~ - I :
.................__ ~ _ ~ ···..····.._····i· ..···..····· ·····f..····· ..···········..··1..· ·..··..·.. ······ ·..·..~············ .. ·········•·· ~ ..
r : · · · r I ·.' ~.- ~ _ ": , ·T········_··············I········_············· ·.. ·1··· "' ~ ~ .
J ! · t f I I i
• ..:
..
I • • •······..····· ..·..·t..···· .. ················~ ....·······.. ······.. ······.... ··~· ..·....·· .. ··· .. ·······~ .. ·· .... ····· .. ······· ..r..·• f •
........................i ~ ··..i·
180
135
90
V)
11) 45l1.)s,0')l1)
0~
c.-
-45/l)(J)
10,- -90a:
-135
-I$)
-180~-------------------------_....-u;I'S)IS)
Frequencv in Hertz ~
~-----------------------------------Figure 5.5. Spectral characteristics of increased magni
tude pulse.
Disk-3 Velocity Frequency Response Func.Magn;tude 10 Pul~e (lax!) Exc;tat;on
82
,.. . .: , : '~A: ' : '20 ,......,.:] 1,....····...·..·r·.._..·..··..·..··.........:..·....,..·..··..r..··....·..·..·..'···r'a
('\J
t t t t t-60 ·_..·_ ·i..· ·· ·~ · · l + +-_ j ..
i : iii ;
t t t t t t .-80 1· ··· t · · l ·..· ·· t..· ·..·..· !..· ·..· ·..·:..· r +.
t t t' ,-10e"'-----------------------------"'"
L(')
~
F~equency In Hertz
D1Sk-3 Velocity Frequency Response Fune.M-a~~.+"~. 11A C •• l .... t"1C\v1't C-v,..;+-a+;o""\""• • ....~ •• 'I v ....... -." .. '-' I " .... ,.. • I ... ,-, "-' • v ... ·." • .. ••
: ;
::Il ; ': j l. f180 r----~---~----~---~---~----__:_---~---~
135
45
I:"\Ju;
-180~ -.l
u-,''::l~
Frequency In Her~z
Figure 5.6. Frequency response function using increasedmagnitude pulse.
Duration 10 Pulse (lx10)Excitation Signal Spectrum
83
40,---~-------.....--~-------~--~--~
-100"----------------------------~Ui
~
c
. . . .
20 r------j':---~;----~j--~:~~.·..~··..~.··;,:,::~.:.-··t:······==.:::·······..·· ..·,····.···················r·····················,.. ··. ~ . . . \' ; ~
o ······················· ·····················r·················..···········....··_···············r····················r··························· ·y~:~:l-·· ",r
-20 : 1.. ·············..···········,·······..···············t···· r-.............................. ...
1 t • t r •-40 ·····················..r:..·········r···························..r······················r·····················T··········· ··..···..·····..···..r·············· . . ! .
.~ t! r t t r f t~ -60 ··· ·_·····..·······f······_·····_-·..·..·r..·····..······ ~ + _ ! · ·..··.._·········· ..·f········· ·······..-f········..··..·..···..··f····!: .!: .j i .i : .J :
!.:. • ~:: I:~ . i i ! . j .-80 ···········..··· ····!..·· ····..···..·····f······..·· ·· ·· ··· ..·f·..··..··..·..···..·····+······..··········..··..f··········· ··_..·· ··..··~······· ..····.. ····..·..+ ··· _ ~ .
~ 1 : ~ : ' ! j+ r t + • ! ! +
Frequency in HertzDur at ion 10 Pu 1 s e (1 x 10 )
Excitation Signal Spectrum180---~---~----------~------------
135
«>(f)
I'd.ca.
I + • I" I +····· ..·..···..·.. ··....!·..·....····· ....·......t··· ..···.. ·····....·..·..··..···:··············..··....··t·..·........·....·..··..i··..·......·..·..·..··· ..·..··..·i· ..·....····..······....7··....·....········..·..;....
: I I ~ · I : r90 ·················r·················:······················..···T············r···············r······················r······t··········--r·
45 .. ···········..·..··..··!·..··············· t· ····..··.._ f _ r _ ~ ~ + ~: r~.,~ + .. + .;:
7 ~ ~ t ~ '. :o ::::·-:::::::::::::::::::+:,::::::::,,:::::,,::,·::::.t::: : f" j _.... .. _. ..ft·..···..· ~... ..; ~ :~ • ~ ".. ' .. f.. + :. :
-45 ··..·..· ·..···j·_· · ·t..·..·..·..······ ·······..··j······· ·· ·..··t· .:z:-··::··.. •·..!··· ..·..····· :, ~ ~ .f ~ ! ~ '~
t • ~ t r I: ;-90 _..···· ..·· ..· ···! ···· ···· ..T..··..··..······ ···· ~ ·..·t··..·..·.. ··· .. ·········:·..··..·.. ~··~·· ······ ·I · , ..• ·· ·.:4 ..
!. ~.: : .
......................+ ~ ·············..i·· ················1·····················1··············· :::1 tt t-
-180lSi - ~"'J
CS\ ~ ~I~
-135
Frequency In Hert~
Figure 5.7. Spectral characteristics of increased durationpulse.
Disk-3 Velocity Frequency Response Fune.Duration 10 Pulse (lx10) Excitation
84
........................! ~ ! + ~ ! + ! .\ : ;. . ~ :
~ r
; ..:
...::I:1;:C:~:::::::::::::I::::::::::::::::::::::::::::::I:::::::::::::::::::::-:::::::::::U'::::::::-::::1 .. ; \ ~
: : '. :
........................: u u..' u u'»K: u u.f + ~ + t·v. ~~IJ J1 ~ i i ~ Jyf'\J.. ····.. ·~······· ..·..t ..····..·..·· t..· ·· ·-r···..··· · f.. ·········..·..·····..·t·..·..·········..·..·· ·· ·t..··..··
i
t : t .... ···.. ·· ..·..·····..···t·..···..····· ..·..··· .. ~·· ······ ·..· j _ + ~ j + j .
-100 '---- ----.1
L(')
i
40
20
aCO"'0
c -20
tl.'1:1 -40:J~
COJ -60Ittz
-80
Frequency in Hertz
Disk-3 Velocity Frequency Response Func.Duration 10 Pulse lx10) Excitation
180
135
90
VlOJ 45l1)
~
01tD aU
c
-45IJj(/.t
m.£ -90c,
-135
(\J-180 '-------------------------------------....
lr')~~
Frequency in Hertz
Figure~ Frequency response function using increasedduration pulse.
Shaped Pulse (lx10)Excitatton Slgnal Spectrum
85
N
i ~:.;:.~. ir · :.......................; + : ~ ; ········..········+·····················r············ +..
··· ····..····..··..r ·····..·····..·1 ··· · ·..· j···· ····..····..··..r·..··..··..····· ···i-··..· ·..·····..·· ···..r·..····_·..·.. ..: .: : : i : : . :It_ _ ~_ + _....•.........•~..............•...._.•.~....•.•......•.•.•......:...••._.........•...••..........:..It" •••••••••••••••••••*!r......... . _~ .. ,.
~ i 1 ~ ! ! ~ .r r 1 r t ! t !
.._ + _ _..+ ! t- ! _ ! +....... . ! .I 1 1 1 i : ~ .
J r r r t : r !
···..·····..············1'..·_···· ····..··1·····..····..·_·····..··..···r·····..· ·········..r..·..···············..··r·········..········..·······_·1'······..·· ···..·····r..···· - j .
+ + + ! ~". + • +! 1 1 1 ; 1 . :..··_···_--····r..........··......·..T...._..·····......···..·..·T....·......·..····....T·....·_..····....·..r·······..··..···..·..·..···..T··..····..·..·........f·..··..···....·......"(..r r t t • ! r h
-10e'---------------------------~i
'40
20
aa.1"'0
c -20.-Il)
-c -40::J~.-cm -69toL
-80
frequency in Hertz
Shaped Pulse (lx10)Excitation Signal Spectrum
180
135
90
f/JIJJ 454)s,0')(1)
0-0
c'- -45ll)(/)
l'O..c -90a...
-135
· • t : t : ,.. •·_·········· ·····..1·····················_·t..················..···..·······j·············.. ·········t······_···_ ·······:··..········..·_····_· ···~···! .\···· ·····~··· ..········..··t·\···l..··
• t • t t t t : +.;, .;, .;, ,;,"" ~ :t~ : ~':. .
...........·············!·····················..·t···.._ ~ ·· ..·················t····.. ···· j ·•· .. ·i···~········ ······\· ..·t·· ·t.. ··~i'···
• I + . t I·' .1;....····..·····..·· : ···..···· ·_·T···..··..··..·..·· ·· ~ ! ······..·· rr..·..··..·····..l ..··..····..·..·7·····t ·
~ t ~ : ::.
Ui
Frequency in Hertz
Figure~ Spectral characteristics of binomial triangleshaped pulse.
Dlsk-3 Velocity Frequency Response Func.Shaped Pulse (lx10) Excitation
86
t·....·......·....····J.·..
U"')
........................~ + ··························r-····..···········-i··············..········,···························..····t···· .1...
::::::::::::::::::::::-::.:::::::::::::':::::::r::::::::::::::::::::::::::.I::::::::::::::::::::::;::::::~::::·::::::::::::~:::I:::::::::::::::]no ..~ ~ ~ f ; l
t + t r i
···..·······........··..t........·....··..·..····t...... ·· ..····..··..······_··..t···....·..···..·..·..···f·..··..··..····....·....t....·....·..· ._.... ..j i i !
t . t . . . ' .•............- -·············_···········4····················_··.···· _ _.......................•- ·..· ···.•..-...· t••••
j i j ~ ~ 1 ; ;: : i :: :.. ...... ..I . i J It.: r...................·····1··_····· f' •••••.! ! _ r: ! _ -~ I •••• - • ..··r···_·····..··· !.. fl
~ ~ ~ ~ ~
f f' t\ i 1
-100 '--__IIlIIi-- ~
i
40
20
aCO'"'0
c -20
Q)
'"'0 -49~
+l'
c0)
-60t'C11:
-80
Frequency in HertzD1Sk-3 Velocity Frequency Response Fune.
Shaped Pulse (lx18) Excitation
: .t~ ... ~ iL n~~ d
•.... ::. .: :::; ::: ii :: ;:; :i:: :: '31
./~ ..! • . • -If .' ;tm~~ ~ln il ~....·..··::····-'· r · · ··~ ·· ..·..· ·r · ··· ..·~ r· ··· ·:!r.. ·..·r·!nnrn r·'f: . ::i+ ~ ;f ~ ~ :
-180 ......------- -..l
\Jj~l~
180
135
99
fI)
IV 45(1)~
UJt[I
0"0
c
-45d)
Chf~
.c -90a..
-135
Frequency In Hertz
Figure 5.10. Frequency response function using binomialtriangle shaped pulse.
87
results; however, it should be noted that their spectral
magnitudes are arbitrarily higher in this case. In the
second group of signals considered in this section, the
relative strengths of the excitation signals are chosen more
equitably.
In the complementary case of the bipolar triangle pulse,
the results are shown in Figures 5.11 and 5.12. Again, at
this excitation level the results show no merit.
The final result from the first group is the stochastic
excitation response function of Figure 5.13. Here the
magnitUde peaks of the two modes are noticeable, but the low
frequency peak is badly spread and the high frequencies are
quite noisy. The phase characteristic is also quite noisy,
but it tracks the expected phase somewhat at low
frequencies.
For the second part of the experiment, the signals were
tailored for use under the constraints of the given system.
Recall that the analog computer realization was magni
tude-scaled based on the results of the digital simulation
excited with zero mean, unity variance gaussian noise.
Therefore, the digital simulation was used to scale the
magnitudes of the excitation signals so the linearity
constraints of the analog computer would be observed.
Furthermore, due to modes of interest near the half-sampling
Shaped Pulse (-lx10)Excitation Signal Spectrum
88
40,...---~--~---~--~--~---~~-----~
20
('\J
-20 ···..· ·····1" ·..··..·..·····+..····..· · ·· ·1..····· ···..···· r ·..·..· ·..·\"· · · ···.. ·····.'1" _ · ·..· 1..
• t r · · r . t-40 ~ _ ~ _.~ ~ ! !.................. ._.! ! .
~ ! ~ ~ ~ ~ I f ~
.~ t t r r r Tit r~ ~: ···..r..· ·r _·r..·_· ·..·..·r..·..·_r..-- · _·· ·~) -1"' 1"'.
..............· r· · r ··· ···· f· ···..·_· ·r · _·..·r ·..·..·..~ · r..·..·..··.-100 '-------------------"-----------~
iFrequency in Hertz
Shaped Pulse (-lx10)Exc;tation Signal Spectrum
• t.. :.: t. ~ t. • *. . . . ! :. ;. ,::...... ·· ~..· ··..·..·· r..·..·..·..·..· T..·..·..··..·..·..·· ..I..· · · ·..r ·· ······ ·..··• ·~· ..··· ~ : ~ .• t t ! • \ ~..............._ ; ~ ·..·..··..····i..·..···..······ ··t·..·····..·····..··· ..··l·· ~ ..:~ ;j .• t t t + .: ~ •
180~--o:___--~---~--~--~------~--~; i+_:.; _:.; ! t ·.i.·· · 1.. ; ~ ~
135 ..·· · ·( · ·..·· t..L\ l _·t · · · 1" · ·..·..1" · ·;..·:1 · ··..[1..
90 1 I··t ····..'. .r :;..: ; .I. ; !..\ :.[.,::..';T •..• ·..··t· ·.......... t .1. :
+/ ...... ~... . t4S ··.. ··.... ·············~ ..··..···....··..····_·ft..······.. ·_··..·_··.. ·_··~·· ..···..····..·..··_··f..···::·':··-:..·..-:--····f·..·_···..·..·..·..·····....····1......····..······~· ....~·····~ ..···.. ···....·1!····
t 1 + t "!- +' : ~o1---- :-"..::::·::·::·..·l · ·..· ·..·..· ;· ··T ·..·..· 1" ,·,:~ · T ·T T ·~ ·..·..·t
-45
-90
(''\,J("\JIi")~
. ; ;:·..··..· ·T ..·..· r · ·· ·T · · T ·..· · T · · ·r ....+· ..! · ·;:::~·..F
-135
-180'-------------------------------~lJ')IS)l~l
Frequency in Hertz
Figure 5.11. Spectral characteristics of bipolar triangleshaped pulse.
Disk-3 Velocity Frequency Response Func.Shaped Pulse (-lx10) Excitatlon
89
40r---~--~---~--~--~---~--~--~
".J
20 , · · i.... ....\ t ; ·· r· ··1'..
: :o . ....· ·T ·..·..··· · r..· \~ ····· T ···..·..· ·..··..·..·..1 · ·· .r..·..· · 1..·
........................i .l : 1 :...; t) 1 i 1
· . t . r . .·•·..···················1"·····················..1································1························1··········..··········..I..···_·..·····················..r· · ! .
. r . r : . . .·..······· ······..1'·····..····· ·······1..·..·..·· ·_········· ·r..·..····..··········..l..···..·····..······..··1'·· 1' ···.. ···..··r..······..············..r···• ••iii ;
• = • ; ; . ~ •...................._ ", _ _ - ········t····..·········· -•............., -•....: i 1 j 1 1 j ~
+ t . r t •-100 "--__"'--__-..-__.......i ----.__---i.__---..J
!
CQ'"C
c -20
~
"'0 -49:J~
CC') -69IU
1:
-80
Frequency in HertzDisk-3 Velocity Frequency Response Func.
Shaped Pulse (-lx10) Excitation180
135
99
"~ 45~
'-0')
o0-0
c
-45Q)(/J
~
.c -90a..
-135
,....J-1813 ~------------------------------------~1.0
~
Frequency 1n Hertz
Figure 5.12. Frequency response function using bipolartriangle shaped pulse.
DiSk-3 Velocity rrequency Response ~unc.
Unity White Noise Excitatton90
40r---~--~---~--~~--~---~--~--~.....
a
: ~
-20 _ ~ ~ ········_········..··1···..···· ·········_+············ ;. . ) _+ ~ ..~ :! j : 1 :
. . . . r t-40 ······ ·T ·· ··· r ···..· ·r ·..· · r ·..· 1".._..·· ..
.. .. . .. ..
! i 1 ! I .-£0 ..·..·..·..· ·..··T·..·_· ·-1"·..·..·_·..· · 1' ·.._·..··..·..·r..··· 1-··..···..·..·· · ·····1..··· .
r r I 1 r . .-80 ! '!' ! _.~ _ ~ _ _ ~ ~...................... ...
1 i 1 j i . j+ t + t t
NIt,)~
-100--------------------------------~If')cgIS)
Frequency in Hertz
Dlsk-3 Velocity Frequency Response Func.Unity ~hlte Noise Excitation
180
135
90
f/Jo
~5Q)
'-rsJQJ
B~
c
-4511)V)
Itti: -90a.
-135
t ;·. : . . ~ ..::: .:::;~ !:If!::
: ~
.......................,t. .J J t J. _ .L. ~.~.fJ.J ~: . . ~; : . ~ ~ j \f j
~1 : .:: :: :....................~ ~ _··,::-··.._· ··_·:· ;..; ···:·;··::;:r· · ·· ~ · ~ ·· ·ljj·lT'..:il· i · · r ·..·..l"' r · ·..r..··_· -r ·..l.rfl'···..ff!1Ii=
t t t~ ~1 ;~ j-180 '------------------------------'
t.(')
~
Frequency In Hertz
Figure 5.13. Frequency response function using unity whitenoise.
91
frequency, the protracted constant amplitude pulse was
omitted. These constraints lead to consideration of the.
following excitation signals:
* Increased magnitude pulse (10x1).
* Binomial triangle shaped pulse (2x10).
* Bipolar triangle shaped pulse (-2x10).
* Unity white gaussian noise N(O,l).
Additionally, the benefits from averaging results are
demonstrated by comparing the average response functions for
one, four, and sixteen records in each case.
First, the results for the increased magnitude pulse are
shown in Figures 5.14, 5.15, and 5.16. The reduction of
noise in the response function estimate due to averaging is
clearly demonstrated in both the magnitude and phase plots,
especially at high frequencies where the signal to noise
ratio is smaller.
Next, the binomial triangle excitation is considered in
Figures 5.17, 5.18, and 5.19. These results improve with
averaging at low frequencies, but the high frequency results
are not usable.
D1Sk-3 Velocity Frequency Response rune.Magnitude 10 Pulse (lext) Excitation
92
·······················f······················r···········..···············1································· >:~!:. ;····..·..·········.. ·· .. ·····l..······················~··· _L.; :....·....·· ..·..........···..················....r·..·····..·..··..·..··········t..··..··....·....···..t..··....·..·· ..···..·..·i..··..·..........·.....r ! .
··········..·· ·· t ······· ····· t..··..····· ..····..·· ···..··i ············ - ..··..·..···· ..·..· ·!..·· ~ .i i .::
r r· ··..· · T..····..· ·· t · · · ··· ·.. '1'..· ·· ·..·..r: · ·'1"· ·· i.. L +..t i' '
40
29
0
£II1)
c -20
dl'7J
-~0:J~
Cm -60ttl
1:
-80
(\J
Frequency In Hertz
Disk-3 Velocity Frequency Response Func.Magnitude 10 Pulse (lexl) Excitation
180
135
90
CI)
l1) 45~s,0)
QJ0-0
c
-45C)(/J
f(j
.c -900...
-135
(\JIf")t\JLfjC;).
('\jIS)
-180 :- ....... ~~..J
l("JIS)tSI
Frequency In Hertz
Figure 5.14. Frequency response function using singleincreased magnitude record.
Average (4) rrequency Response runctionMagnitude 10 Pulse (t0xl) Excitation
93
N
········· ··j_..·_-_·_·..·t.._·..· ··· ·..·..·l·· _·-· ·1 'l _ i ~ : .
. i
·······..············..i· ·..··_···· ····t···········..···· ·····..····t·····..·· ··..··..···t······..······..···· !......................... ··!·..······..·······..····t·····..····· !._.1 ! ! 1 1 !
: 1 t t t . '•.............._ -...- _ -- _.............•......_.•..............._ .1 [ ~ 1 j j
1 r f f r : ; ......................... !•............_ ~ _ ! ..- ~ _ _ ! _ _ ~ ······t························!····i ! 1 ill ~ 1: l : :: ~:
~ t t r r . t-100""------------------------------......
Uj
~
40
20
0a:J"'0
c -20
4)
'7J -49~....cen -S0Itj
1:
-80
~requency in HertzAverage (4) Frequency Response Function
Magnitude 10 Pulse (l0xl) Excitation
uj-~
.; ~:.
-~.:: . : :: a:HE: ~!!
. . ~ . . : ~ ~.: Hjill
· . . ...-_ ....-· 1. ~ ! mi!i~!! !! ~
. . . f~1. .: j ll~r!! 'i I=........· · + _.._..t · · +..· ·_· tT · ..\ ·..i- ··-i;~rt~·ff·TI=-
: I Y : . .1!1t 1 If ~I;:::::.1:..·-....::::.:::..::..::..r:::::::::::::::::..::·1······· ....··· ..·······T··..····_..···· ..··············\······..········t..1TT···:····rlhi· . : ' :... 116............ ··· .. ····.. ·~ .... ··..····..··..···..·r·····..·····....····.. ········-r·····..········..·..·..r·· ..·····..······..·····1·....···..·....·····..·....······r··.. ···········+·..!..r·..···..·····'1Ir:~.
~ + , t t ~ ~ ~ img ~ ~~t t ~ . .:: ;.~~. ffil
-180 ~----------------------------'IJ1s
180
135
90
"It) 454)~
mII) e-0
c::
-45~III~
s: -90a.
-135
Frequency In Hertz
Figure 5.15. Frequency response function using fourincreased magnitude records.
Average (16) Frequency Response ~unct1on
Magnitude 10 Pulse (lext) Excitation
94
(\J
················~..·_····..·· · t··· ..···_·..·..····..·_· i..·..··..··_··..·······t i i ~ _;••••
..............._1 =:= t \ ..t. ...- _ L. L.: : : : x : : : ;r t r t ~ t T t
..· · ·..·j· · · l ·..·..· ·r ·..l· ·..r~ ·..r..· + · ·r..
.. ···..· ········..··!·..······..····· ··..t···· ..····· ··..···· ···t..······· ·..·..· t·····..···..··..·· f· ···................. .. ~ t".......... ·..······f..·j; ~ i ! it t t t + t~ ~ , 1 ; .._ _.................•..........- .- _ _ - .J J ! : ! ;; : : : : :
f r t ttl........................! + ! + ! ! + ..
~ ! ~ ~ ; 1 ~
. 1 .• •-100 .......-------------------------......
i
40
29
0CO
"c -29
G)
""C -40~
+oi
C0') -60tfj
~
-80
~requency in HertzAverage (16) Frequency Response Function
Magnitude 10 Pulse (10xl) Excitation
(\j('\J(\JtS)-~
t :: ~f
T . , . . . . ~~~ j~..........................! ~ .•.........._ _-~ __ +......•........••..••... ~......•••..._......•.•.......••• ~...•.•.....•.......••...~ :+:.~:
~ ~ ~ ~ ; ~ ; ! g:1 ~
•• •• +:':;". :'::::, - , ·"IE..:; 1 j ~~~~7 T T !?iii:. : ; :: . :::=:
t ~ , 1\ + . ;+1' i ~ i
..............· · , · · ·..·f· · · · i · _ t.L · ·; · · ·j" · 'l...· _... L ~'
":'; ~:' : : ;: :iii!t-----.......---.;.---:.,:.:::..:..: :.--:::+ :.:::::.•::::.•:::.~•.••........•••...••.... ~ ··········~························~······~······f··;;
: :: :===t ... III
~
........................! ; ~ ! ! ! ········~······ ..·····iiif!·: : : ~ . .. . :::=:
t . t t "1~ ~~.t ~ ~ ~ 1TI-t8e Lo- ........ --.I
L(')
~
180
135
90
~
11) 45t1)~
enl1)
0'"0
c
-45lU"4)Itt.c -90a..
-135
Frequency in Hertz
Figure 5.16. Frequency response function using sixteenincreased magnitude records.
Disk-3 Velocity Frequency Response Func.Shaped Pulse (2xlB) Excitation
95
40,......--~--~---~--~--~----~--~--~
(\J
20 ····················..·i···············~····_·~········ · · · · · · · · · · · · · · · · · · ..····i························+····· ..··· ··········~···· ·..·..·..·..·····..·.._·i·· ..··..·· ·· + ~ ..
-40 ..·..·..·..···· ..··..·..r·····..··············..r ··········_··_.. ·····.._·j"··_··············..·.. j·..··· ······· ···r· .: : : : ;
r ~ ; . r . . .-60 ·..····················t··_·················T-················..··········r- · ·····..r· ·..·..-r _ '1" ·_..r..· ·..--r... .. ..
. . I . I J . J-80 ..· · · T· · ·_..· ~..··..·_ ·_·· ·-T..· T_·..· ·..1"··..·..··.._..· · ~ ···..··..·..···..·..r·..· ·r..+ t t t . ti ~ \ ~
-20
-100 "----------------------------------'Lf")
~
Frequency in HertzDisk-3 Velocity Frequency Response Fune.
Shaped Pulse (2x10) Excitation180
135
90
~
C) 45~s,
m(I
0'1J
c
-45«)
("111.c -90a.
-135
-180~ - (\j l(")
~ IS) IS)IS)
Frequency In Hertz
Figure 5.17. Frequency response function using singlebinomial triangle record.
Average (4) Frequency Response ~unct1on
Shaped Pulse (2x10) Excitation
96
N('\J('\JIS)
· .
· i ' j 1 \
........................1"' + ······················r· ···..············+·······················T ·····························r···················..·t .
r . r . . i·····..··..····..·······j ·······..·····..·..·1 ···················..·······r········..······· r..···..··..··· ..···· 1' .
1 • : • : , . .·······················I···············..··•··..r..···..···..·..·····..···· ..··..~·· ..··· ··· ..······T·····..· ·..···· 1' _ 1' ·..7·········__·..··..··..r
t• • : . I : . •..··..··· ·· ·····1'···········..·····..···1·· j j ·..······..···..····r·················..·····..······r..·..· ·····..··1··..···.._·_·····_·[..·
• i
-10a '---------------------------......Lni
40
20
0t:O"'0
c -20
l1)
'"7J -49~
+'
cC"J -60ffj
L
-80
Frequency in Hertz *Average (4) Frequency Response Function
Shaped Pulse C2x10) Excitatl0n180
135
90
VJt) 4SQ.)s,
mo a-0
c
-45Q)
VJtfj .:
s: -90Q..
-135
-180IJ") N If") ('"I..j
~ l::) l=s;) ~lSI
.. , 1~~
Frequency In Hertz
Figure 5.18. Frequency response function using fourbinomial triangle records.
Average (16) Frequency Response FunctionShaped Pulse (2x10) Excitat;on
97
<'J
" ".................................................! ! ~ ~ .. ·············-f······_····· ~ . ~ : ~ ~
, 1 . , . 1.......................~ _r..· ·..·······..r····.. ·· ..··· ·.. ·f ···..··· ········r..·····..·..· · ··· ..~ _.._.
· r i ! r . . 1··········..··· ·..l-··········_·_····..·j·..·..·..···..···-· ··..·..·I..····_·· ·····..·•········..········..····r·· _.._.._ _J_ _. ..~.: ···_··············-1;: .: : i ::
· r 1 f r t···•·..····..··..·..····!··_..···········..•·..·r..···········..··· · ·..··!····..···..··..··..··..·r..··· ····..·..······l..···· ··..····..····..··..t:..· ·..·.._·t··· ..·..· ·· ~ ..
t t t t-100 '--------------------------....
~
40
20
0CD-0
c -20
G)
'"0 -40:J+ol
C0) -6£1t'd
1:
-80
Frequency in HertzAverage (16) Frequency Response Function
Shaped Pulse (2x18) Excitation180r---~--~---~--~--~---~--~--~
90
(1)
Il) 45~
~
C')
o e"1J
c
-45d.)
CI.I
""...
s: -90a..
-135
-180lJ'j .-4 N In N U1 ('\Jf3) eg cs:> t'S)1St
Frequency in Hertz
Figure 5.19. Frequency response function using sixteenbinomial triangle records.
98
The bipolar triangle signal's performance is displayed
in Figures 5.20, 5.21, and 5.22. These results near the
half-sampling frequency improve with averaging, but the low
frequency regions are not meaningful. In Figure 5.22, the
high frequency mode is identifiable. The spliced results of
the complementary pulse pair are shown in Figure 5.23. In
this method, both the flexible modes are distinguishable.
Finally, the results from noise excitation are presented
in Figures 5.24, 5.25, and 5.26. Here, the low frequency
results are not of the quality achieved with deterministic
excitations. However, the high frequency fidelity improves
dramatically with averaging.
The results of the experiment show that there are some
clear distinctions in frequency response function estimate
fidelity due to the excitation signal. For low frequency
fidelity, a deterministic method using a signal of high
spectral magnitude is desirable. Also, a deterministic tack
may be indicated if averaging is not to be performed, or if
only a few records will be used in an average. On the other
hand, with a moderate number of records to include in an
average, excellent high frequency fidelity is achievable
with stochastic methods. Of course, for broad-spectrum
applications, a combination of deterministic and stochastic
Dlsk-3 Velocity rrequency Response rune.Shaped Pulse (-2x10) Excitation
99
40..----~--.....,..---~---~--~---~--~---~
NN
·· ..··············..··..I..··················..··t-······_·····_···········T"7\·····.._·_··+······················l· _ : .
T ! ~\, , r T !
-1.... ·····..r..··..---··..········..·;······..·· ~·· ..·..·..··········t················..···············!..·····················T·..·················..·j····
..................._..I I 1. 1 r.. f. . 1. .
r . · · . , ... ····..· ·.._···..·r···..····..···· ·..·r ··..~ ·_.._····· ··!..·· ········..·_·;·······..·..·· ·······r······..· !' : .
r r . . . r :····..·.._···..···--r········..·..····..····r....··..·······..····..·.._·..l·····_····..····....···I-··..··_..········..··r·····..··········....··....····r··..··..···..····..··..1..··..··..·····.....
t r f f r , · .···..·..·..·..·......···i..··..···..··....·..····r-·....··_·....·....···..······r·....··..···..·_·..···I···..·..············....r·....·..·..···..····..·····..···!·....·..·....···......··~·....·.....··..···_..r-·t! ! t
20
-100'----.-....----...--------------------~\i)
i
t11"0
c -20
4)
1J -49:J.,:-
C0') -60,'1L:
-80
Frequency in Hertz
D1Sk-3 Velocity Frequency Response Func.Shaped Pulse (-2x10) Excitation
t :
...:.•..:: .~.::': ::::.. • ! . J.., 1
1 I .! ': HUi :i;' :;..·········-······ ~.:········ .._···..·······..Jt.. ··..··..· ··-..· · ·~.·r.·.; ~.; i ·..•..........•. t····.······..·.,.·· -M ••••••• -,.
! ~ ~ L ~ ~ i H~Hi.: ~h ~: ji
! it !II ~ ...····..···_·..··....··t·....·····....·..·..·r·tr....··_........·..·..··.. ··.. ·t·r~1 ..·..··......_....·+··..·..·_·..··....·..··t· ....··.... ·I#·· ~.....
! {f1 !U
180
135
90
(I)
Q,) 4S4)
~
0)
lIJ0-0
c
-45()
iIJl1j
.c -90a.
-135
N-CS)
-180 .......---.......----..........----------....0.----------..~
Frequency in Hertz
Figure 5.20. Frequency response function using singlebipolar triangle record.
Average (4) rrequency Response FunctionShaped Pulse (-2x10) Excitation
100
40.----~--~-----.---~--~----~--"""!"---~
('\oj
..1·.. ···......··..·....···1....
.......................
·······················r······..······..·····~········· ; + : .-." ·······T.· ~ ..
\. :..·..I..··..··_···_··..·..·····..·t·····..· ·~·~·..·· ·..;..··· ..
........................j'" ··..·..·..f· ! ..
· . • r . . .·..·····..····..·..··..·1'..··..·_· ·..···1..·..·_··_··..·······..··;·· · · ·····r..···..·· ····..···..!..··..·..···..·..·· .,. - .
· . r . . . .··········..····_····r·········....·....·····f··......·..···_··......·..····r-..·..·....···....·..·1..···_·..·..··_·..···r·_..··..···....··....····..··..r·······..··....·······"7···....········ ···....·1·
· . . r , r , t·..···..·..·············r·· ·· ···..··..··~· ..···..···..·..·..·..···..·····[..·..··..········· ··1····· ·..·········.. ·1 1' ) ~ ..
t 1
" .
20
-100"'-----------------------------'U1
i
ttl1'J
c -20
G)
'"'C --49:J.+oJ
c0') -6910~
-80
rrequency in Hertz
Average (4) Frequency Response FunctionShaped Pulse (-2x10) Excitation
180
135
90
UJI) 45vs,
enQ)
0"'0
c
-45e(/)
~
..c -90a.
-135
-180Ln~~
N
Frequency in Hertz
Figure 5.21. Frequency response function using fourbipolar triangle records.
Average (IS) Frequency Response FunctionShaped Pulse (-2x10) Excitation
101
('\J('\J
. .
··················_·-1······················1·········· -r J..... + . 1
: 1 · . . ' .························r·················..··..1·_···..····..····..·····..····1..····..····..··..···..·1····..········· ·1·· ····· ·· ···..···..·····1..··..·· ··· ·_·1·..
I f I . . . .·..···..·..··..·--····r·....·········--···l·········...··....·..·..······..1'·..·__··__·······1·..········_....··....·[..·....·..···..····....··....·..r···..···...._·..···..·1..········......! + + + + . .1 1 1 1 r ; . .··..··..··--·..·-r···..···· ·..···..r·..·..·..··..··· ·.._·..··~ _·..···..r· ! 1'· ····· ..·..·..· 1·..···..····"'-'''1''
+ r t-100 "'-"---------------------------~
i
40
20
0CO-0
c -20
l)
~ -49:J~
Cm -b1)10~
-S0
rrequency in HertzAverage (16) Frequency Response Function
Shaped Pulse (-2x10) Excitation180
135
30
ent) 45G)s,
O"J11)
0-0
c
-45Q)(J)
~
s: -90a.
-135
('\J
:: 10:
('\J-100'--------------------------------.........
lJj
iFrequency in Hertz
Figure 5.22. Frequency response function using sixteenbipolar triangle records.
40
20
0ID1'J
c -20
wf.J -40::J.+oJ
cen -60~~
"-
-80
Average (16) trequency Response ~unc.
Spliced Shaped Inputs (2x10) & (-2x10)
: :·····_····-r········ ~·..··.._..·r"·"···..·....····_·T..·....·_··..·....T·"·....·........·_·r
! i ~ j ~ j ! j
·..···· · ···· ·..· ···r..·· ·· ··..· ·!···..· " +......................... . !..................... • 1"..
+ r . ; •..····,,·····..····1·· ·..·· 1'·..·_··..····..·····1-..··"·..·_.._··t..·_··· _· ·..•· ·..··· _..·..·,,..t·..·........ .. .-t-......_ .
..._ ; 1.. _._ ~ + ! \ L..- +-
102
-100~-------------------------_-..~~
rrequency in Hertz
Average (16) Frequency Response Func.Spl iced Shaped Inputs (2x10) & (-2x10)
180
135
90
I/J1]) 45~
~
0')~
0"'0
c
-45Q)r/)
rU..c -90CL
-135
.. ;
.:: • :::::jt,.;
. ~ uL ~ i i:: ~m~!H L.......................j ~ " :.. + " t· · · · ·..· ·ri~lI~J,l..li!
: , ' : T. ~ ,:L! :~iggjIL rdf~
! :
.......................l _ " l " L I L L ....ll!..I:.\ii + ; :,! · i If IiIi'll.
..·..· ····..··::L:···.:·:·.·· .:s:"".+" "".:r: " L ~ ~,i.··..··..·..· ·..i l+..i-H·t-III: :;~ ;:: :;~: =c:s::=, ., : = :::: ::::::=: .: ·1 :;:: ;'i:=:~:
: : ~ : : :' :: ; ~~ ;~ ;~ i!·! g~ ;................................................. ") ~ ... .. .'j lHl ~!~,
N-180 '-------------------------------
lJ)IS)(!j
Fr~quency in Hertz
Figure 5.23. Frequency response function using sixteenspliced triangle records.
Djsk-3 Velocity Frequency Response rune.Unity White Noise Excitation
103
40
20
'3~'"0
c -20
II"'0 -40~
+-'
c0) -60~
L
-80
..·····················..··..··················T·····..···················.. ·~··.._··_············r··..·· ~ _ _ ~ _~ .
........................: ·..· ·· ·..t ····..K ·..· ·· · ·..· i · r ..
~::: .: : ; I . . : .·····..··· ..· ·.. ····~ ..· ···..·..·····1.. ·····..········· , ~ , · ···..··..·..·..· r..· · ··_..·•·..·..· · ·..· .
('\Ju')I:'\J-100 \.----------------------------....
lJ')lSI~
rrequency in Hertz
DiSk-3 Velocity Frequency Response Func.Unity White Noise Excitation
180,.---~--~---~-~--~---~------ .....
90
45
; .: j! .: :::::: ;
. . . ~ !:;. i di~ ~1h;135 ··..·······; ········ ···· t..·..·..·· ·..··· ..······;··.._ ··..·····..···t· ..···..·· ··· ..·····~ ..·········· ..· ·..·i..···i ·; ·f·t·.. ·~··j ..t~ .. ~ ~ ilffi
.)
-45 ··················· ..··~····· ....·······..·..··T·......·..·..········..·......r·····..···_..···..···!!·:··· ··..··· .i.,..··..·....··..··· ·..: ···..·······..·t~ ~f..l=!
-913 -,,·_'-,·,-·· ·--·f ·""·····..··;·:l· · · ·..··..·..·..·..·· ·..·· · ········.. ···· _·HI·_·H~: ~~~ I: ~fil
; ~ : ~i :: ;;g iii-135 ,·.. ··· ..·· 1 _..···..··..·..··t· · · ·· ··..···i.. ······.._..·..·_·..·t..·..· ·········..·t··· .. ·· ·.. ·· ·-!..·· · ···..·..··ti~··..r··..ff·ill:.
~ f r ! t t t~ ~ ff j~,r+ +? ; ~ t; ~j
('\J-180~----------------------------'
If":\c:gIS)
Frequency in Hertz
Figure 5.24. Frequency response function using singleunity white noise record.
Average (4) Frequency Response FunctionUnity White Noise Excitation
104
('"\J('\J~
. . . . , . , .···..······..·..·..·....1..······ ········..··..··t············..··..··......··....1......·.._··.. ··..-··r..·.._·········..·_..l···..·..··..······....····..····· t····..·················r·················.._··\·..·
. i _.:.: ;
: ! ~
······················r····················r···············_··········r···..·····..········r··············_·····r············~~···············T··· .. ······ C'
............._.........•..............._......•............._ - _._ - _ , _.................................. ..: : : ! ! :
I t f f ;. . . L j : : i
••••••••••••••••••_ _ _ ••••••••••• t· .. ••••.. -···,,···_·_·.····_········_·.. ·-··.············ ••••••••••••••••••••.,.•••••.••••••••••••••••••••••••••- ••••.••••.•••• t· •
j ~ . ~ 1 j j 1
. + r ~ i
-100 L....-__......._ ....... ......... ........ ........
If')
~
40
29
0CO-0
c -20
Q)
~ -40~
~
cr.n -60~
z
-80
Frequency ;n HertzAverage (4) ~r.quency Response Function
Unity White Noise Excitation
('\J-rs-
-45
-180 '--__...i--__....... --'
LI"J
~
-135
180
135
90
II)
o 4511)
\..0')l1) e""0
c
'1)(,')
1'0
5: -90 ·:::.:..._".::::::.:.:.:::..:::-::.c........
Frequency in Hertz
Figure 5.25. Frequency response function using four unitywhite noise records.
Average (IS) rrequency Response ~unction
Unity White Noise Excitation105
-~
I. i
···•···········..······j····..····_·..········i···············..··············r···.._·· ..··········r·····..·············T- ·······················..···')""··..·····..········..·r..··········_·······;····~t:\ t
:::::::::.:.::::::.I:::::::·::.:.:::~I::::::::::.:.::...:::::::I.:::::::::::..::·:I::.:~:::::.:]:::::.:::::::.:~ ..:':::::I::: ~ ~ ~ : t ~ :
, ~ • • . r 1........................! _ ··········r..···· '" ·······r· ~ -1' __ ···..····..1 ·· ··_··1 ·············
r ! t ;-lee L- ........
Lf':l
i
40
20
0CO-0
c -20
tD'7J -40:5+"
Cen -60~
L
-80
Frequency in Hertz
Average (16) Frequency Response FunctionUnity ~hite Noise Excitation
t'\J-(S)
•.! ~ ~~l:. ~
......:.::.....
f :
..................·I..· ···..·..+ .•:.:.:.:.:.:.:.·:·.J.t::::::::::::·:::::::I.~:::::·::::·::'::::':::::::':::::!.::::::::::':"'::"":~.:.i·':::::::-:··::::.-::::'-:-a-:::-::':. .-
• \~e; iae
........................ j t..·········· ·····..·· ·j······· ········..···! j ·t..··· ····+· ·..· t·j~
r ! ~ l~'f: ~jr~:::~: :-180 L..-. .......,
If')
~
180
135
90
~
to 45l1)
~
0')G)
0fJ
c
-454)CI)
tf:J.c -90a.
-135
Frequency in Hertz
Figure 5.26. Frequency response function using sixteenunity white noise records.
106
approaches may be employed. In any case, the results of
this chapter should provide some insight into the successful
selection of suitable excitation signals.
107
CHAPTER 6.
CONCLUSION
6.1 Observations
In this thesis, methods of improving the fidelity of
empirical sampled-data frequency response function estimates
have been investigated. The successful identification of a
high-fidelity system response function is the cornerstone of
the LSS design methodology, Design-To-Performance, which was
outlined in Chapter One. The analytical methods reviewed in
Chapter Two permit the identification of the frequency
response function matrix necessary for the One-Con
troller-At-A-Time design procedure discussed in the
introduction.
In order to achieve high-fidelity estimates, several
methods of enhancing the properties of excitation signals
were considered in Chapter Three. These methods include
increasing the signal's magnitude and/or duration, shaping
the signal, and using complementary signal pairs. Other
means of improving results include averaging several
responses and combining deterministic and stochastic methods
advantageously.
The system identification process and signals of
interest were explored via digital simulation in Chapter
108
Four. Finally, the practical application of the methods
discussed was demonstrated with the hybrid system experiment
and results presented in Chapter Five.
6.2 Applications
The methods of enhancing excitation signals used in
system identification discussed in this thesis were
developed from the viewpoint of LSS applications. However,
the same principles apply to any application that requires
the empirical identification of a sampled-data frequency
response function. Likely applications include very high
order systems or systems whose dynamics are not well
understood. For instance, many biomedical systems may lend
themselves to a similar analysis.
As a direct application of the results detailed in this
thesis, the identified response functions could be used to
design controllers for the hybrid system. In this manner,
the influence of the identification methods on the
closed-loop system performance could be investigated. since
this is the purpose for the identification process, it is
the true test of adequate fidelity of the identified
response function estimate. As a more ambitious project,
the analog computer simulation could be replaced with a
109
physical system, e.g. a pointing system. Such a demonstra
tion should establish the usefulness of the methods
described herein.
110
REFERENCES
1. Jerrel R. Mitchell, Sherman M. Seltzer, and H. Eugene
Worley, "Design-to-Performance", Proceedings of the 1985
IFAC Conference on Automatic Controls in Space,
Toulouse, France, June 1985.
2. Jerrel R. Mitchell and J.C. Lucas, "l-CAT: A MIMO Design
Methodology", Workshop on Structural Dynamics and
Control Interaction of Flexible structures, Sponsored by
OAST and MSFC, NASA Marshal Space Flight Center,
Alabama, April 1986.
3. Julius S. Bendat and Allan G. Piersol, Engineering
Applications of Correlation and Spectral Analysis, John
Wiley & Sons,Inc., New York, 1980.
4. Julius S. Bendat and Allan G. Piersol, Random Data
Analysis and Measurement Procedures, Second Edition,
John Wiley & Sons, Inc., New York, 1986.
5. Charles P. Plant, Schemes for Improving the Fidelity of
Empirical Sample-data Frequency-domain Models of Large
Space Structures, Masters Thesis, Mississippi State
University, Mississippi, May 1987.
6. Jerrel R. Mitchell, Advising Session, Ohio University,
Ohio, April 1988.
111
7. William P. Maggard, Adaptive Control of Flexible Systems
Using Self-tuning Digital Notch Filters, Masters Thesis,
Ohio University, Ohio, September 1987.
8. Electronics Associates, Inc., "Model TR-20 Desk Top
Analog Computer Operation Manual", Electronics Associ
ates, Inc., Long Branch, New Jersey.
9. Hewlett-Packard Co., "HP-3852A Data Acquisition and
Control unit Plug-in Accessories Configuration and
Programming Manual", Hewlett-Packard Co., Everett,
Washington, 1983.
10. Raymond G. Jacquot, Modern Digital Control Systems,
Marcel Dekker, Inc., New York, 1981.
11. Daniel Graupe, Identification of Systems, Krieger
PUblishing Co., Inc., New York, 1972.
12. Charles L. Phillips and H. Troy Nagle, Jr., Digital
Control System Analysis and Design, Prentice-Hall, Inc.,
New York, 1984.
13. L. Ljung, "Identification: Basic Problem", System .i
Control Encyclopedia: Theory, Technology, Applications,
Editor-in-Chief M.C. Singh, Pergamon Press, U.K., 1987.
112
14. K. Glover, "Identification: Frequency Domain Methods",
System ~ Control Encyclopedia: Theory, Technology,
Applications, Editor-in-Chief M.C. Singh, Pergamon
Press, U.K., 1987.
15. G.S. Hope and O.P. Malik, "Identification: Pseudo-random
Signal Method", System ~ Control Encyclopedia: Theory,
Technology, Applications, Editor-in-Chief M.C. Singh,
Pergamon Press, U.K., 1987.
16. J.C. Lucas, I-CAT: A MIMO Design Methodology for the
Control of Large Space Structures, Masters Thesis,
Mississippi State University, Mississippi, December
1986.
17. Daniel DeBra, Jerrel R. Mitchell, Sherman M. Seltzer,
and H. Eugene Worley, "Practical Dynamics & Control of
Large Space structures", AIAA Professional study Series
Book Three, Huntsville, Alabama, August 1986.
114
This appendix contains the source code listings for the
digital simulation of Chapter Four. The programs are exe
cutable command files for the MatrixX control system analy
sis and design software package.
The three disk system described in Chapter Four is in
corporated into MatrixX variables in the ANCMPINI program,
which initializes the digital simulation of the analog com
puter. The programs for the simulation provide three major
functions. The CALCFRSP program permits the generation of
continuous and sampled-data frequency response functions for
the system. The RTLOCUS program provides the capability to
examine the continuous and discrete root loci of the system.
Finally, the SYSTEMID program accesses several subprograms
to perform a system identification simulation.
There are five subprograms called by the SYSTEMID pro
gram. The first is GETPARAM, which gets the system identi
fication parameters for the user. The DISCRTCL module forms
the discrete closed-loop system. Then, the GENINPUT module
generates the input signal defined by the user. Finally,
the time simulation and frequency analysis are performed by
the DTRMNRSP program for deterministic analysis or by the
STOCHRSP program for stochastic analysis.
115NOTE: Comment lines in MatrixX cannot be placed in com-
pound statements. In the listings that follow, comments
precede the compound statements which they describe, e.g.
WHILE loops.
II PROGRAM116
--------~-~~~--~~~--~~-~-~~-~~~-~~~-~-~~~--~~~~--~~------~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-~~~~-~~~~~~~~~~~~~~~~~~~~~~---
II NAME: ANCMPINI
II
II AUTHOR: T T HAJVIIVIOND
II
II DATE: 10 APR 1988
II
II PURPOSE:
II This file is a MatrixX executable command file. Its purpose is
II to initialize the MatrixX main program with a state-space
II description of the system analyzed in the thesis. A modal
II realization is used paralleling that implemented on the analog
II computer. This permits sirrulation of the analog ccmputer's
II responses to various signals. In order to effectively magnitude
II scale the analog computer simulation~ the dynamic ranges of the
II states of the realization must be determined. Then the variables
I I generated by t he ana log canputer can be norma1; zed to prevent
II "their exceeding the dynamic range of the computer's amplifiers.
II Additionally, the MatrixX variables initialized by this program
I I are SAVEd in the file ANCMPSIM. VAR for use by other program
1,1 modu1es .
II The rigid body mode of the system's response is modeled by
II subsystem A. The 0.1 Hz flexible body mode ;s modeled by
I I subsystem 8, and the 1.5 Hz flexible body mode is modeled
I I by subsystem C.
II
II VARIABLES ==--==--===========================--===============--========
II AA, AS, AC
II
II BA, 88, EC
II
II CA, C8, CC
II
II DA, 08, DC
II
II NSA, NSB,
II NSC
II SA, 58, SC
II
117coefficient matrices of state vectors in the
state equations of the subsystems
coefficient matrices of input vectors in the
state equations of the subsystems
coefficient matrices of state vectors in the
OLJtput equatrions of the subsystems
coefficient matrices of input vectors in "tt1e
output equations of the subsystems
integers indicating the subsystems T orders
matrices composed of coefficient matrices "to form
the Matrix)( representatioras of the subsystems
II
II BEGIN ===============================================================
II CLEAR SCREEN AND DISPLAY PROGRAM FUNCTION -------------------------
ERASE
DISPLAY(TINITIALIZING ANALOG COMPUTER SIMULATION ... T)
SHORT E
II DEFINE ANALOG COMPUTER SIMULATION SUBSYSTEMS -----------------------
AA= [0 1 ; 0 OJ ;
SA = [0; 1/3];
CA = [1 0; 0 1] ;
DA = [0 ; 0] ;
SA = [AA SA; CA DAJ;
NSA = 2;
A8 = [0 1; -0.39478 OJ;
88 = [0; -0.33482];
118C8 = [1 0; 0 1] ;
08 = [0; 0] ;
S8 = [A8 88; CB D8J;
NS8 = 2;
AC = [0 1 ; -88.82686 OJ ;
Be= [0; 1.48807E-3];
CC = [1 0; 0 1] ;
DC = [0; OJ ;
SC = [AC Be; CC OC];
NSC = 2;
II CLEAR SUBSYSTEM COEFFICIENTS AND SAVE SUBSYSTEM DESCRIPTIONS ------
CLEAR AA A8 AC SA 88 Be CA C8 CC DA DB DC
SAVE(tANCMPSIM.VAR')
II CLEAR VARIABLE STACK ----------------------------------------------
CLEAR
II END OF ANCMPINI PROGRAM =======================--====================RETURN
II
II
PROGRAM119
-----~~-~~~~~~~-~~-~~--~~~-~~~----------~~~~~~~~~--~~~---~~~~~~~~~~~~~~~~~-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
I I NAME: CALCFRSP
II
I I AUTHOR: T T HAMMOND
II
II DATE: 10 APR 1988
II
II PURPOSE:
II This file is a MatrixX executable command file. Its purpose is to
II generate the calculated frequency response for the continuous-time
II analog computer or the discretized analog computer system. The
II file ANCMPSIM.VAR must be placed in the current directory prior to
I I "the executian of this program.
II The response functions calculated are for Disk-3 velocity.
II
II VARIABLES
II CHOICE
II DELTAT
II DSNFRESP
II DSFRESP
II K
II KFB
II MFRESP
II NAB, NABC
// NFRESP
II NSA~ NSB,
-----------~~~--~-~~~----------~-~~-~-~~~~--~-~-~~-~----~~~~~~~~~~~~~~~~~~~~~~~~~~---~~~~~~~~~~~~~~~~~-~~~~-~~-~~~
index of menu selection
sampling period in seconds
order of discrete system
MatrixX description of discrete system
compensator gain
feedback matrix
connect function input matrix
intermediate subsystems f orders
connect function output matrix
120II NSC
II NSFRESP
II SA, S8, SC
II
II SAB, SABC
II SFRESP
II
integers indicating the subsystems' orders
order of analog system
matrices composed of coefficient matrices to form
the MatrixX representations of the subsystems
intermediate subsystem matrices
MatrixX description of analog system
II CONSTANTS ==========================================================
II
II SELECTION MENUS:
TI MEMENU = [f * * MENU * * f
, CONTINUOUS r
DISCRETE f];
II
IVOREMEJ\IU = [f * * MENU * * r
MORE
QUIT -i.
II
I I QUERY RESPONSES:
NO = fN';
y = 'Y":
YES = fyt;
II
II BEGIN ===============--==============================================
II CLEAR SCREEN AND DISPLAY PROGRAM FUNCTION -------------------------
ERASE
121DISPLAYC'FREQUENCY RESPONSE FUNCTION GENERATION ... T)
II RETRIEVE SYSTEM SIIVlILATION AND FORM DISK-3 VELOCITY OUTPUT -------
LOAD('ANCMPSIM.VAR')
[SAB, NABJ = APPEND(SA~ NSA, S8, NSB);
[SABe, NA8CJ = APPEND(SA8, NAB, SC, NSC);
MFRESP = [1 1 1]';
NFRESP = [0 1 0 1 0 1J;
KFB = [0 0 0 0 0 0; 0 0 000 0; 000 0 0 OJ;
[SFRESP, NFRESP] = CONNECT(SA8C~ NABe, KFB, MFRESP, NFRESP);
II WHILE LOOP COMMENTS ==========~=====================================
II SET UP LOOP FOR MULTIPLE RESPONSES --------------------------------
II QUERY USER FOR CONTINUOUS OR DISCRETE RESPONSE --------------------
II SPACE CONTINUOUS FREQUENCY DATA LOGARITHMICLY ---------------------
II DISPLAY CONTINUOUS FREQUENCY RESPONSE FUNCTION --------------------
II INQUIRE IF PRINTING IS DESIRED ------------------------------------
II QUERY USER FOR SAMPLING PERIOD ------------------------------------
II SPACE DISCRETE FREQUENCY DATA LOGARITHMICLY ----------------------~
II DISPLAY SAMPLED-DATA FREQUENCY RESPONSE FUNCTION ------------------
II INQUIRE IF PRINTING IS DESIRED ------------------------------------
II QUERY USER FOR MORE RESPONSES -------------------------------------
CHOICE = 0;
WHILE CHOICE <> 2; ...
CHOICE = MENU(TIMEMENU,1); ...
IF CHOICE = 1; ...
GEOMRATIO = 10**(1/250); ...
FRANGE = [O.OS*PI O*ONES(1,500)J' ; ...
FOR 1=1:500, FRANGE(I+1)=GEOMRATIO*FRANGE(I); END, ...
122H = FREQCSFRESP, NFRESP, FRANGE); ...
FHZ = FRANGE/(2*PI); ...
PLOT(FHZ, 20*LOGCA8S(H))/LOG(10), ...
'TITL/DISK 3 VELOCITY CONTINUOUS FREQUENCY RESPONSE FUNCTION/ ...
YLAB/dB/ ...
XLAB/FREQjENCY IN HZ/...
LOGX t ) ; •••
INQUIRE CHOICE 'DO YOU WANT TO PRINT THIS FUNCTION
IF CHOICE = 'Y' , ...
PLOT('PRINTER HOLD'); ...
PLOT(FHZ, 20*LOG(A8SCH))/LOG(10), ...
, .. , ...
'TITL/DISK 3 VELOCITY CONTINUOUS FREQUENCY RESPONSE FUNCTION/ ...
YLA8/dB/ ...
XLAB/FREQJENCY IN HZ/...
ERASE; ...
PLOTC'OISPLAY HOLD'); ...
END; •..
ERASE; ...
ELSE INQUIRE DELTAT TENTER SAMPLING PERIOD IN SECONDS'; ...
[DSFRESP, DNFRESPJ = DISCRETIZE(SFRESP, NFRESP, DELTAT); ...
GEOMRATIO = 10**(1/250); ...
FRANGE = [O.OS*PI*OELTAT O*ONES(1,SOO)J'; ...
FOR 1=1:500, FRANGE(I+1)=GEOMRATIO*FRANGE(I); END~ ...
H = FREQ(DSFRESP, DNFRESP, FRANGE, 'DIS'); ...
PHZ = FRANGE/(2*PI*OELTAT); ...
PLOT(FHZ, 20*LOG(A8S(H))/LOG(10), ...
123'TITL/DISK 3 VELOCITY SAMPLED-DATA FREQUENCY RESPONSE FUNCTION/ ...
YLAB/dB/ ...
XLAB/FREQUENCY IN HZ/...
LCX3X' ) ; •••
INQUIRE CHOICE '00 YOU WANT TO PRINT THIS FUNCTION
IF CHOICE = 'y t~ •••
PLOT(tPRINTER HOLD'); ...
PLOT(FHZ~ 20*LOG(A8S(H))/LOG(10)~...
, .. ,. ....
'TITL/DISK 3 VELOCITY SAMPLED-DATA FREQUENCY RESPONSE FUNCTION/ ...
YLAB/dB/ •..
XLAB/FREQUENCY IN HZ/...
ERASE; •..
PLOT('DISPLAY HOlD t) ; •••
END; ••.
ERASE; •..
END; •.•
CHOICE = MENU(MOREMENU,1); ...
END
II CLEAR VARIABLE STACK -----------------------------------------------
CLEAR
/ / END OF CALCFRSP PROGRAM ================--==========--===
RETURN
124II PROORAM
/ / NAME: RTLOCUS
II
II AUTHOR: T T HAMVIOND
II
II DATE: 10 APR 1988
1,1
I I PURPOSE:
II This file is a MatrixX executable command file. Its purpose is to
II generate the r-oot; locus for ,the con"tinuous-'time analog canputer
II simulation or the discretized analog computer system. The file
II ANCMPSIM.VAR must be placed in the current directory prior to
I I the execution of this program.
II
II VARIABLES
II CHOICE
II DELTAT
II DSNRTLOC
II DSRTLOC
II EV
II GAIN
II K
II KFB
II MRTLOC
1/ NAB, NAOC
II NRTLOC
1/ NSA p NSB p
~~-~-~~-~~~~-~-~---~~~~~~~~~~~-~~~~~~~-~--~-~------~~~~~~~~~~~~~-~~--~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-~~~~~-
index of menu selection
sampling period in seconds
order of discrete system
MatrixX description of discrete system
matrix of discrete poles corresponding 'to GAIN vector
row vector of gains for desired discrete roots
compensator gain
feedback matrix
connect function input matrix
intermediate subsystems' orders
connect function output matrix
125II NSC
II NSRTLOC
II SA, S8, SC
II
II SA8, SABC
II SRTLOC
II
integers indicating the subsystems' orders
order of ana log system
rllc."9trices composed of coefficient matrices to form
the Matrix)( representatior1s of the subsystems
inotermediate subsystem matrices
MatrixX description of analog system
II CONSTANTS ==--=======================================================
II
II SELECTION MENUS:
TIMEMENU = ['* * MENU * *'
, CO\ITI NUOUS t
DISCRETE '];
II
IVOREMENU = [t * * MENU * * t
MORE
QUIT T];
II
II QJERY RESPOI\ISES:
N = TN';
Y = Ty';
YES = Ty';
II
II BEGIN
/1 CLEAR SCREEN AND DISPLAY PROGRAM FUNCTION -------------------------
ERASE
126DISPLAY('ROOT LOCUS GENERATION ... ')
II RETRIEVE SIMULATED SYSTEM AND FORM DISK-1 VELOCITY OUTPUT ----------
LOAD('ANCMPSIM.VAR')
[SAB~ NAB] = APPEND(SA t NSA, S8, NS8);
[SABe, NABeJ = APPEND(SAB, NAB, SC, NSC);
MRTLOC = [1 1J';
NRTLOC = [0 1 0 -1.9909 0 -3.4905E-3];
KF8 = [0 0 0 0 0 0; 0 0 000 0; 0 0 0 0 0 OJ;
[SRTLOC, NRTLOCJ = CONNECT(SABC, NABC, KF8, MRTLOC, NRTLOC);
II WHILE LOOP COMMENTS ================================================
II SET UP LOOP FOR MULTIPLE LOCI --------------------------------------
II QUERY USER FOR CONTINUOUS OR DISCRETE ROOT LOCUS -------------------
II DISPLAY CONTINUOUS ROOT LOCUS --------------------------------------
II QUERY USER FOR PRINTING CONTINUOUS ROOT LOCUS ----------------------
II INQUIRE FOR SAMPLING PERIOD AND GAIN VALUES ------------------------
II DISPLAY DISCRETE ROOT LOCUS ----------------------------------------
II QUERY USER FOR PRINTING DISCRETE ROOT LOCUS ------------------------
II QUERY USER FOR MORE LOCI -------------------------------------------
CHOICE = 0;
WHILE CHOICE <> 2; ...
CHOICE = MENU(TIMEMENU,1); ...
IF CHOICE = 1; ...
K = RLOCUS(SRTLOC, NRTLOC); ...
INQUIRE CHOICE 'DO YOU WANT TO PRINT THIS LOCUS
IF CHOICE = 'Y' , ...
PLOT('PRINTER HOLD'); ...
EV = RLOCUS(SRTLOC, NRTLOC, K); ...
, .. , ...
127ERASE; ...
PLOT(TDISPLAY HOLD'); ...
END; ...
ERASE; ...
ELSE INQUIRE DELTAT 'ENTER SAMPLING PERIOD IN SECONDS'; ...
[DSRTLOC, DNRTLOC] = DISCRETIZE(SRTLOC, NRTLOC, DELTAT); ...
INQUIRE GAIN 'ENTER VECTOR FOR GAINS OF ROOTS TO BE SHOWN'; ...
ev = DRLOCUS( DSRTLOC, DNRTLOC, GAIN); ...
GAIN, ...
INQUIRE CHOICE 'DO YOU WANT TO PRINT THIS LOCUS
IF CHOICE = 'Y' , ...
PLOT('PRINTER HOLD'); ...
EV = DRLOCUS( DSRTLOC, DNRTLOC, GAIN); ...
ERASE; ...
PLOT('DISPLAY HOLD'); ...
END; ...
ERASE; ...
END; ..•
CHOICE = MENU(MOREMENU,1); ...
END
, .. , ...
II CLEAR VARIABLE STACK -----------------------------------------------
CLEAR
II END OF RTLOCUS PROGRAM =============================================RETURI\I
128II PRCX3RAM --
II
II NAME: SYSTEMI0
II
II AUTHOR: T T H.AJVI'JtOI\ID
II
II DATE: 9 APR 1988
II
II PURPOSE:
II This file is a MatrixX executable command file. Its purpose is
I I to drive the system identification process. This is accompl ished
II by calling smaller MatrixX executable command files to perform
I I required tasks.
II Preparation to use this file must include creating the ANCMPSIM.VAR
II file in the DOS default directory. That file contains the analog
I I computer simulation MatrixX description. It is created by
II executing the ANCMPINI.EXC file.
II Files read during the execution of this file must be in the OOS
II default directory. Files written during execution will be placed
II in this same directory.
II
II VARIABLES
II ANSWER
II TYPE
II
II
~~~~~~-~~~~--~~~~~~~~~~~~~---~~~-~~-~~----~-~~~~----~~~-~~~~~~~~~~~~~~~---~~~~~-~~~~--~~~~~~~~~~~~~~~~~~~~~~~-~~~
response to question
= for deterministic input signal
= 2 for stochastic input signal
I I CO\ISTANTS .................-..-...-
129N = TN t ;
NO = TN t ;
II
y = 'yt;
YES = "( ' ;
II
II BEGIN
II CLEAR SCREEN AND DISPLAY PROGRAM INSTRUCTIONS ---------------------
ERASE
II
* * * * * * * * * *')DISP('* * * * * * * * * SYSTEM IDENTIFICATION
DISP('*
DISP('* This program simulates the sampled-data frequency
DISP( '* response function method of system identification
DISP('* applied to a highly flexible structure.
DISP('*
*' )*' )
*' )
*' )
*' )
DISP('* A variety of input signals may be specified by menu *t)
DISP('* responses. Also -the sampling period and length of -time *')
DISP(f* for each sampled record is selectable. The maximum *')
DISP(T* number of samples must be restricted to 256.
DISP('*
DISP('* The program pauses at several statements until the user *')
is ready to continue. To cant;nue~ enter <RETURN> at
DISP('* the PAUSE> prompt.
DISP('*
*' )
*' )
*' )
DISP(T* Time responses and frequency responses of pertinent *' )
DISP('* data are displayed graphically. These are plotted on *T)
130DISP('* the default plotting device (video display at power-on). *')
*' )
DISP('* * * * * * * * * T T HAMMOND 21 APR 88 * * * * * * * * *')
II
II WAIT FOR USER RESPONSE ---------------------------------------------
OISP(' ')
PAUSE
II
II WHILE LOOP COMMENTS =============================================--==
II SET UP LOOP FOR MULTIPLE IDENTIFICATIONS -------------~-------------
I I GET SYSTEM AND INPUT SIGI\IAL PARAMETERS FROM USER -------------------
II FORM DISCRETE CLOSED-LOOP SYSTEM -----------------------------------
II GENERATE INPUT SIGNAL SEQUENCE -------------------------------------
II DETERMINISTIC INPUT SIGNAL SYSTEM IDENTIFICATION -------------------
II STOCHASTIC INPUT SIGNAL SYSTEM IDENTIFICATION ----------------------
II QUERY USER FOR ANOTHER IDENTIFICATION ------------------------------
ANSWER = 'Y';
WHILE ANSWER = 'Y' , ...
EXECUTE('GETPARAM.EXC'), ...
EXECUTE('DISCRTCL.EXC')~...I
EXECUTE('GENINPUT.EXC'), ...
IF TYPE = 1, ...
EXECUTE('DTRMNRSP.EXC'), ...
ELSE~ ...
EXECUTE('STOCHRSP.EXC'), ...
EJ\lD t •••
INQJIRE Ar\ISWER 'DO YOU WANT TO DO ANOTHER RESPO\ISE . . ,. , ...
II
II
PROGRAM132
----------~-~~~~~~~~-~~~~-~~~~------~~-~~~-~~~~~~-------~--~~~~~~~~~~--~~~~~~~~~~~~~~~~~~~--~~~~~~~~~~~~~~~~~~~~~~~~~~~
I I NAME: GETPARAM
II
I I AUTHOR: T T HAJV1IVtO\I 0
II
II DATE: 6 APR 1988
II
II PURPOSE:
II This file is a MatrixX executable command file. Its purpose is to
II establish values for the parameters necessary to run the thesis
II simulation. Default parameters are set and the user is prompted
I I to change any parameters that re/she desi res .
II
II VARIABLES
II CHOICE
II DELTAT
II NSAMPLES
II PULSAMPLTD
II PULSDURATN
II RECORWR
II SHAPE
II
II
II TYPE
II
II
~-~-~~-~~-~~~~~~~-~~~~---~~-~~--~-~~~~------~~~~--~~-----~~~~~~~~~~~~~~~~~~~~----~~~~~~~~~~~~~~~~~~~~--~~~-~~-~~~~
index of menu se1set;on
sampling period in seconds
number of samples rounded to next higher power of 2
input pulse amplitude (normalized -10 to 10)
length of inpu"t pu l se in sample periods
1ength of time s;gna1s observed in seconds
= for rectangular input pulse
= 2 for binomial triangle shaped input pulse
= 3 for bipolar triangle shaped input pulse
= for deterministic input signal
= 2 for stochastic input signal
133II CONSTANTS ==========================================================BLANK15 = '
II
II SELECTION MENUS:
CHANGEMENU = [' * MENU *
, ACCEPT
'CHANGE '];
II
SHAPEMENU == [t * SHAPE MENU *
RECTANGLE
BINOMIAL TRIANGLE'
BIPOLAR TRIANGLE f];
II
TYPEMENU == I.'* TYPE MENU *
DETERMINISTIC
f STOCHASTIC '];
II
II BEGIN ===============================================================
II
SHORTI
II
II SET DEFAULT PARAMETERS ---------------------------------------------
RECORDUR == 40;
DELTAT = 0.2;
NSAMPLES ::: 2**ROUND(0.5 + LOG(RECORDUR/DELTAT)/LOG(2));
TYPE::: 1;
PULSOORATN = 1;
134SHAPE = 1;
PULSAMPLTD = 1;
II
ERASE
II QUERY USER FOR RECORD DURATION ------------------------------------
OISPLAY( 'CURRENT TIME RECORD DURATIO\I IS') t RECOROOR
PAUSE
CHOI CE = MENU (CHANGEMENU r 1) ;
IF CHOICE = 2, ...
INQUIRE RECORDUR 'ENTER TIME RECORD DURATION IN SECONDS' , ...
RECOROORt • • •
NSAMPLES= 2**ROUND(O.5 + LOG(RECORDUR/DELTAT)/LOG(2)); ...
PAUSE, ...
END
II
ERASE
II QUERY USER FOR SAMPLING PERIOD ------------------------------------
DISPLAY('CURRENT SAMPLING PERIOD IS'), DELTAT
PAUSE
CHOI CE = MENU (CHANGEMENU , 1) ;
IF CHOICE = 2, ...
INQUIRE DELTAT 'ENTER SAMPLING PERIOD IN SECONDS' , ...
DELTAT, ...
NSAMPLES= 2**ROUND(O.5 + LOG(RECORDUR/DELTAT)/LOG(2)); ...
PAUSE, ...
END
II
135ERASE
II QUERY USER FOR DETERMINISTIC OR STOCHASTIC EXCITATION -------------
DISPLAY('CURRENT INPUT SIGNAL TYPE IS')
DISPLAY(TYPEMENU(TYPE + 1,:))
PAUSE
CHO I CE = MENU (CHANGEMENU r 1) ;
IF CHOICE = 2, ...
TYPE = MENU(TYPEMENU,1), ...
DISPLAY(TYPEMENU(TYPE + 1,:)), ...
PAUSE, ...
END
II
II DETERMINISTIC SIGNAL SECTION COMMENTS ==============================
II QUERY USER FOR PULSE DURATION -------------------------------------
II QUERY USER FOR PULSE SHAPE ----------------------------------------
II QUERY USER FOR PEAK PULSE AMPLITUDE -------------------------------
IF TYPE = 1, ...
ERASE, .•.
DISPLAY('CURRENT PULSE DURATION IS'), PULSDURATN, ...
PAUSE, •••
CHOICE = MENU(CHANGEMENU,1); ...
IF CHOICE = 2, ...
INQJIRE PULS{)JRATN 'EJ\ITER PULSE DURATION IN SAMPLE PERIODS T , •••
PULS()JRATN = ROJND(PULSDURATN), ...
PAUSE, ...
END, ...
ERASE, •••
136DISPLAY('CURRENT PULSE SHAPE IS'), .
DISPLAY(SHAPEMENU(SHAPE + 1,:)), .
PAUSE, •••
CHOICE = MENU(CHANGEMENU,1); ...
IF CHOICE = 2, ...
SHAPE = MENU(SHAPEMENU,1), ...
DISPLAY(SHAPEMENU(SHAPE + 1,:)), ...
PAUSE, ••.
END, .
ERASE, .
DISPLAY('CURRENT PULSE AMPLITUDE IS'), PULSAMPLTD, ...
PAUSE, •••
CHOICE = MENU (CHANGEMENU , 1); ...
IF CHOICE = 2, ...
INQUIRE PULSAMPLTD TENTER PEAK PULSE AMPLITUDE (-10 TO 10)T , ...
PULSAMPLTD, ...
PAUSE, .•.
EJ\ID, •••
END
II
II CLEAR UNNECESSARY VARIABLES ---------------------------------------
CLEAR BLANK15 CHANGEMENU TYPEMENU. SHAPEMENU CHOICE RECORDUR
II END OF GETPARAM PROGRAM ============================================
RETURJ\l
137II PROGRAM ============================================================
II
II NAME: DISCRTCL
II
II AUTHOR: T T HAMMOND
II
II DATE: 5 APR 1988
II
II PURPOSE:
II This file is a MatrixX executable command file. Its purpose is to
II create a MatrixX description of "the sampled-data system examined
II in the tt'1esis. Tlie MatrixX description of the analog computer
II realization rrust be available in the file ANCMPSIM.VAR in the OOS
II default directory.
II This program incorporates a stabilizing feedback loop to the input
II using collocated, complementary sensing (i.e. the rotational
II velocity of disk one is the feedback to the torque input of disk
I lone).
II
II VARIABLES
II DA8, DA8C
II DELTAT
II DSA, DS8,
II DSC
II K
/1' KF8
II NAB, NAOC
----------~~~~---------~~-~~~~~~~~~~~~~~-~~~~~-~---------~~~~~~~~~~~-----~~~~~~~------~~~~~~~~~~~~~~~~~~-~~~~-~~~~~
intermediate subsystem descriptions
sampling period in seconds
descriptions of digital subsystems
compensator gain
feedback matrix
intermediate subsystems' orders
138II NCL
II NSA,. NS8,
II NSC
II SA, S8, SC
II
II SCL
II
order of c Iosed-Toop system description
integers indicating the subsystems' orders
matrices composed of coefficient matrices to form
,the MatrixX represerttations of the subsystems
closed-loop system IVIatrixX descriptiQli
II CONSTANTS ==========================================================I I LOOP CLOSING PARAMETERS:
INGAIN = [1 1 1J';
K = -1;
KFB = K*[O 0 -1.9909 0 -3.4905E-3; .
o 0 -1.9909 0 -3.4905E-3; .
o 0 -1.9909 0 -3.4905E-3];
II
II BEGIN ==============================================================
II CLEAR SCREEN AND DISPLAY PROGRAM FUNCTION -------------------------
ERASE
DISPLAy(tDISCRETIZING THE CLOSED-LOOP SYSTEM ... T)
SHORT E
II RETRIEVE SIMULATED SYSTEM AND DISCRETIZE EACH SUBSYSTEM -----------
LOAD(tANCMPSIM.VAR t)
DSA = DISCRETIZE(SA, NSA, DELTAT);
058 = DISCRETIZE(SB, NS8, DELTAT);
DSC = DISCRETIZE(SC, NSC, DELTAT);
II CLOSE DISK-1 VELOCITY FEEDBACK LOOP -------------------------------
[DAB, NABJ = APPEND(DSA, NSA, DSB, NSB);
139[DABC, NA8C] = APPEND(DAB, NAB, DSC, NSC);
[SCL, NCLJ = CONNECT(DABC, NABe, KFB, INGAIN);
II CLEAR UNNECESSARY VARIABLES --------------~------------------------
CLEAR DAB DABC DSA DSB DSC KFB NAB NABC NSA NS8 NSC SA S8 SC
II STORE CLOSED-LOOP DISCRETE SYSTEM ON DISKETTE ---------------------
SAVECtDISCRTCL.VAR t)
II END OF DISCRTCL PROGRAM =========================================--==RETURN
140II P~RAM ============================================================
II
II NAME: GENINPUT
II
II AUTHOR: T T HAMMOND
II
II DATE: 7 APR 1988
II
II PURPOSE:
II This file is a MatrixX executable command file. Its purpose is to
I I gerterate tile inpu"t vector TAU for the digita1 system simulation.
II The input signal is a rectangular pulse, a binomial triangle
II shaped pulse, or unity white noise signal as defined by the user
II in the calling program.
II
II VARIABLES
II CHOICE
II COLUMNOX
II DELTAT
II LEVEL
II NEXTPASCAL
II NOISEQ
II NSAMPLES
1/ PPS::,AL
~~~~~---------~~~~~---~~~~~~~~~~~--~~~~~~~~~~~~~~~~~~-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
index of menu selection
column index counter
sampling period in seconds
level of Pascal's triangle
next level of Pascal's triangle
noise sequence
number of samples rounded to next higher power of 2
level of Pascal's triangle
II PULSAMPLTD input pulse amplitude (normalized -10 to 10)
II PULSDURATN length of input pulse in sample periods
II SHAPE = 1 for rectangular input pulse
141
determinis"tic vector input signal ~ or
= 2 for stochastic inpu"t signal
= 2 for binomial triangle shaped input pulse
= 3 for bipolar triangle shaped input pulse
stochastic matrix of vector input noise signals
= for deterministic input signal
II
II
II TAU
II
II TYPE
II
II
II BEGIN ==--===--=======================================================
II CLEAR SCREEN AND DISPLAY PROGRAM FUNCTION -------------------------
ERASE
DISPLAy(tGENERATING SYSTEM INPUT SIGNALS ... f)
II SIGNAL GENERATION COMMENTS =========================================
II GENERATE DETERMINISTIC SIGNAL -------------------------------------
II GENERATE RECTANGULAR SIGNAL ---------------------------------------
II GENERATE SHAPED SIGNAL --------------------------------------------
II SCALE PEAK MAGNITUDE OF DETERMINISTIC SIGNAL -----------------------
II ALTERNATE SIGNS FOR BIPOLAR TRIANGLE ------------------------------
II GENERATE PSEUDO-RANDOM UNITY WHITE NOISE SEQUENCE SIGNAL ----------
IF TYPE ;: 1~ ...
I F SHAPE = 1 ~ ...
TAU ;: [PULSAMPLTD*ONES(1 ~PULSDURATN)J T; •••
ELSE PASCAL = 1; •••
LEVEL;: 1; ...
ltt-tILE LEVEL < PULSruRATN, ...
NEXTPASCAL = [1 O*ONES(1~LEVEL-1) 1J; ...
COLUM\IOX = 2; •••
If.tiILE COLUM\lOX < LEVEL+1 ~ ...
142NEXTPASCAL(1,COLUMNDX) = PASCAL(1,COLUMNDX-1)+ ...
PASCAL(1,COLUMNDX); ...
COLUMNDX = COLUMNDX+1; ...
END, •••
PASCAL = NEXTPASCAL; ...
LEVEL = LEVEL +1; ...
END, •••
TAU = PULSAMPLTD*[PASCAL/MAX(PASCAL)Jf; ...
I F SHAPE = 3, ...
FOR COLUMI\IDX = 1: PULSDURATN , ...
TAU(COLUMNDX,1)=TAU(COLUMNDX,1)*(-1)**(COLUMNDX-1); ...
EJ\ID; •••
EJ\ID; •••
END, .••
TAU = [TAU; O*ONES(NSAMPLES-PULSDURATN, 1)J; ...
ELSE RAND(tNORMAL t) , •••
TAU = RAND(NSAMPLES, 1); ...
END
II
II CLEAR. UNNECESSARY VARIABLES ---------------------------------------
CLEAR COLUMNDX
CLEAR LEVEL NEXTPASCAL NOISEQ PASCAL PULSAMPLTD PULSDURATN SHAPE
ERASE
II END OF GENINPUT PROGRAM ============================================RETURN
II
II
PROORAM143
--~~~~~~-~~~~--~~-~~~~~--~-~~--~--~-~~~~~---------~~~~~~~~-~~~~~~~~~~~~~~~~~~~-~~~~~~~~~~~~~~~~~~~~~~~~--~~~~~~~~~~~~~~~
II NAJVIE: DTRMNRSP
II
II AUTHOR: T T HAMVIOND
II
II DATE: 9 APR 1988
II
II PURPOSE:
II This file is a Matrix)( executable command file. Its purpose is to
I I compute the time and frequency responses for the exper i menta1
II system simulation due to deterministic inputs. Also the desired
I I frequency response function is extracted from the c Iosed-Toop
II configuration.
II The system description variables and input signal must be provided
II by the calling program.
II
II VARIABLES
II DELTAT
II FHZ
II G5FEST
II K
II IVIODES
II NCL
II NOISEQ
II NSAMPLES
II NSIM
--~~~~~~~~--~~--~-~~-~~-~~~~~-~-~~~--~~-~~~~---~~~----~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sampling period in seconds
frequency in Hz
es"timated disk 3 velocity frequency response function
comper1sator gain
rncda1 responses
order of closed-lcx:>p system descriptiQr1
noise sequence
number of samples rounded to next higher power of 2
discrete simulatiQr1 index vector
144II SCL closed-loop system MatrixX description
II TAU deterministic vector input signal~ or
II stochastic matrix of vector input noise signals
II TIME time in seconds
II VEL1CL disk closed-loop velocity
II VEL1CLF disk c1osed-loop veloci ty f r-equency s pect rum
II VEL3CL disk 3 closed-loop velocity
II VEL3CLF disk 3 closed-loop velocity frequency spectrum
II
II BEGIN ==============================================================
II CLEAR SCREEN AND DISPLAY PROGRAM FUNCTION -------------------------
ERASE
DISPLAY('COMPUTING SYSTEM RESPONSES ... ')
II GENERATE TIME RESPONSES --------------------------------------------
[NSIM, MODESJ = DLSIM(SCL, NCL, TAU);
TIME = NSIM*DELTAT;
CLEAR NSIM
II DISPLAY INPUT SIGNAL AND DISK-3 RIGID BODY MODES ------------------
PLOT(TIME, [TAU MODES(:,[1 2J)J, ...
'TITL/DETERMINISTIC INPUT AND DISK-3 RIGID BODY MODES/ ...
YLAB/MAGNITUDE/ ...
XLAB/TIME IN SECONDS/')
PAUSE
ERASE
II DISPLAY DISK-3 LOW FREQUENCY MODES --------------------------------
PLOT(TIME, [TAU MODES(:,[3 4J)J, ...
'TITL/DETERMINISTIC INPUT AND DISK-3 0.1 HZ MODES/ ...
145YLAB/MAGNITUDE/ ...
XLAB/TIME IN SECONDS/')
PAUSE
ERASE
II DISPLAY DISK-3 HIGH FREQUENCY MODES -------------------------------
PLOT(TIME, [TAU MODES(:,[5 6J)J, ...
fTITL/DETERMINISTIC INPUT AND DISK-3 1.5 HZ MODES/ ...
YLAB/MAGNITUDE/ ...
XLAB/TIME IN SECONDS/ r )
PAUSE
ERASE
CLEAR TIME
I/next section added to find max values for normalization==============
I/P3L=MAX(A8S(MODES(:,3)))
/IPAUSE
I/P3H=MAX(A8S(MODES(:,5)))
I/PAUSE
//V3R=MAX(A8S(MODES(:,2)))
IIPAUSE
//V3L=MAX(A8S(MODES(:,4)))
//PAUSE
//V3H=MAX(A8S(MODES(:,6)))
//PAUSE
I/A3L=MAX(A8S(-O.335*TAU-O.395*MODES(:,3)))
//PAUSE
I/A3H=MAX(A8S(O.0015*TAU-88.8269*MODES(:,5)))
//PAUSE
146Ilend first max values section
II DELETE POSITION RESPONSES FOR MEMORY CONSERVATION -----------------
MODES = [MODES(:,2) MODES(:,4) MODES(:,6)];
II FORM DISK-1 AND DISK-3 VELOCITY SIGNALS ---------------------------
VEL1CL = [MODES(:,1)-1.9909*MODES(:,2)-3.4905E-3*MODES(:,3)];
VEL3CL = [MODES(:,1)+MODES(:,2)+MODES(:,3)];
Iisecond max values section begins
1/V1CL=MAX(ABS(VEL1CL))
IIPAUSE
1/V3CL=MAX(ABS(VEL3CL))
IIPAUSE
Ilend second max values section
II CLEAR SUBSYSTEM TIME RESPONSES FOR MEMORY CONSERVATION ------------
CLEAR MODES
II PERFORM FFTs ON INPUT SIGNAL, AND DISK-1 AND DISK-3 VELOCITIES ----
TAUF = FFT(TAU, TRECT T);
CLEAR TAU
TAUF = [TAUF([2:(NSAMPLES/2)],1)J;
VEL1CLF = FFT(VEL1CL, 'RECT T) ;
CLEAR VEL1CL
VEL1CLF = [VEL1CLF([2:(NSAMPLES/2)],1)];
VEL3CLF = FFT(VEL3CL, TRECT T) ;
CLEAR VEL3CL
VEL3CLF = [VEL3CLF([2:(NSAMPLES/2)],1)];
I I COMPUTE DISK-3 VELOCITY OPEN-LOOP FREQUENCY RESPONSE FUNCTION -----
G5FEST = [VEL3CLF./CTAUF+K*VEL1CLF)];
FHZ = [1:(NSAMPLES/2)-1]f/(NSAMPLES*DELTAT);
147// DISPLAY INPUT SIGNAL SPECTRUM --------------------------------------
PLOT(FHZ~ 20*LOG(A8S(TAUF))/LOG(10), ...
fTITL/DETERMINISTIC INPUT FREQUENCY SPECTRUM! ...
YLAB/dB! ...
XLAB/FREQUENCY IN HZ/...
LOGX t)
PAUSE
ERASE
II DISPLAY DISK-1 CLOSED-LOOP VELOCITY SPECTRUM -----------------------
PLOT(FHZ 7 20*LOG(A8S(VEL1CLF))/LOG(10), ...
fTITL/DISK-1 CLOSED-LOOP VELOCITY FREQUENCY SPECTRUM/ ...
YLAB/dB/ ...
XLAB/FREQUENCY IN HZ/...
LOGX t)
PAUSE
ERASE
II DISPLAY DISK-3 CLOSED-LOOP VELOCITY SPECTRUM -----------------------
PLOT(FHZ~ 20*LOG(A8S(VEL3CLF))/LOG(10), ...
fTITL/DISK-3 CLOSED-LOOP VELOCITY FREQUENCY SPECTRUM/ ...
YLA8/d8/ ...
XLAB/FREQUENCY IN HZ/...
LOGX t)
PAUSE
ERASE
1/ DISPLAY DISK-3 OPEN-LOOP VELOCITY FREQUENCY RESPONSE FUNCTION ------
PLOT(FHZ~ 20*LOG(A8S(G5FEST))/LOG(10), ...
TTITL/DISK-3 OPEN-LOOP VELOCITY FREQUENCY RESPONSE FUNCTION/ ...
148YLAB/d8/ ...
XLAB/FREQJENCY IN HZ/ ..•
LOGX T)
PAUSE
ERASE
II CLEAR UNNECESSARY VARIABLES ---------------------------------------
CLEAR DELTAT FHZ G5FEST K NCL NSAMPLES SCL TAU TAUF
CLEAR TYPE VEL1CL VEL1CLF VEL3CL VEL3CLF
II END OF DTRMNRSP PROGRAM ============================================RETURJ\I
II PROGRAM149
~~~~~~~~-------~~~~~~~~~~~--~-~~~~~~~~~~~~~~~~~~~--~~~-----~~~~~~~~~~~~~~~~~~--~~~~~~~~~~~~~~--~~~~~~~~~~~~~~~~~~~~~~--
II
II NAME: STOCHRSP
II
II AUTHOR: T T HAMMOND
II
II DATE: 10 APR 1988
II
II PURPOSE:
II This file is a MatrixX executable command file. Its purpose is to
I I ccmout.e the time and frequency responses for the exper trnerrce 1
I I system sirrulation due to stochastic inputs . Also the desired
I I frequency response funct i on is extracted from the C 1osed-l oop
II configuration.
II The system description variables and input signal must be provided
II by the calling program.
II
II VARIABLES
II DELTAT
II FHZ
II G5FEST
II K
II MODES
II NCL
II NOISEQ
II NSAMPLES
II NSIM
-~~~~-~---~~~-----~----~~~-~-----~~~~~-----------~~-~----~~~~~~~~~~~~--~~~~~~~-~~~~~~~~---~~~~~~~~~~~~~~~-~~~~~~~-~
sampling period in seconds
frequency in Hz
estimated disk 3 velocity frequency response function
compensator gain
mcxJa 1 responses
order of closed-loop system description
noise sequence
number of samples rounded to next higher power of 2
discrete simulation index vector
150II SCL
II TAU
II
II TIME
II VEL1CL
II VEL1CLF
II VEL3CL
II VEL3CLF
II
closed-loop system MatrixX description
deterministic vector input signal, or
stochastic ma"trix of vector input noise signals
time ; n seconds
disk closed-loop velocity
disk closed-loop velocity frequency spectrum
disk 3 closed-loop velocity
disk 3 closed-loop velocity frequency spectrum
II BEGIN ==============================================================
II CLEAR SCREEN AND DISPLAY PROGRAM FUNCTION -------------------------
ERASE
DISPLAy(tCOMPUTING SYSTEM RESPONSES ... t)
II GENERATE TIME RESPONSES --------------------------------------------
[NSIM, MODESJ = DLSIM(SCL, NCL, TAU);
TIME = NSIM*DELTAT;
CLEAR NSIM
II DISPLAY INPUT SIGNAL AND DISK-3 RIGID BODY MODES ------------------
PLOT(TIME, [TAU MODES(:,[1 2J)J, ...
tTITL/STOCHASTIC INPUT AND DISK-3 RIGID BODY MODES/ ...
YLA8/MAGNITUDE/ ...
XLAB/TIME IN SECONDS/ f)
PAUSE
ERASE
II DISPLAY DISK-3 LOW FREQUENCY MODES --------------------------------
PLOT(TIME, [TAU MODES(:,[3 4J)J, ...
fTITL/STOCHASTIC INPUT AND DISK-3 0.1 HZ MODES/ ...
151YLAB/MAGNITUDE/ ...
XLAB/TIME IN SECONDS/ T)
PAUSE
ERASE
// DISPLAY DISK-3 HIGH FREQUENCY MODES -------------------------------
PLOT(TIME~ [TAU MODES(:,[5 6])], ...
fTITL/STOCHASTIC INPUT AND DISK-3 1.5 HZ MODES/ ...
YLAB/MAGNITUDE/ ...
XLAB/TIME IN SECONDS/f)
PAUSE
ERASE
CLEAR TIME
Iinext section added to find max values for normalization===============
I/P3L=MAX(ABS(MODES(:~3)))
//PAUSE
//P3H=MAX(A8S(MODES(:,5)))
//PAUSE
//V3R=MAX(A8S(MODES(:,2)))
I/PAUSE
//V3L=MAX(A8S(MODES(:~4)))
//PAUSE
//V3H=MAX(A8S(MODES(:,6)))
//PAUSE
//A3L=MAX(ABS(-O.335*TAU-O.395*MODES(:,3)))
//PAUSE
//A3H=MAX(A8S(O.0015*TAU-88.8269*MODES(:~5)))
//PAUSE
152Ilend first max values section
II DELETE POSITION RESPONSES FOR MEMORY CONSERVATION -----------------
MODES = [MODES(:,2) MODES(:,4) MODES(:,6)];
II FORM DISK-1 AND DISK-3 VELOCITIES ---------------------------------
VEL1CL = [MODES(:,1)-1.9909*MODES(:,2)-3.4905E-3*MODES(:,3)];
VEL3CL = [MODES(:,1)+MODES(:,2)+MODES(:,3)];
Iisecond max values section begins
I/V1CL=MAX(A8S(VEL1CL))
IIPAUSE
//V3CL=MAX(A8S(VEL3CL))
I/PAUSE
Ilend second max va 1ues section
II CLEAR TIME RESPONSES FOR MEMORY CONSERVATION ----------------------
CLEAR MODES
II FORM FEEDBACK ERROR SIGNAL ----------------------------------------
U = TAU + K*VEL1CL;
II COMPUTE AUTo-PSD OF INPUT SIGNAL -----------------------------------
[TAUF, FIN] = SPECTRUM(TAU, TAU);
CLEAR TAU
TAUF = [TAUF([(NSAMPLES/2+2):(NSAMPLES)])];
II COMPUTE AUTo-PSD OF DISK-1 VELOCITY --------------------------------
[VEL1CLF, FIN] = SPECTRUM(VEL1CL, VEL1CL);
CLEAR VEL1CL
VEL1CLF = [VEL1CLF([(NSAMPLES/2+2):(NSAMPLES)])];
II COMPUTE CROSS-PSD OF ERROR SIGNAL AND DISK-3 VELOCITY -------------
[UVEL3CLF, FIN] = SPECTRUM(VEL3CL, U);
CLEAR VEL3CL U
153UVEL3CLF = [UVEL3CLF([(NSAMPLES/2+2):(NSAMPLES)])];
II COMPUTE DISK-3 OPEN-LOOP VELOCITY FREQUENCY RESPONSE FUNCTION -----
G5FEST = [UVEL3CLF./(TAUF+K*VEL1CLF)];
FHZ = FIN([(NSAMPLES/2+2):(NSAMPLES)])/(DELTAT);
II DISPLAY INPUT SIGNAL SPECTRUM --------------------------------------
PLOT(FHZ, 20*LOG(ABS(TAUF))/LOG(10), .
'TITL/STOCHASTIC INPUT AUTO-PSD/ .
YLAB/dB/ ...
XLAB/FREQUENCY IN HZ/...
LOGX f)
PAUSE
ERASE
II DISPLAY DISK-1 CLOSED-LOOP VELOCITY SPECTRUM ----------------------
PLOT(FHZ, 20*LOG(A8S(VEL1CLF))/LOG(10), ...
'TITL/DISK-1 CLOSED-LOOP VELOCITY AUTO-PSD/ ...
YLAB/dB/ ...
XLAB/FREQUENCY IN HZ/...
LOGX t)
PAUSE
ERASE
1/ DISPLAY DISK-3 CROSS-PSD ------------------------------------------
PLOT(FHZ, 20*LOG(A8S(UVEL3CLF))/LOG(10), ...
'TITL/ERROR SIGI\IAL AND DISK-3 CLOSED-LOOP VELOCITY CROSS-PSD/ ...
YLAB/dB/ ...
XLA8/FREQUENCY IN HZ/...
LOGX T)
PAUSE
154ERASE
II DISPLAY DISK-3 OPEN-LOOP VELOCITY FREQUENCY RESPONSE FUNCTION -----
PLOT(FHZ, 20*LOG(A8S(G5FEST))/LOG(10), ...
'TITL/DISK-3 OPEN-LOOP VELOCITY FREQUENCY RESPONSE FUNCTION/ ...
YLAB/dB/ ...
XLAB/FREQUENCY IN HZ/...
LOGX')
PAUSE
ERASE
II CLEAR UNNECESSARY VARIABLES ---------------------------------------
CLEAR DELTAT FHZ G5FEST K NCL NSAMPLES SCL TAU TAUF
CLEAR TYPE VEL1CL VEL1CLF VEL3CL VEL3CLF
II END OF STOCHRSP PROGRAM ============================================R8U~
156
This appendix contains the source code listings for the
hybrid simulation of Chapter Five. The programs are written
in HP BASIC for use on a HP Model 9000 computer.
The programs listed perform three major tasks. The SIM
ULATION program drives the hybrid simulation to collect the
time data for the identification process. The SIUDI (System
Identification Using Deterministic Inputs) program performs
frequency analysis and response function computation for
deterministic analysis. The SIUSI (System Identification
Using Stochastic Inputs) program provides the corresponding
functions for stochastic analysis.
15710 PROGRAM ======================================================
20 NaMe: SIMULATION
30 Author: T T HAMMOND
40 Date: MAY 1988
50 Purpose: This prograM drives the three-dIsk saMpled-data
60 control sY5teM 5iMulation. First the HP-3852A Data Acquisi-
70 tion/Control Unit is readied by initializing the HP-44702A
80 HIgh Speed VoltMeter and HP-44727A Olgital to Analog Voltage
90 Converter, and downloading the data acquisition/control
100 routine. Next the user is queried for the Input 5Ignal flle-
1 10 naMe and the record nUMbering sequence for the archIve of tlMe
120 records. At thIS pOInt) the Data AcquisitIon/Control UnIt 15
130 cOMManded to run the siMulatIon, and this prograM concurrently
140 displays a strip chari view of the lnput signal and the
150 Measured velocIties of disk one and disk three. DurIng the
160 siMulation, the user is proMpted to operate the analog
170 COMputer and confIrM these actions. FInally, the data col-
180 lected 15 stored on diskette in the approprlate archive} and
190 the prograM recycles to begin another siMulatIon If the user
200 desires.
210
220 SUBPROGRAMS ==================================================
230 The following subprograMs froM the HP Data AcquI5ition Manager
240 OACQ/300 package are used:
280
290
300
310
330
340
Y_M.in ,Y_Ma>·< ,[ ,Buf_5ize[ ,Strlp_buf$J J)
158
350
360
37G
390
400
410
420
Arch_open[ . Ar-ch i ve name s I ,Arch.i'lesize[ ,Last_pageslze ] J
Page_store(BooknaMe$( ,Inst_addr[ ,Pagelabel$[,
Des t inat ion$]] ] )
430 VARIABLES ====================================================
440 ! J)e.l1 ,')e13 d i s k r ! , disk-3 ve l c c i t i e s In radians/sec
450
460 ! Scandata
analog COMputer norMall:lng voltage ln volts
raw d r s k r ] ,disk-3 vo l t ne t er- readings in v o l t s
470
480
I Insignal input signal sequence in radians/sec
chart abSCIssa in seconds
490
500
520
530
Last rQr'
i Record_count
saMple nUMber Index
first record nUMber to append to archive
last record nUMber to append to archIve
record nuno e.r i ride x
strip chart graph record
540
550
560
570
! Stri.p_count
! SaMpling
! InsQnl$
strip chart saMple nUMber index
=1 while s anp l i no data
=0 c t her u i s e
naMe of input signal vector
159
5812) More$
590 I Title$
600 ! Sub titleS
6i0 ! Trace_labels$
620 ! X label$
630 Y' Lab e l s
640
query response
5trip chart title
5trip chart 5ubtitle
strtp chart trace labels
strlp chart abscissa label
strIp c har t ordlnate label
650 REAL Vell(0:1023),Ue13(0:1023),NorM_volt
560 REAL Scandata(0: 1)
670 REAL Insignal(0:1023)
710 OIM Insgnl$[10J ,More$(3J
7:0 DIM Tltle$[40J ,-Sub_title$[40J ,Tr3ce_labels$[40J
730 OI~l X_label$[ 20 J ,Y_label$[ 20]
740 CONSTANTS ====================================================
750
760
770
! Nor-nve l t
I NorMve13
! NorMsgnl
disk-l velocity norMallzlng constant
disk-3 velocity norMalizing constant
input signal norMalizing constant
790 REAL Nor-mve l l .Nornve l S .Nor-n s qn I
800 NorMvell=5
160
810 Nornv e 13-=5
820 NorMsgnl=10
830 BEGIN ========================================================
840 INITIALIZE DATA ACQUISITION SUBPROGRAMS ----------------------
850 Sys_inlt
860 CONFIGURE HP-3852A HARDWARE ----------------------------------
870 OUTPUT 709;"RST"
880 OUTPUT 709; JIIJSE 200"
890 OUTPUT 709; iI CONF DC\) /I
90!~ OUTPUT 709 j II PACER 0.200 J 1()24 iI
9110 C)UTPt.JT 709; II PDELAY 0.200 It
920 OUTPUT 709. IlSWRITE 100, 2,13"
930 OUTPUT 709;IISIJJRITE 100,6,00,00 11
940 UNMASK HP-3852A SOFTWARE INTERRUPT ---------------------------
95fb OUTPUT 709; II RQS FPS 11
960 DECLARE HP-3852A VARIABLES -----------------------------------
970 OUTPUT 709; II REAL C:NTRLERR, I NS I GNAL ( 11:Q23 ) II
980 ;JUTPUT 709 ; iJ REAL 'S GNLSeAL, t) EL t 1(02 3 ) 1 l.)EL3 ( 1023 ) Ii
990 OUTPUT 709; "REAL SCANDATA( 1 ) ~ FBGAIN II
1000 OUTPUT 709;;1 I NTEGER COUNT II
1010 SET FEEDBACk GAIN --------------------------------------------
1020 OUTPUT 709; /I FBGAIN = 1 II
1030 HP-3852A DATA ACQUISITION AND CONTROL ROUTINE ----------------
1040 OUTPUT 709;"SUB CNTRLOOP"
1;J50 OUTPUT 709;!' FOR COUNT=0 TO 1023"
1060 OUTPUT 709;IlCNTRLERR=0.667*SGNLSCAL*INSIGNAL(COUNT)-0.333*FBGAIN*S
161
CANOATA(0)"
1070 OUTPUT 709; II INDEX SCANDATA 0 11
1080 OUTPUT 709; III;jAITFOR PACERII
1090 OUTPUT 709;IIMEAS DCV, 401-412'2 INTO SCANDATA"
1100 OUTPUT 709; 11 APPLY DCIJ 100, CNTRLERR"
1110 OUTPUT 709;lIlJEL1(COUNT)=SCANDATA<0)"
112i~ OUTPUT 709; II VEL3 ( COUNT) =SCANOATA ( 1 ) II
1130 OUTPUT 709; II I F COUNT 100 THE~r'
1 i 40 OUTPUT 709; II ~JREAD SCANDATA II
1150 OUTPUT 709; !lEND IFII
1160 OUTPUT 709;"NEXT COUNTlI
1170 OUTPUT 709;IlSRQII
118f~ OUTPUT 709; "SUBENO"
1190 DEFINE INTERRUPT SERVICE ROUTINE -----------------------------
1200 ON INTR 7 GOSUB Service_routine
1210 LOOP WHILE MORE SIGNALS TO BE APPLIED ------------------------
1230 t~ HI LE i10r e $ [ 1 , 1 J=1\ Y"
1240 GET uSER SUPPLIED INFORMATION --------------------------------
1250 BEEP
1260 OISP "ENTER NAME OF INPUT SIGNAL '.JECTOR, e.g. RT1Xl_\)EC 11
1270 LINPUT II 1'1Insgnl$
12B0 BEEP
1290 OISP IlENTER FIRST RECORD NUMBER";
1310 BEEP
1320 DISP "ENTER LAST RECORD NUMBER";
162
1330 INPUT II II ,Last_rec
1340 IF Insgnl$<>·IINOISE_RECIl THEN
1350 Get_rarray(Insgnl$,Insignal<*»
1360 MAT Insignal= Insignal/(NorMsgnl)
1370 END IF
1380 LOOP FOR EACH RECORD -----------------------------------------
1400 PARSE NOISE SIGNAL -------------------------------------------
1410 IF Insgnl$=ilNOISE_REC" THEN
E_UEC" )
1430 Get_rarray( "NOISE_I)EC" ,Insignal( *»
1440 MAT Insignal= Inslgnal/(NorMsgnl)
1450 END IF
1460 INITIALIZE STRIP CHART ---------------------------------------
1470 Strip_init(Strip_rec(*),3,0,20,1 ~-10,10)
150 0 Tit 1e $ =il Thr e e - 0 i 5 k S i MU 1a t ion II
15i0 Sub_title$=lISa Mp l e d- da t a Strip Char t "
1530 Y_label$="Radians per second"
1540 Trace_labels$=IIInput:Vel-t :Vel-3 11
1560 OUTPUT ~~BD;CHR$(255)&Jlf<";
163
1570 GRAPHICS ON
1580 Graph_show(Strip_rec(*))
1590 INITIALIZE HP-3852A VARIABLES --------------------------------
1600 OUTPUT 709; "0ISP OFF"
1610 OUTPUT 709;J1SCANDATA(0)=0 11
1620 OUTPUT 709;/IINDEX SCANDATA, 0 11
1630 OUTPUT 709;"INOEX INSIGNAL, 0"
1640 FOR SaMple_count=0 TO 1023
1650 OUTPUT 709;1I\)WRITE INSIf3NAL "&VAL$(Inslgnal(SaMple_count»)
1670 ENABLE INTERRUPT FOR HP-3852A FINISH -------------------------
1680 SaMpling=1
1690 ENABLE INTR 7;2
1700 PROMPT USER TO ACTIVATE ANALOG COMPUTER ----------------------
1710 BEEP
1720 BEEP
1730 DISP llACTIIJATE ANALOG COMPUTER AND PRES'S RETURN .. II,
174 0 LIN PUT II II ,N u 1$
1750 DISPLAY NUMBER OF RECORD BEING RECORDED ----------------------
1760 DISP IIRECORDING RECORD NUMBER ";Record_count
1770 MEASURE ANALOG COMPUTER NORMALIZING VOLTAGE ------------------
1780 OUTPUT 709; II MEAS DCtJ, 403 INTO S.GNLSCAL II
1790 OUTPUT 709; II ()READ SGNLSCAL"
1800 ENTER 709;NorM_volt
1810 =TART HP-3852A ROUTINE ---------------------------------------
1820 OUTPUT 709;"PTRIG"
1830 OUTPUT 709;"CALL CNTRLOOplI
164
1840 ADO DATA POINTS TO STRIP CHART FOR 100 SAMPLES ---------------
1850 FOR Strip_count=0 TO 99
1860 ENTER 709;Scandata(*)
1880 S t r i p_p 0 i n t 5 ( S t. r i p_rec ( * ) ,S t r i p_ t i ne , II 1 : 2 : 3 II .Nor n 5 gn 1* In 5 i gn
al (Strip_count ) .Normve l l *Scandata( 0) ,NorMve13*Scandata( 1 »
1890 NEXT Strip_count
1900 WAIT UNTIL HP-3852A FINISHES AND INTERRUPTS ------------------
1910 WHILE SaMpling=l
1920 END WHILE
1930 PROMPT USER TO RESET ANALOG COMPUTER -------------------------
1940 BEEP
1950 BEEP
1 1360 BEEP
1!370 0ISP II SA~1PL I NG COMPLETED: RESET ANALOf3 COr1PUTER II
1980 GET RAW UELOCITY MEASUREMENTS FROM HP-3852A ------------------
1990 OUTPUT 709; IIlJREAO UEL1 II
2000 ENTER 709;Ve11 *)
2010 OUTPUT 709;IIVREAD VEL3 u
2020 ENTER 709;Ve13(*)
2030 CONVERT UELOCITIES TO RADIANS/SEC ----------------------------
2040 ~1AT \Jell = (NorMvell )*\)e11
2050 MAT Vell= Vell/(NorM_volt)
2060 MAT Ve13= (NorMve13)*Ve13
2070 MAT Vel3= Ve13/(NorM_volt)
2080 STORE VELOCITIES IN ARCHIVE ----------------------------------
165
2090 Save_rvect (VeIl (*) ,1I0I51<_1 VEL II)
2110 Arch_open( Insgnl$[ 117J&"ARC
Il)
2140 ACTIVATE HP-3852A DISPLAY ------------------------------------
2 150 0UTPUT 709; II 0 I SP 0N II
2160 END OF LOOP FOR EACH RECORD ----------------------------------
2170 NEXT Record_count
2130 QUERY USER FOR MORE SIGNALS TO BE APPLIED --------------------
2190 BEEP
2200 OISP -oc YOU WISH TO 00 ANOTHER INPUT SI(3NAL? (Y or N)";
2210
2220
2230
2240
2270
2230
2290
2300
2310
2320
L I NPUT II II ~ More$
END OF LOOP FOR MORE SIGNALS ---------------------------------
END WHILE
STOP
HP-3852A INTERRUPT SERVICE ROUTINE ---------------------------
ServIce_routine:
Ser_poll=SPOLL 709)
OUTPUT 709;IlSTA?"
ENTER 709;Stat_word
SaMpling=0
RETURN
END OF HP-3852A INTERRUPT SERVICE ROUTINE --------------------
::30 END OF SIMULATION PROGRAM :===================================
2340 END
10 PROGRAM ===================================================)=~ 6
20 NaMe: SIUDT (SysteM Identification U5ing DeterMinistic Input)
30 Author: T T HAMMOND
40 Date: MAY 1988
50 Purpose: This prograM COMputes and averages the saMpled-data
60
70
80
90
frequency re5ponse function for disk-3 velocity for deterMin
Istic excitations. First a raM disk is established for COMpU-
tational efficiency. Next the U5er is querIed for input
filenaMe and nUMber of records to average. Then the FFT of
100 the input signal is perforMed. Now FFTs for the disk-1 and
1 0 disk-3 ~elocitie5 for each record to be averaged are perf8r~ed
120 and sUMMed. Finally the average disk-3 open-loop frequency
130 response function is calculated and stored on diskette.
140
150 SUBPROGRAMS ==================================================
160 The following subprograMs froM the HP Data Acqu151tlon Manager
170 DACQ/300 package are used:
200 Arch_query(ArchivenaMe$~BooknaMe$,Pagelabel$,FieldnaMe$~
:10 Record$1Results$)
230
240 The following subprograMs appear at the end of this prograM:
250 FNMag_db(Real_part ,IMaQ_part,Nr_saMples)
270
167
280 VARIABLES ====================================================
290 I SaMple_rec$ input 5ignal filenaMe
310
320
330
I Nr_records
! All zeros
/ Frequencies
nUMber of records in average
nUMber of records to average
array of zeros
FFT frequency array
350 ! Disk_l,iMag real and iMaginary parts of FFT for velocity
370 ! Disk_3iMc.g resl and iMaginary parts ~f FFT for ~elaclt ~
390 real and IMaginary SUM5 of FFTs for velOCIty
410 real and IMaginary SUM5 of FFTs for velOCIty j
430 real and iMaginary parts of FFT for lnput
real and iMaginary parts of denOMinator
460
470
480
490
500
510
52C0
OenOM faz
NUMer fa::
denOMInator MagnItude In dB
denOMinator phase in degrees
nUMerator Magnitude In dB
nUMerator phase in degrees
frequency response function Magnitude in dB
frequency response function phase in degrees
168
540 REAL Nr_records
550
560 CONSTANTS ====================================================
570 ! Nr_saMples nUMber of saMples
580 I Nr sectors nUMber of raM disk sectors
saMple period in seconds
650 SaMple_per=.2
660
670 BEGIN ========================================================
680 INITIALIZE DATA ACQUISITION SUBPROGRAMS ----------------------
700 SET UP RAM DISK FOR EFFICIENT EXECUTION ----------------------
710 INITIALIZE ":MEMORY,f~,14"1Nr_5ector5
720 QUERY USER FOR INPUT SIGNAL NAME AND NUMBER OF RECORDS TO ----
730 AVERAGE ------------------------------------------------------
74.0 BEEP
750 OISP ~ENTER NAME OF INPUT SIGNAL VECTOR USED FOR SYSTEM EXCITATION
760 LINPUT 1111 ,SaMple_rec$
770 BEEP
780 0ISP II ENTER NUMBER OF RECORDS TO I NCLUOE IN At)ERAGE II ;
790 INPUT 1111 ,Nr_records
169
800 GENERATE ZERO VECTOR AND FREQUENCY VECTOR --------------------
810 ALLOCATE REAL All_zero5( 1 :Nr_sBMples),Frequencie5( 1 :Nr_saMples)
820 FOR 1=1 TO Nr_saMples
840 Fr euue nc i e e t I )=( I-l )/( SaMple_per*Nr_saMples)
850 NEXT I
860 Save_rvec t ( AII_zero5 ( *' ) , II ALL_ZEROS: r1EMORY ~ fD ,14 II )
870 Save_r'v'ec t ( Frequenc i e s ( * ) , II FREQ_I')ECT: ,700 , 1 II )
880 DEALLOCATE Frequencies(*)
890 PERFORM FFT ON INPUT SIGNAL ----------------------------------
900 USE RECTANGULAR WINDOW -------------------------------------
REAL: MEMORY) 0 ~ 14: INPUT_I MAG:MEMORY, 0 ~ 14" , II NIf )
920 SET UP LOOP TO PROCESS NUMBER OF RECORDS SPECIFIED -----------
gS0 ALLOCATE REAL Oisk_1real(1:Nr_sBMples),Disk_1iMag(1:r\ir_saMple:=,)
970 MAT SUM_lreal= All zer05
990 MAT SUM_3real= AII_zero5
1000 MAT SUM_3iMag= All zer05
1010 FOR J=l TO Nr_records
1020 RECALL SAMPLED-DATA FOR OISK-l ,OISK-3 VELOCITIES FROM ARCHIVE
1030 --------------------------------------------------------------.
170
VAL$ ( J ) 1 II * II , II * II 1 '10 I SK_l UEL : r1EMORY ,0 ,14 : 0 I S~~_31')EL : MEMORY, 0 , 14 11)
1050 PERFORM FFT ON DISr~-l CLOSED-LOOP VELOCITY SIGNAL ------------
1060 USE RECTANGULAR WINDOW -------------------------------------
SK_1 REAL: MEMORY J 0 , 14:0I S~:_l I MAG: MEMORY, 0 , 14 11, l/ Nl/ )
1080 PERFORM FFT ON OISK-3 CLOSED-LOOP VELOCITY SIGNAL ------------
1090 USE RECTANGULAR WINDOW -------------------------------------
1100 \)ec t_ f f t ( II FFT II , if 0I Sr<_3VEL : MEMORY' ,0 ,14 I ALL_ZEROS: MEMORY, 0 , 1411, II 0 I
S~~_3REAL: MEMORY, 0 , 14: 0 I -S~~_3 I r1AG: MEMORY ~ 0 ,14" , II Nil)
1110 ACCUMULATE SUM OF VELOCITY FFT'S -----------------------------
1130 !3et_rarray( 1I0IS~~_1 IMAG:MEMORY,0, 14 11 ,Oisk_1 iMag("')
1140 Get_rarray( 1I0ISt<_3REAL:MEMORY,0, 14" ,Disk._3real( *))
1150 Get_rarray( IIDIS~~_3IMAG:MEMORY,0, 14 11 ,Dlsk_31Mag( *))
1160 MAT SUM_1real= SUM_lreal+Oisk_lreal
1170 MAT SUM_liMag= SUM_liMag+Disk_liMag
1230 COMPUTE OISK-3 VELOCITY OPEN-LOOP FREQUENCY RESPONSE FUNCTION
1240 DENOMINATOR --------------------------------------------------
1250 ALLOCATE REAL Input_real( 1:Nr_saMples),Input_iMag( 1:Nr_saMples)
1270 ALLOCATE REAL DenoM_~ag( 1:Nr_5aMple5),Oeno~_faz( 1 :Nr_50Mples)
1280 Get_rarray< "INPUT_REAL:MEMORY ,O,14" ,Input_real<*))
171
1300 MAT Input_real= (Nr_records)*Input_real
1310 MAT Input_iMag= (Nr_records/*Input_iMag
1320 MAT Denom- real= Input_real-SuM_lreal
1330 MAT DenoM_iMag= Input_iMag-Su~_liMag
1340 FOR I =1 TO Nr_saMples
5 ) )
1370 NE/T I
1380 DEALLOCATE Input_real(*),Input_i~ag(*)
1390 DEALLOCATE OenoM_real(*),DenoM_iMag(*)
1400 DEALLOCATE SUM_lreal(+),SuM_liMag(*)
1410 COMPUTE OISK-3 VELOCITY OPEN-LOOP FREQUENCY RESPONSE FUNCTION
1420 -------------------------------------------------------------
1430 ALLOCATE REAL NUMer_Mag( 1:Nr_sBMple5),NuMer_faz( 1:Nr_sBMples)
1440 ALLOCATE REAL Xferfn_Mag( 1 :Nr_scMples),Xferfn_faz( 1:Nr_sBMples
1450 FOR 1=1 TO Nr_saMples
1470 NUMer_faz(! )=FNPhase_deg«(SuM_3real( I») ,(SuM_3iMBg( I»))
i480 Xferfn_Mag( I )=NuMer_Mag(I )-DenoM_Mag(I)
1490 Xferfn_faz(I )=NuMer_faz(I)-DenoM_faz(I)
1500 IF ABS(Xferfn_faz(! »))=180 THEN Xferfn_faz(I )=Xferfn fa:(! )-360*
1510 NEXT I
ON
172
1520 STORE OISK-3 OPEN-LOOP VELOCITY FREQUENCY RESPONSE FUNCTION
1530 DISKETTE -----------------------------------------------------
1540 IF Nr_record5<10 THEN
1550 Nr"'_a···.;g$="_11 &UAL$( Nr_records)
1560 ELSE
1570 Nr_avg$=VAL$(Nr_records)
1580 END IF
1610 END OF SIMULATION PROGRAM ====================================
1620 END
1630 SUBPROGRAMS ==================================================
1640 COMPUTE MAGNITUDE IN dB OF REAL AND IMAGINARY FFT RESULTS
1650 OEF FNMaQ_db(REAL Real_part ,IMaQ_part ,INTEGER Nr_sBMples)
166 If) RETURN 20* LGT( Nr _ 5 amp 1e 5 *" SQR< Rea l_p art ,. 2+ I Ma Q_P art ,\:: ) )
1670 FNENO
1580
1590 COMPUTE PHASE IN DEGREES OF REAL AND IMAGINARY FFT RESULTS ---
1720 CASE -1 ,1
1730 IF Real_part 0 THEN
1740 RETURN 180*ATN(IMag_part/Real_part )/PI
1750 ELSE
1770 END IF
1780 CASE 0
173
1790 IF Real_part<>0 THEN
1800 RETURN 180*( l-SGN(Real_part))
181r3 ELSE
1820 RETURN 90*SGN(IMag_part)
1830 END IF
1840 END SELECT
1851~ FNEND
1860
i870 END OF SUBPROGRAMS ===========================================
10
20
30
PROGRAM ====================================================~Z 4
NaMe: SIUSI (SysteM Identiflcation Using Stochastic Input)
Author: T T HAMMOND
40 Date: ~A 1\ \ !:IM r i988
50
60
70
90
Purpose: This prograM COMputes and averages the saMpled-data
frequency response function for disk-3 velocity for stochastic
excitations. First a raM disk is established for COMputa-
tional efficiency. Next the user is queried for input 5ignal
filenaMe and nUMber of records to average. Then FFTs are
100 perforMed on the input signal and disk-l and dlSk-3 velocitIes
! 10 for each record. Also tne nUMerator and denOMInator of the
120 average response function are forMed. Finally the average
130 disk-3 open-loop frequency response function 15 calculated
140 and stored on diskette.
150
160 SUBPROGRAMS ==================================================
170 The following subprograMs frOM the HP Data Acquisition Manager
1B0 OACQ/300 package ar-e u s e d :
220 Arch_query(ArchivenaMe$,BooknaMe$,Pagelabel$,FieldnaMe$,
230 Record$,Result$)
250
250 The following subcragraM5 appear at the end ~f thIS prograM:
270
175
280
290
300
310 lJARIABLES ====================================================
320
330
340
360
, Nr_records
I All zer05
j Frequencies
input signal filenaMe
nUMber of record5 in average
nUMber of record5 to average
array of zeros
FFT frequency array
330 I T +".. npu ,,_1 Mag real and iMaginary part5 of FFT for input
390 I Disk-"'real,
400 real and iMaginary parts of FFT for velocIty
410 Disk-3real,
.4.30 j Error_rea 1 1
Error_1Mag
rea 1 and i Mag i ne r y par t 5 Gf FFT f ":J r "/ e 1eel t y -..i
real and iMaginary parts or error signal
460 Error_1Mag_sq: squares of real and IMaginary part5 of error
470 DenoM_5uM_real: cUMulative denOMinator of response funct:on
481~ ! NUMer 5UM rea 1. ~
490 NUMer_5uM_IMag: cUMulative nUMerator of response function
530 Product 4 teMporary product terMS
540
550
560
570
580
! OenoM_Mag
I NUMer_Mag
I NUMer_faz
I Xferfn fa:
denOMinator Magnitude in dB
nUMerator Magnitude in dB
nUMerator phase in degrees
frequency response function Magnitude in dB
frequency response function phase in degrees
176
590
600 OIM SaMple_rec$[10J ,Nr_avg$[2J
510 REAL Nr_records
620
530 CONSTANTS ====================================================
640
650
650
670
I Nr_saMples
l'··lr_sectors
nUMber of saMples
nUMber of sectors
saMple perIod in seconds
730
740 BEGIN ========================================================
750 INITIALIZE DATA ACQUISITION SUBPROGRAMS ----------------------
760 5Y5 i n i t
770 SET UP RAM OISK FOR EFFICIENT EXECUTION ----------------------
73(~ INITIALIZE 11:MEMORY,Z),14Il
JNr _ s e c t or s
790 QUERY USER FOR INPUT SIGNAL NAME AND NUMBER OF RECORDS TO ----
800 AVERAGE ------------------------------------------------------
177
810 BEEP
820 OISP ~ENTER NAME OF INPUT SIGNAL VECTOR USED FOR SYSTEM EXCITATION
830 LINPUT II 1I ,SaMple_rec$
840 BEEP
850 0 I SP II ENTER NUMBER OF RECORDS TO I NCLUDE IN A'.JERAGE \I ;
860 INPUT II II ;!Nr_records
870 GENERATE ZERO VECTOR AND FREQUENCY UECTOR --------------------
:2·80 ALLOCATE REAL AII_zeros ( 1 : Nr_sarflp 1es ) .Frequenc i e5 ( 1 : Nr_saMp 1es
900 All zeros < I )=fl)
9 1(2) Fr eq ue nc i e 5 ( I )=( 1-1 ) / ( Sa mp 1e_p er '*' Nr _ 5 anp 1e 5 )
930Save_rvec t ( All_zeros ( * ) , II ALL_ZEROS: MEMORY, 0 ,14 'I :>
940 Save_rvec t ( Frequenc i e s ( * ) , II FREQ__'JECT: ,700 , i \I )
950 DEALLOCATE Frequencles(*)
960 SET UP LOOP TO PROCESS NUMBER OF RECORDS SPECIFIED -----------
1370 ALLOCATE REAL Input_real(1:Nr_5aMple5),Input_iMag<1:~\Jr_sa!"lple5)
980 ALLOCATE REAL Disk_lreal< 1 :Nr_saMples),Olsk_liMag( 1 :Nr_saMple5)
990 ALLOCATE REAL Oisk_3real(1:Nr_saMples)10isk_3iMag(1:Nr_saMples}
1000 ALLOCATE REAL Error_real( 1:Nr_saMples),Error_iMag( 1 :Nr_sBMple5)
1010 ALLOCATE REAL Error_real_sq( 1:Nr_saMples)lError_iMaQ_sq( 1:Nr_sdMpl
e s )
pIes)
178
1040 ALLOCATE REAL Product_1( 1:Nr_saMples),Product_2( 1:Nr_soMples)
1050 ALLOCATE REAL Product_3( 1:Nr_saMples),Product_4( 1:Nr_sBMples)
1060 r1AT OenoM- SUM- real= All- zeros
1070 MAT NUMer- 5UJ'r) ._-real:: All - zeros
108!J MAT NUMer_5uM_iMag= All - zer05
1090 FOR J=l TO Nr- records
1100 RECALL INPUT SIGNAL FOR JTH RECORD ---------------------------
1120 PERFORM FFT ON INPUT SIGNAL ----------------------------------
1130 USE HANNING WINDOW -----------------------------------------
1140 l,}ec t._ f f t ( II FFT" , II INPUT_SGNL : i1EMOR\{ ~ 0 , i 4 I ALL_ZEROS: MEMOR'Y ,0 ,14 \I , q I
NPt.JT_REAL: MEMORY, 0 , 14: INPUT_ I 11AG : r1EMORY ,0 , 14 II , \I U II )
1150 RECALL SAMPLED-DATA FOR DIS~~-l ,OISt<-3 UELOCITIES, FROM ARCHIl..)E
1150 --------------------------------------------------------------
1180 PERFORM FFT ON OISK-l CLOSED-LOOP VELOCITY SIGNAL ------------
1 90 USE HANNING WINDOW -----------------------------------------
·S.~~_1 REAL: t1EMORY ,0 ,14: or S~~_1 Ir1AG: r1EMORY .e ,14" ,II!.J II )
1210 PERFORM FFT ON DISk-3 CLOSED-LOOP VELOCITY SIGNAL ------------
1220 USE HANNING WINDOW -----------------------------------------
1230 Vect_fft( IIFFTII,"DIS~~_3VEL:MEMORY,0,14:ALL_ZEROS:MEMORY ,0~14" ,ClDI
1240 ACCUMULATE NUMERATOR AND DENOMINATOR OF RESPONSE FUNCTION ----
1250 SUMS ---------------------------------------------------------
1260 Get_rarray( "INPUT_REAL:t1EMORY,0, 14 11 ,Input_.real( *))
1270 Get_rarr'ay( "INPUT_IMAG:MEMORY ,0,14" ,Input_iMBg( *)
12:80 Get_rarray( "0IS~<_1REAL:l'-'lEMORY,0,14" ,Oisk_1real(*)")
1290 Get_rarray( 1I0IS~~_1 IMAG:MEMORY,0, 14" ,Oisk_l iMag( *)
1301" Ge t_rarray ( "0 I Sr<_3REAL: ~1EMORY ,0 ,14 II ,0 i 5 k_3rea 1( * ) )
1310 6et_rarray ( II 0 I Sr<_3 I MAG: r1EMORY ,0 ,14 II ,0 i 5 k_3 i Mag ( * ) )
1320 MAT Error_real= Input_real-Oi5k_~real
1370 MAT OenoM_5uM_real= OenoM_5uM_real+Error_iMsg_sq
1380 MAT Product 1= Error_real. Oisk_3real
1390 MAT Product 2= Error_iMsg . Oisk_3iMag
1400 MAT Product_3= Error_real. Olsk_3iMag
1410 MAT Product 4= Error_iM8g . Oisk_3real
1430 MAT NUMer_5uM_real= NUMer 5UM real+Product 2
1460 NEXT J
179
180
1520 DEALLOCATE Product_l(*)~Product_2(*)
1530 DEALLOCATE Product_3(*),Product_4(*)
1540 COMPUTE OISK-3 VELOCITY OPEN-LOOP FREQUENCY RESPONSE FUNCTION
1550 ------------------------------------------------------------.--
1560 ALLOCATE REAL OenoM_Mag( 1 :Nr_saMples)
1570 FOR 1=1 TO Nr_saMples
1580 OenoM_Mag(I }=FNMag_db«(OenoM_5uM_realCI )),(All_zeros(I)),(Nr_saM
p l e s : )
1590 NE)<T I
15 1(0 ALL0CATERE AL NU ne r _M a g ( 1: Nr _ 5 amp 1e 5 ) ,NU jY) e I~_ fa: ( 1: Nr _ 5 aMp 1e 5 )
1620 ALLOCATE REAL Xferfn_Mag( 1:Nr_sBMples),Xferfn_fa:( 1 :Nr_saMples)
1630 FOR 1=1 TO Nr_sBMples
1680 IF ABS(Xferfn_faz(I))=180 THEN Xferfn_faz<I )=Xferfn_faz( I )-360*
SG~j(Xferfn_faz<I))
1630 NEXT I
1700 STORE DISK-3 OPEN-LOOP VELOCITY FREQUENCY RESPONSE FUNCTION ON
1710 DISKETTE -----------------------------------------------------
1720 IF Nr_records 10 THEN
1730 Nr_avg$=ll_"&UAL$(Nr_records)
1740 ELSE
181
1750 Nr_avg$=VAL$(Nr_records)
1760 END IF
1790 END
1800 SUBPROGRAMS ==================================================
1810 COMPUTE MAGNITUDE IN dB OF REAL AND IMAGINARY FFT RESULTS
1·850
1860 COMPUTE PHASE IN DEGREES OF REAL AND IMAGINARY FFT RESULTS ---
1870 DEF FNPhase_deg(REAL Real_part 1IMag_part)
1880 SELECT SGN(Real_part*IMag_part)
: 8 gO CASE -1 11
1900 IF Real_part 0 THEN
1910
1940 Ei'·jD IF
1950 CASE 0
1960 IF Real_part 0 THEN
1970 RETURN 180*( 1-SGN(Real_part)
1980 ELS·E
2000 END IF