An automated digital computer technique of control system design is presented. The emphasis is on compensator design but the method is applicable to the design of any system with free parameters. Signal representation and system response are in the time domain. The inputs required from the engineer are the system configuration, the desired output response and the free
DD
1
JAN 73 1^73
1)
(Page
S/ N
EDITION OF NOV 61 0102-014-6601
<
I
IS
OBSOLETE
SECURITY CLASSIFICATION OF THIS PAOC (Whan Data
tntarad)
ftCUWITY CLASSIFICATION OF THIS P>GEfW»..n Hrtm
£»l.f..'
parameters. A parameter minimization routine is then used to minimize a specific cost function and to set the free parameter^ A graphical output of the desired response and actual system response is then produced for comparison by the engineer.
DD
Form
1
1473
SECURITY CLASSIFICATION OF THIS F*OEfWi»n Data Enlmfd)
Jan 73 S/N 0102-014-6601
Computer Automated Design
of
Systems
by
Larry Paul Vines
Stati Lieutenant, iant. United States Navy
B.S.E.E., Purdue University,
1970
Submitted in partial fulfillment of the
requirements for the degree cf
MAST5H OF SCIENCE IN ELECTRICAL ENGINEERING
from the NAVAL POSTGRADUATE SCHOOL June 1976
An automated digital computer technique of control
systea
design
is
presented.
with
free
The
emphasis
is
on
compensator design but the method is applicable to the
design cf any system
parameters.
Signal
representation
domain
The
and
system
response
are in the time
inputs
required
A
from
the
engineer are the
and
systen configuration, the desired output response
the free parameters.
is then used tc minimize a specific cost function
parameter minimization routine
and
A
to set the free parameters.
graphical output of the
response
is
desired response and actual system
then
produced fcr comparison by the engineer.
TABLE Cf CONTENTS
I.
IN1FCDUCTION
EROGEAM DEVELOPMENT AND IMPLEMENTATION
A
E
10
II.
12
12
15
GENERAL
MAIN PROGRAM
C
FLANT
1.
15
16
Block Data Card Block Connections
17
17
Drives
Standard Transfer Function Blccks
18
Parameters to be Optimized
D
19
DESIRED RESPONSE
CCST FUNCTION
(XDATA)
20 20
E
F
(PERFORMANCE INDEX, PI)
,
MNIMIZATICN ROUTINE
1.
22
22
General
2.
Explicit/Implicit Constraints and
Start Points
,
22
GRAPHICAL OUTPUT
III.
A B
23
..,
INVESTIGATION OF PROGRAM PERFORMANCE
IATA INPUT, AN EXAMPLE
25
25
1ACHOMETEF FEEDBACK
CASCADE COMPENSATION CASCADE LEAD COMPENSATION
SERVO SYSTEM COMPENSATION
28
34
41
C
D
E
47
...
IV
A E
DISCUSSION, CONCLUSIONS AND RECOMMENDATIONS
DISCUSSION
61
61 64
CONCLUSIONS
C
V.
RECCMMINDATIONS FOR FUTURE WORK APPENDIX A BLOCK DATA CARD FORMAT
APPENDIX APPENDIX
E
65
67
VI.
PLANT FLOW CHART
CADS MODIFICATIONS FOR
69
VII.
C
PECEIEM III.
VIII.
E
80
82
CMS
EROGEAM LISTING
LIST OF FIGURES
1
Ijfical System to be Optimized
CADS Erogram Flow Chart
14
2
3
24
Hard-Leonard Speed Control System
lachcmeter Compensated System
CADS Elcck Diagram of Tachometer System
CADS Compensated Tachometer Feedback System
26
28
4 5
29
6
Sesfcnse
7
30
Type Three Block used for Pseudc Tachometer
Feedback
8
31
Eseudc Tachometer Feedback
32
9
Pseudc Tachometer Compensated System^
Response
10
33
34 35
Third Order Plant to he Compensated
lag Compensated System
11
12
13
Occcirpensated, Compensated BODE PLOT
36
....
37 33
Conventionally Compensated System Response
CADS Lag Compensated System
CADS Compensated System Response
14
15
16
40
,
Plant for Lead Compensation
41
.
17
18
Conventional Lead Compensated System Response
CADS Elock Diagram for Lead Compensation
42
43 44
45 46
..
19
CADS lead Compensated System Response
CADS Compensated System Block Diagram
CADS Relaxed Boundary System Response
20
21
22
Serve Drive Mechanism, Original Compensation
CAES lagraa of Servo Drive Mechanism
48 49
51
23
24 25
Original Servo System Response
CADS Compensated C
System Response
52
26
CAES Compensated C -1 System Response
N
54
27 28
C2ES Compensated C -2 System Response
N CfiDS
55
Optimized System Response
0.5
57
29
Serve System Response, CADS Values with
8
=
60 63
30
Elcck Reduced Ward-Leonard Drive
ACKNOWLEDGEMENTS
The
author wishes to thank Dr. George Thaler fcr his
and
guidance
encouragement
during
the
course
M.
of
this
investigation.
fcr
Kris Butler, Ed Donnellan and
Anderson of
Lastly,
I
the W. R. Church Computer Center deserve special recogcitioc
their
patient
and professional assistance.
thank my «ife, Patricia,
without
whom
this
thesis
could
never have teen accoaplished.
I.
IHTBODUCTION
The past two
decades
have
been
witness
to
an
ever
increasing
use
ci the digital computer.
Engineering usage
Classes in
have
become
the
and design are prctably the most important justif icaticn for
the large memory, high speed computers of tcday.
computer application to
part
of
engineering
curricula
on
problems
the
established
engineering
or
at
at most universities.
Numerous papers have appeared
the
[ 1 J
application
long,
of
computer
were
to
problems
best
which until recently
unsclvatle
required
tedious
procedures which gave only approximate solutions.
Contrcl system engineering has
relied
met
increasingly
and
a
on
computer
design
sinulaticn
of
large scale systems to verify that
have
teen
to
specifications
a
make
any
mcdif icaticns to
system even before producing
prototype.
There are
programs
circuit,
available
or
of
to
simulate
first
order
virtually
electrical
eguations.
control system, either in transfer
function form or as a system
differential
design
ethers
which
draw the Bode, Nichols or Nyguist plots
of open and clcsed loop systems.
Some programs help
compensators
use
an iterative method to achieve the
desired frequency response [2] [3]» Researchers continue to better adapt the computer to engineering usage.
Cantalapiedra £4] has used an iterative method to find model for large order systems. the optimum reduced order MacNamara [5 1 went even further and used an iterative method
to
find the optimum compensation for an aircraft autopilot.
and
to
It would appear that these techniques could be extended
applied
the
direct
simulation
and
design of control
10
system circuits in the time dcmain.
The ictert of this thesis was to develop
a
user oriented
of
program which could
systems
means
of and
simulate
the
a
a
wide
variety
control poles and
determine
values of the gains,
zeros necessary to produce
desired respoDse.
A
convenient
(1)
data card input was to be provided to specify
the control system which was to be
optimized
optimize
and
(2)
the
desired response. routine
(ECXE1X)
fi
locally available function minimization
the
was to be used to
simulated
system's output.
To
simulate
the
system
which
which
is
to
be
optinized,.
transfer
functions
are
commonly
so.
encountered
were
reduced to first order linear differential eguations.
These
data
egcations were then programmed
card
the
that the transfer function
blocks could be connected in an arbitrary
input.
fashion
the all the
by
Several common nonlinear transfer blocks were
The program will simulate
also provided.
known
system
unknown
with
or
parameters
parameters
to
and
be
then
allow
by
adjustable
fixed
computer
optimization routine to achieve the desired response.
11
II.
PROGRAM DEVELOPMENT AND IMPLEMENTATION
A.
GENEEA1
Any program which is to be of
maximum
benifit
tc
the
user
which
must
is-
fcave
a
simple means of data input and an output
easy to interpret and apply to the problem at hand.
The input data should have a physical significance that does net lose its relevance through the programming
of
numerous
is
eguatiens.
"present such
(CADS)
,
The program should be a readily usable tool and
not a profclei in itself.
a
The ictent of this
used
thesis
to
program, Computer Automated Design of Systems
is
of
which
readily
the output.
for
simulation
as
with
optimization
Optimum in the sense that the
near
output response cf a given system configuration is
the desired response as possible.
Most
ccntrol
system
CADS has
design
the
starts
common
with
a
a
proposed
transfer
built
which
or
function block schematic to achieve
desired time
functions
tlocks
domain response.
transfer
The
into the program and the input data can come directly
from the proposed system
are
schematic.
transfer
available for system simulation are presented in
These blocks should be adequate, either seperately
Table I.
in
various
cascade
combinations,
to
represent
most
ccntrol systems.
12
CP
SYMBOL
EQUATION SOLVED
EICCK
9i
.
G
9,.
9o= Q9 V
9i
,
G
9o
S*P
9
»
-P
9
*
G
9;
ei.
G
S+ 2
•
9
ft
q
»
-P
9o +
G(9i + H9\)
0L_
G
S 2 +Z6W5
2
*• co
-
9
o
^
9
=
n
-25^6o-^9
+6e
«uVx~ X
8i.
6
=
G 9C
6
«.
9L ^
G9
i
fie
9
"«.
9U
X
6l
9c
-9-0—
9o
9
--
9
-
*
—Zj
8;
=
°w
G6 ^ <
u
|9.
r
—S.0—
9;
9o
9
*
M
G
9|9 e
|
<
9
6
19
=
>
9.
3
)
t
TAELE I Program Transfer Functicr, Elocks
13
-
To
use
the
program
the engineer must know the system
configuration in transfer block form and the desired second second order If other than a order output response. response is desired, this may be easily specified but
requires
response.
it,
seme
the
knowledge
given
of
how
to
input
to
the the
desired
desired
CALS will take the transfer block system, cennect
compare
system
response
response and set any free parameters to achieve the
closest
match possible of the system output to the desired response.
Figure
program
1
is representative of the
type
In
of
systeir
1,
the
the are
X,
car
simulate
and
optimize.
figure
parameters of the
I,
numbered
transfer
function
blocks
known or fixed by equipment limitations.
and
Z
Transfer blocks
contain variable parameters which must be
selected
by the prcgrair to make the system output reproduce a desired
output function as closely as possible.
x
<y
i
o-
3
O
+ -6- * -Q
—
IIGQBE
1
Typical System to be Optimized
14
E.
MAIK EBCGFAM
The
KAIK program was developed to control the selection
to
of the sutf unctions used in the optimization process and
compute the desired response cf the systems which were to be
optimized.
A
second order step response is calculated
from
eguation
XEAIA.
(I)
as the desired response and stored in the array
G
e
=
u(t> S2 + 2
8
(i)
WtjS + W
2
The MAIN is essentially a bookkeeping routine which controls
the
prccraa
execution.
The
sufcfunction
operations
it
controls are defined in the following sections.
C.
ELANT
The ccrnncn simulation routines LISA, DSL, ECAP, CS£P and
INTEG were investigated in an attempt to adapt them
tc
the
general
crotlem
specified
equations
all
above.
These programs reguired
either the incut cf the system's state
variable
eguaticns,
are
useful
Laplace
for
transform
if
or the node-transfer function
pair for each system simulated.
These programs
simulation
of a system's parameters have been
specified.
these
Eowever, each simulation is problem specific and
are
A
technigues
cf
cf
not
readily
card
integrated
with other
sutfuncticn ticgrams.
capable
variety
simulation
data
subprogram
of
which
a
was
wide
accepting
input tc simulate
ccntrcl
systems
and
working
with
other
existing subprograms had to be developed.
15
Subfurcticn PLANT was developed to simulate the systems It provides the versatility tc he optimized. which were
necessary to simulate numerous system configurations and is capable cf working with a minimization routine to optimize
the variable parameters.
T^e
program
reads
the
transfer
in
fucction
state
the
blcck
connections
format.
from data cards.
This data is
then used to automatically set up the
system
eguatiocs
variable
These state variable eguatiocs are
solved as first order, ordinary
differential
eguatiocs
are
in
by
Bunge-Kutta-Gill
The
forth
order method.
of
The programming
was done in Jortran IV and all calculations
double
precision.
capability
in
connecting
the
transfer
forward,
fucction
f eedback
blccks
,etc
.)
any
configuration
having
each
(feed
is
possible
with this program.
Simulation
withic
the
flexibility is provided by
blcck
system capable of accepting an external forcing function.
1 -
JiccJS Data
Card
Several
possible
means
ways
of
inputting
a
the
data
necessary tc simulate arbitrary system
investigated.
directly fica
a
A
configurations
chosen
as
were
the
of defining
system configuration
blcck diagram schematic was
most preferable method.
of the system.
Using this approach, each data card
was designed to be directly related to a
transfer
fucction
to
This resulted in having direct access
the
transfer fucction parameters for optimization.
1c
simulate the system,
a
a
data card is prepared for
The data card
each transfer blcck in the schematic diagram.
ccctains
the
field cf cumbers which specify the block number,
the type cf transfer functioo ccctained
within
the
block,
input
cede
and
the output node to which the blcck is
connected and the values of any parameters associated with general fcrmat of the data the transfer function. The
16
.
cards used tc input the system configuration is shown telcw.
|1
]11
|
20
G
|
40
P
|60
Z
(Column Numker)
ELKCCE=EEVV
Hhere:
CC
D
=
-
Position number of the block
Type of tlocic (number)
EE = Input node number
VV = Cutput ncde number
G,
P,
and Z are parameters of the transfer function.
2.
Ilcck Connections
The rumfcer of transfer blocks in the system and
the
data
cards associated with each of the blocks are read upon
The prograi then initial entry into the simulation routine. connects the transfer function blocks in the proper order by
comparing the input node number of a
ncde
nunfcer
block
to
the
cutput
cf
every
other
block.
Whenever
these two
numbers are equal, the blocks are knowc to be connected and The flags are then used to is set equal to 1.0. a flag
identify the input drives to each block.
3
Er i v es
The input guantity
to
each
block
to
is
called
the
EBIVE.
The
program
was
designed
allow for multiple
was
inputs to the system being simulated.
This flexibility
to
the
achieved
plus
anj
by making the input to each block equal to the sum
of the outputs of all blocks connected
input
node
external
forcing
function
(DBVIN)
feeding the
transfer tlcck.
shewn in equation
The input DRIVE to a block is determined as
(II)
.
17
DBIVE(i) =
IHA(j)
j
£
THA
(j)
*FLAG
(j
f
i)
DRVIN
(i)
(II)
is the output of block j.
FLAG(j,i)
is 1.0 if block
is connected to block i and zero if it is
(i)
cot
connected.
EEVIN
is
any external forcing function specified by the
i.
user whicfc drives block
DRVINs
must
be
inserted
in
subroutine
the
IIANT
as
=
Fortran
IV statements.
The standard
program has CSVIN(1)
system.
1.0 specified as a unit step incut to
Problem III-E in the section Investigation of
varying
Prcgram Eerfcrmance demonstrates how multiple, time
inputs are tc be inserted in the program.
4.
Standard Transf er Function Blocks
The
transfer
function
I.
blocks available for system
simulation are shown in Table
were alsc found to
These blccks were selected
They
cr
because cf tteir ccmmcn usage in the modeling process.
be
adequate
to
either
separately
most
in
various
systems.
cascade
combinations
represent
control
The
transfer
function
equations
for each tjpe of
block were written in state variable format and stored in an
array
named THA.
with
The program reads a number from a block's the
type
of
data card which specifies
transfer
type
of
function
equation
associated
solved
the
block.
by
The system equations ar€ then
the
seguentially
selecting
assccitaed with block number one, solving for its output and integral etc. The then sequencing tc block number two,
eguations
were
are
solved by a modified RKL fourth order method
routines.
a
in the subfunction
Three
integration
routines
The
necessary
tc
store the intermediate results obtained
for those equations involving
double
integration.
four
sutfunctions,
RKLDE2,
RKLDE3,
CCPIX
and RKIDI4 are
of
a
called by IIAKI.
BKLDE2 is used for the
integration
18
type
twc
blcck.
and
BKLDE3 is used for integration of
to
a
type
for
three
blcck
store
intermediate
calls
quantities
and
sutfuncticn
CCPLX.
CCPLX
RKLDE3
RKLDE4
for
integraticn cf a type four block.
Nc
prevision has been made for the input of initial
cenditiers cr the integrators.
must
run
as
Therefore, the
(DT)
integrations
the
start
time
at
fc€
.
time equal zero.
The user must specify the
and
step size tc
used for integration
possible.
DT
=
problem
as
(IF)
To conserve computer time ET should be made
large
as
TF/1000
is
suggested
shewn
appropriate
letting
was
TF
in most cases.
Equally important is that TF be
as small as pcssible..
te
Use of the
program
has
that
greater than the transient response time of
For preliminary analysis TF
the systea is rarely justified.
kept tc enly slightly longer than the time of the first
undershoot fcr second order dominant
teen included as Appendix B.
systems.
Because
of
the complexity of subroutine PLANT a flew chart of PLAKT has
5
.
Pa^am eter s to be Optimized
The
parameters
ty
of
the
system
which
are
to
be
be
optimized
the
minimization
routine
EOXPLX
must
specified in PLAN1.
labeled
variable
•C
l
The minimization
routine
returns
equate
the
trial values cf the variable parameters to PLANT in an array
.
fiegular to
Fortran
the
be
to
IV
statements
in
the
the
If a
parameters
pair
=
values
in
this array.
P(i)
pcle-zerc
statements
and
Z (i)
were
be
optimized/
PLANT:
following
C(1)
wculd
C (2)
.
inserted
The location fcr the preceding
statements
is clearly
iEdicated in the program listing for PLANT.
19
D.
DESIBEI* BESPONSE
(XDATA)
The ciitciia against which the system response
will
be
compared
system.
en
will
vary
of
according
second
a
to
the
application cf the
is
Since most design work on control systems tasis
tc
done
the
order dominance, the program was
order
written
provide
second
step
response
with
adjustable
5,8
n
and
gain as the basic criteria against
The desired 6,
which the sinulated system will be compared.
H
and gain are read in as input data.
n
The step response is
and
then computed in the main program by subfunction RKLEEC
stored
order
by
in
the
array
called XDATA.
Any ether time domain
response nay be specified by the user by removing the second
step
response
III-E
equations and replacing them with the
An example
is
equations of the desired response.
problem
is
provided
in Investigation of Program Performance.
If the program is being used for simulation only and nc data
curve
desired, setting the input variable LEAP =
1
will
cause the program to bypass the computation cf rhe
secend order data equation.
standard
E.
C0S1 FOSCTION
(PERFORMANCE INEEX, PI)
The
achieved
response
this
tc
system
and
for
the
response
is
compared
is
witt
the
The
desired
ainimize
the
difference
parameter
system
the error.
program searches
user
the
settings
as
which
will
error. The cost function may be specified by
weight
the
outputs
desired
in
subfuncticn FE.
The default cost function cf the program is
20
rr
tte integral errcr sguared.
J =
/
(Err) 2 dt.
An
example
of a weighted cost function is presented in Section III.
21
.
F.
MINIMIZATION RCOTINE
1
G ene r al
lh€ free parameters are optimized to reduce the cost
function hy the complex
method
of
M.
J.
Box. [6]
Box's
constrained
called
optimization
method has been programmed at the
Naval Postgraduate School by B. B. Hilleary as a
subroutine
ECXFLX.
This subroutine will find the minimum cf an
(cost
arbitrary function
function)
subject
lower
to
arbitrary
Explicit
on
explicit constraints and for implicit constraints.
ccnstraints are defined as upper and
free
bounds
may
178).
the
parameters.
Implicit
constraints
P
1
be arbitrary
functions of the free parameters (e.g.
P
2
<
Two
function
and
The
subprograms
implicit
EOXPLX
method
are used to evaluate the
objective function
constraints,
uses
which
FE
and
KE
respectively.
values of
the
to search for the
free
parameters
minimize
the
cost
function is explained in the computer program listing.
2.
Ex elicit/ Implicit Ccnstraints and Start Points
ECXFLX
space, where
and
n
searches a feasibility region
=
(n-dimensicnal
number of free parameters)
defined ty upper
a
lower
bcunds on the free parameters for
The
the
minimun cost
by
function.
sialler
nore
the
region
defined
these
rapidly the program will converge to judgement Good engineering the optinum parameter settings. will be necessary to keep the feasibility region as small as
boundaries,
22
possible.
data
The boundaries of the search regicn are read from
by the main program as the upper bound
(XL)
(XG)
cards
and
the lower bcurd
for each free parameter.
may be any arbitrary furcticn
If an
Implicit
of the free
constraints
parameter desired.
implicit
constraint
bj
such
as
the
product of a pole-zero pair must be less than
the
seme number is to be evaluated, it must be supplied
user
to
the
subfunction KE.
No implicit constraints were
used in this thesis.
The
starting values cf the free parameters
data
cards.
A
(XS)
are
of
read in by the MAIN frcm
good
or
choice
Bode
starting methed
of
values
will dramatically reduce the time required
A
for optiii2aticn.
preliminary root
locus
plot
estimating the best values of the free variables
should be accomplished whenever possible.
G.
GRAPHICAL OOIPUT
Two
subroutines
were
written
to
provide
for
the
graphical output of the desired response and the best system
response achieved by the optimization process.
select, by data card input,
The user may
PPLI
which
either
Every
subroutine
fifth
and
provides
a
a
high speed printer plot or subroutine PIC which
provides
calcomp graph.
the
integration
THAODT
is
point
stored
in
arrays
XDATA
plotted.
Subroutines PPIT and PIC call the subroutines PLOTP and CRAW respectively. PLOTP and DRAM are standard plotting routines
at the NP£ computer facility and
are
not
a
part
of
the
simulation
program.
Figure
2
diagrams the information flow
and data input to the program.
23
1
*
*
No
.
Runs
r
XDATA
6
>
No. Var dt, TF
• •
i
wn
Type Graph
No. Trials XU, XL XS
•
•
,
«
t
•»«
1
1 1 1
MAIN
i
—.
.»
—
m
w»
—
—
Plot output response
L
1
1
t
I
1
1
1
1
i
t
1 1
BOXPLX
n
i
~"
*
]
s.E
1
1
1
1
BLOCK Data Input Drives
[
1
NO
'
J
I
t
FE
1
YES
/K>| < ^
l
>v
I
/^Implicit
/ Met
xTli Constraints
i
i
r
*4»
t
1
i
PLANT System
Si .mulation
i
—
«...
RKLDEs
Optimization
Simulation
FIGG3E
2
CADS Program Flow Chart
24
III.
INVESTIGATION OF PROGRAM PERFORMANCE
The example problems presented below were used to aid in
tie development cf the optimization program.
The
order
of
difficulty
cf
the
problems
progresses frcm a simple text
a
bock single variable, single input system to
multivariafcle
operatioDal servo drive system which has multiple inputs and An example of how a schematic discrete level feedback.
representation of a system is prepared for input to example the progran is presented prior to considering the
diagram
problems.
A-
EATJ INECT, AN EXAMPLE
Ihe Rard-Ieonard drive system [7] shown schematically in The gain cf the (a) has two variable parameters. Figure 3
amplifier and
adjust
tie
the
tachometer
response.
feedback
To
are
available
to
a
system's
simulate the system
I.
blcck diacraa representation of the system is drawn as shown
in Figure 3
(c)
using the transfer blocks frcm Table
Cortjt
fTAc*
<a)
25
URvjnCj)
GO)
S*P<0
t\-
-5G>
GCO
GO)
S+PW
-($
nn
G(4)
EC
Gfs)
(c)
fIGCBE
3
Ward - Leonard Speed Ccntrcl Systei
(A)
tsing Feedback.
(B)
Schematic Diagram
(C)
Elock Diagram
Blcck Diagram
Using CADS Blocks.
= K/t
=
Hfcere
G(1)
G(2)
P(1)
=
1/T
Km/Ba
1/J
P(3)
G(3)
G <4)
=
=
=
=0
-Km
-Kt
= Er =
G (5)
DEVIN DEVIN
The
(1)
(3)
-T t
are
then
as
nodes
cf
the
block
diagram
The
nunbered
shcwc
in
sequentially
also
1,2, ...,n.
blocks between the nodes are
1,2, ...N
(N)
nunbered
3
sequentially
Data
Figure
input
within
(c)
.
cards
are then prepared for each
blcck which specify the block number,
ncde,
tfce
type
cf
blcck,
the
the
output node, and the parameters contained
In Figure
3
blcck.
(c)
,
for example, block
1
is
26
a
type
two transfer block connected between nodes
1
and 2.
The data care input for this block would be
BIKC 12=0 102
K/T
1/ T
The program reads the data card input and
connects
number
the
blocks
any
by
a
setting
block.
a
FLAG
=
1.
whenever the input node
of
nunber to
other
block is the same as the output ncde
The
input
of
to a transfer block is then
determined tc be the
connected
step
tc
sum
the
outputs
any
The
of
all
has
a
blocks
unit
the
input
node
plus
external forcing
function driving the input node.
specified for DEVIN
(1).
program
If this is the only input to
the system, no action is necessary on the part of the
If if there
user.
ether than a unit step input to the system is desired or
are
,
other
external
forcing
functions
such
For
as
EEVIN
(3)
they must be specified and placed within the body
the
in
cf the subfurction PLANT as Fortran IV statements.
example
shc%n
Figure
DEVIN
(3)
3
(c)
,
a
)
card with the equation
=
f (T
wculd have tc be inserted preceding the drive equations.
An
example cf hew multiple, time varying drives
is
are
specified
given in section III. E.
When cptinizing a system, seme of the
input
quantities
will
be unkcewn or variables.
These variables must also be
To optimize the variables
assigned *itfcin subroutine PLANT.
K
1
and
K
t
cf Figure
3
the following two statements would be
inserted in IIANT:
G(1)
= C(1)
G(5)
= C(2)
where
C(1)
and
C (2)
are
the
variables
which
will
be
optimized by subroutine BOXPLX.
27
£.
IflCBCMJlIE JEIDBACK
The
had an
first
exact
Ac
optimization problem attempted was one which
solution
that
can
be
found
by
algehraic
methods.
instrument
servo [8] with unity feedback and
1000
forward transfer function
G(S)
=
S(S+10)
was to be ccopensated with tachometer feedback as
(III)
shewn
in
Figure
the
5
4.
The
enly
specification
that
fcr
the
system's
a
performance cf this single variable, secend order system was
simple
=
requirement
the closed leep roots have
0.7.
D
f
i
O
\ f-
\OOa
SCs+io)
K4 S
.
IIGUEE
4
Tachometer Compensated System
The
systeii shewn in Figure 4
tc
was redrawn as Figure
5
in
order
achieve
the
tachometer
feedback.
5
The
is
characteristic
S2
equation
(10 +
= t
of the system shown in figure
103 k
t
)
S
+
103 =
(IV)
The
IV.
required
K
0.0343
may be calculated from equation
28
-) >Q
loeo
S +10
/
«C
m
s
___
«t
1
FIGDBE
5
CADS Block Diagram of Tachometer System
CADS was programned to
optimize
to
the
system
with
the
variable,
0.C1
the
Wn =
K
t
F
,
specified
be
between
the
limits
tc
5
<
K t
<
1.0.
The desired response was specified
be
0.7,
be
standard
100C.
-
seccnd
order
step
response
for
K
=
CADS determined the optimum value for
to
0.03*43.
Figure
6
shows the system's step response and
the desired response are virtually superimposed.
29
CM
00
o
Ho D O < S
Eh
o
O
0.0-
0.05
0.10
0'.15
0.20
0.25
Time
(sec)
E1GUBI
6
CADS Compensated Tachometer Feedback
System Response
30
The
tyce one system, in the preceding example, allowed
derivative feedback from the forward path without
a
requiring
One may
be
block
which
a
was
capable
of
differentiation.
system
a
encounter
configured
path.
type zero system or a
to
of
which
cannot
provide derivative feedback from the forward
Tbe possibility
a
providing
pseudo
derivative
feedback using
cases.
type
7.
type three block was investigated for these
A
three
block
with
Z
=
and
G =
P
is shewn in
Figure
If the pole is placed far out or the real
axis,
this block actroximates derivative feedback.
FIGUFE
7
Type Three Block used for
Pseudo Tachometer Feedback
The effect of using this pseudo tachometer feedback
the
on
complex
a
roots of the closed loop system is negligible.
real root at
P =
It does add
-964.
The system of Figure
5
was redrawn as shown in Figure
8
using the pseudo tachometer feedback.
31
+
1
S
—rv
-(oV+S
iO
J
,
r "\
S+10
\J
SHO 3
-i
FIGURE
The
8
Pseudo Tachometer Feedback
calculated
a
5
optimization
program
K
t =
=
0.0352 fcr the
abcve confignration.
This gave
0.715.
The
in
system response and the desired response are shown
9.
.
figure
The
two
responses
are
again
step
nearly
tc
sup erimpcsed
This method of providing derivative feedback
the
has the disadvantage of adding an integration
problem
solution
with
the concomitant increase in problem
sclution time.
32
0.0-
0.05
0.10 Time
0.15
(sec)
0.20
0.25
fIGURE
9
Eseudo Tachometer Compensated System's Response
33
c.
CASCADE CCMEEJJSATION
Having
a
jrcven the feasibility of the program optiaizing
single
variable
third
system
order
where
an
exact
two
solution
was
available, the next problem considered extending the problem
scope to
a
plant
to
with
be
variables.
The
unstable
Figure 10.
plant
that
was
compensated
is shewn in
o
II
—
(i«»i)(j«+062jft»+i)
EIGUEE
10
Third Order Plant to be Compensated
The
have
riant
<2,
was
to be stabilized using a single section
cascade cempensatcr.
H
The compensated plant was required
tc
without reducing the error coefficient. [6] The
PW
Tc keep the error compensated system is shown in Pigure 11. coefficient constant, the compensator used was a simple lag
network
(
t,< t%
)
.
34
r,sn
S (S + i)
(.2Sf)
FIGOBE
11
Lag Compensated System
A
compensator
or
the
which
wculd
meet
the
required
as
a
specif icaticrs was calculated by conventional methods
check
program's performance.
The values of
=
The Bode plot cf the
uncompensated and conventionally compensated system is shewn
in Figure
12.
tfce
t
r,
and
r.
for Figure
11
which
tc
would meet
t,
design requirements
2
were
determined
shewn
in
be
-
1C,
100.
The compensated system's response
figure
using these values for the compensator is
15.
35
'a
«d <y CO 4-J <D
CO
ac e o a
a>
DU
o g c o
2 H U
jQ
o
O
CO
o
CN
36
0.0
2.0
13
4.0
6.0
(sec)
8.0
10.
Time
FIGCBZ
Cor ventionally Compensated System Response
37
.
The
conpensated
14
system
was
redrawn
program
as
shewn
in
Figure
using
the
standard
transfer
blocks
available for system simulation.
Q^Z^loJjL -O s+i (SfPc)
-I
O
S*5
-o
FIGURE 14 CADS Lag Compensated System
A
H
stardard
was
second
as
order
the
match.
<
P
response
of
&
=
.3,
=
n
0.75
chesen
program
desired
response
for
the
optimization
to
The limits placed en the
< C
.1
optimization program were .001
Arhitrarj
E
and .01
<
Z
<
1.
C
values
Z
to
.1
begin
in the
optimization were specified as
logarithmic
T,
=
.01 anc
centers
and
15
of
=
the
Co
Co
search
2cnes.
CAES determined
=
8.078
f2
67.47
the
as the optin.cn parameter
settings.
the
Figure
shows
desired
response
response
and
program
compensated
system
response cf the system compensated by CADS is more nearly the desired response than is the conventionally compensated sjstem. One should remember that the specified
The
response is for
a
true
is
second
forth
order
order.
systen
whereas,
a
the
ccnpensat€d
system
Therefore,
perfect
38
match cf
desired
versus
actual
responses
could
net
be
cttained.
39
T
•
rH
/f^\
CN
•
/ /
//
\
\^
r-\
II
o
O S B
\
•
HH D
•
/
/
//
\v ^
\
"'"^
\
,
R
//
\
//
//
\
^\
//
00
^
D
o
•
11
/
o
•
/
<*
•
o
/
/
/
CN
o
•
y?
/
0.0
y
2.0
4.0
6.0
8.0
10.0
Time
(sec)
FIGDBE 15
CADS Compensated Systeir Response
40
r.
CASCADE 1EAE COMPENSATION
The complexity of the next problem to be solved by
was
CADS
extended
was
to to
five free parameters.
be
The plant shewn in
a
figure 16
used
to
follow
unit
amplitude
anc
the Two
sice-wave input of 200 rad/sec.
output
cculd
not
The output amplitude was to
by
be almost exactly the same as the input amplitude,
lag
the
input
mere than 10°.
sections cf cascade compensation were to be used. [8]
SiH
ZOOi
*
K
S
x
G*
j
EIGURE 16
Plant for Lead Compensation
The specified requirements may be interpreted as an open
leep gain
a
> 15
db and a phase angle
A
^
90° at
W
=
n
200
from
Nichols
ttat
jlct.
a
cut
of
3
and
X
try solution on a Ecde plot
10 6
Z
shewed
gain
and
=
two
phase
lead
compensators
requirements.
of
with a double zero at
70 and a double pole
at P = 7CC will satisfy the closed loop magnitude and
phase
figure
17
shows
the
system
response and
The
desired response obtained for these values.
the
magnitude
compensated system's response is 33% of the desired
response and lags by 8.12°.
U1
V)
a o W
u>
C3
a
M
W
>i
UJ
i>
P
0)
x>
Ui
a
u
«
C
o
•H
C
> a o
u
-'
03 =3
1*1
J-nOVHi
42
The
prchlem
was
then run on the optimization program
The
free
*ith the tlcck connections as shown in Figure 18.
parameters
the
were the poles and zeros of the compensators and
of
gain
=
the
plant.
was
The
desired
data
curve
of
generated from the standard second order step response by setting 5 = Q, w = 200 and using
XDAIA
sir(.200t)
n
X(2)
as
tre
desired
response.
<
P
The
<800,
i
initial search zone
60
<
Z
<
linits were specified as 600
2.6 X 10« <
G (3)
80,
i
<
3.2 X 10*.
SM
2aa±*0_
A
ti±zJ
S+T>,
<y
-J
o^ 0-J&-
FIGUBI
18
CAES Block Diagram for Lead Compensation
The optinizaticn program solution
went
to
the
lower
were
lioits
fcr
tcth
500
<
poles
and
the upper limits for roth the
the
80
zeros and the gain.
The limits of
<
search
<
Z
zones
< i
relaxed
3.2 X
tc
P
i
<600,
9C
and
the and
106 < g(3)
3.4 X 10*.
the
The program again placed
of P
i
free
G (3)
variables
= 3.4
X
ce
.
limits
=
500,
Z
= i
9C,
1C 6
The system and desired response for these
in
values
is
shewn
Figure
19.
The system magnitude and
phase are much closer to the desired
cutput
response
than
the respcese cf the conventionally designed compensator.
43
o
•ji
a o
—
a>
Q3
<v
-M
W
>i
en
M
<y (0
w s
(i/
a.
u
n3
a o
a u
o\
en
w m U H
Cm
44
The phase difference is only 4.33°. the
fcouncaries
produced
the
Continued relaxation of compensated system shewn in
Figure 2C.
these
lags
the
figure 21 is a plot of the system's response for
The system's response is improved in that it
ty
values.
irput
only
3.46°
and
the
magnitude
is
essentially the sane as the input magnitude.
Si N
LQQl
Cs+a& Si-SOO
-*
O
ifo+'^Q.
S+*too
3-*fo/o»
O
-I
FIGUEE 20
CAES Compensated System Elcck Diagram
The
OEtinizaticn
the
program
was
activated
were
at
T
=
although
prccess
problem
specifications
for the steady
state resicnse.
The problem was rerun with the optimization The same
effctained.
started after the transient had died out.
fcr the
the
values
optimum
compensator
Mere
Apparently
snail
initial transient did not effect the
prctlea scluticn.
45
<v VI
c
o
Oi
VI <p
P3
a
+j
w
pi
en
>i
M
(d
n
x
(0 .-»
s 3 o
03
a «s u
to
W
05
H
&4
T15
iftOYHi
{TT-
46
E.
SEBVC S1S1EM CO KPENSATICN
The
foregoing examples of the simulaticn - optimization
were
program
guickly
simple
with
examples
the
which
of
could
obviously
usace
be
A
sclved
standard cut and try methods.
program
mere complex and challenging example
is
presented
fcy
the sjstem shown in Figure 22.
The system is an operational servo drive mechanism
with
multiple
as
inputs
and
and
discrete level feedback.
(t> 75ms)
This bighly
nonlinear system's output was to follow the input as closely
possitle
the
in steady state
there was to be
and the
very little ccise ripple.
are
The free parameters of the system
of the compensator, C
Li
pcles
and
zeros
,
pcles cf the ncise suppressor C
n
.
To siaulcte and optimize the system it was redrawn using the available program blocks as shown in Figure 23.
Euring
I
simulaticr
it was found that the current limiter for
|I
D
|
was
D
net needed because to
< 25 A and the
limiter
was
removed
(XDATA)
decrease program run time.
three
then
in
The desired response
was written as a set of
equations.
These
eguations
and
were
used
to
replace the second order step response
eguations
were
the standard program.
DRVIN(1),
(4)
(5)
alsc
written
as
a
set
of
equations and placed in
subroutine E1ANT. The discrete level feedback to block two These was achieved by making DRVIN(2) = INTGEB (TBA (12) changes tc tte mair program and PLANT were all that were neccessarj tc sinulate this system. The inplementation of
)
.
these changes to the program is shewn on pages 80 and 81.
U7
%
<
6 o
II
o
^
tx
a o
•H
-P
(0
UJ
a
Pi
W°
a;
tti
t=
u
o
c
>0
o
it
•H en •H
u
O
>
ui
M
5
a
2 o
•H
C
«3
X!
<J <y
^^
§
*
/-».
3
+
•4
3?
+
*4
>
•H
c
C
V)
>?
-t-
3
3
<^>
v^
-t
tf
"7
n
6
«M
Q
M
O > M
<D
CO
a
o
c
h £
u
(N
W
a
•
» •
c
a
n;
*
>
i.
si
,c
-H
ii ii
m|«o
ii
•
<i
ii
<*
•
5
•5
a?"
>"
48
•H
a w S
(TJ
JO,
o
•H
Q
O >
M
u
VI
M-l
O
(0
M
en
a
Vi
a u
««»
w « M
49
The
parts.
to
system
optimization was initially broken intc two
This sas tc reduce the number of variables that were
The above steps were
be optimized per run and to obtain near optimum starting
values fci the free parameters.
in
taken
an
effort
to decrease the computer time required for a
solution.
C
i
The first run was tc
of
optimize
the
compensator,
independent
run
C
8
.
the
to
noise
suppressor.
values
for
The
the
second
noise
cptimizaticn
suppressor,
was
select
The
rationale behind this separation was
that the two circuits perform different functions and should
therefore
system.
free
be
initially
separable
in their effects en the
The final optimization run was to be made with
all
parameters
available to the program fcr optimization.
The search zene centered on the values found above.
Figure values
Z
2
24
shows
the
in
response
Z
*
for
the
run
to
P
l
system
set
=
as
originally compensated.
fcr
£
2
The optimization
=
the
C
resulted
L
=
43.5,
21.0,
of
=
47.5,
5S2.C.
Figure 25 shows the
values.
The
simulation
the
system
using
these
initial
velocity
after
overshoot has teen reduced and the
the
average
velocity
transient
appears
more
equally distributed above and
below the desired velocity.
50
W
c o
Q4
V)
0)
u-
a
<v
*j
>i
</)
o > M
en
H
(0
a
•H o> •H
O
CM
M
-
H
O'Ofr
o*oe
cot o*03 (39S/UT) A^TOOXSA
51
W G O
Ui
o H
<y
=3
w
W
W
>i
o
00
o
Q) CO
u>
<0
6
w a
a>
O
E-t
u Q U
en
a o
o
m
CM
W «
H
O
O'Ofr
o'oe
o*o3 cot (09S/UT) Aq.TOO-[SA
52
The
iiiitial optimization
run for setting the values of
the noise suppressor used the same cost function
which
had
been
J
W
used
/
for
(upper
all
previous
<^
optimization
0.2
runs
and
=
Err 2 dt.
15C0
This resulted in
bound)
.
=
(lower bound),
=
n
Figure 26 shews the result of
was
to
using
this
cost
function
it
further
decrease
the
overshoot.
mere
Ecwever,
the
allowed larger switching cr noise
transient
Tc
J
transients as a result of weighting large
than
errors
this
lesser
noise
jitter.
overcome
=
/
prcrlem, the ccst function was changed to
the transient errors.
|Err|*t*dt.
This was tc weight the steady state errors mere heavilj than
Using this cost function
<f
the
H
values
=
set
by
the optimization program were
,
=
.
2
and
1430.
n
The damping factor, S The
was again placed on the lower limit.
initial
overshoot was still improved ever the original
the
system's response tut
iaproved
27.
switching
transients
were
not
ever the previous optimization trial.
The results
rigure
of the siiulction run using these values is shown in
53
%
V
H o
<p UJ
s o
cu
V} <y
03
o
iH
a
Ul
o
•
u w
Pi
CO
f~
1
00
o o
^
<u
u'
•o
•
w
CD
w
+J
id en
g
•H
vo Eh
c
CD
o o
CU
•
u
a o
a
*s
CO
u
o o
•
vO CN
=q as
<J>
o M
«N
PM
0*0fr
0*02 (TOT Oe (D3S/UT) Aq.TOOI3A
o o
54
'
%
V
CM
o
•
<y
W a o
cu
CO
o H o
«
a
(V
cu
•
-M
CO
PH
CO
CN
1
o—
O
•
CO
o
u'
'O
CD
—
CO
P
s
CD
(V
CD
<d CO
e
•H KO Eh
O o
a,
•
u
a < u
CO
O
B
T
o o
•
r* CN
W 03 a
C3 tn
O
CN
O'Ofr
o*oe
o*oz
o*oi
o o
«
55
A
final
optimization
The
run
was
made
with
all
six
performance index was changed tc the weighted cost function shown in eguation V. The weighting factor was computed to equally weight the start up transient
variables
free.
and the steacy state responses in an effort
to
reduce
the
switching transients.
J
=/*|Err|dt
-
0.0
<
T
<
0.045
TF
(V)
J
In. 14*|Err|dt
0.045
= 521.,
< t
<
The optimuff compensator
P
2
fi
values
P
3
established
Z
3
by
6
CADS
=
were
=
27.1,
1266.
Z
2
= 52.5,
=
48.1,
0.32 and
=
o
figure 28 shows the fully
Ihe
compensated
system's
response.
initial
overshoot has been reduced acd the
tc
average velocity in steady state is clcser
value
than
the
the
the
desired
II.
criginal
system's response.
still
the
However, the
Table
transient summarizes
system.
The parts.
encr
en shut off is
present.
results
of
optimization of the servo
CPD time reguired for the optimization pr-ccess was
net decreased by splitting the
Ihe
problem
into
two
separate
time reguired for each optimization run en the
individual compensators was the same as the time reguired tc
optimize tie complete system with six variables.
56
'
<N
iH
O
•
0> to
e o
O
rH
•
w
CW
Oi
O
ua
a
u>
-P
W
>i
CO
00
o *-» o O
•
"
T3
—
<D CO
V N
•«1
a
•H
4->
CD
g
•H
vo En
O
U4
O o
•
u <J u
CO CN
CO
•^r
o o
•
w 03 o
CS
CM
O'Ofr
cot o*03 o*oe (oas/ui) At}TOO-[aA
57
o o m
o o en
o o
IT)
O m
^r
oo KO CM
CM in
«o
in
CM
CM
f* iT
eg
r* j»
riT
r^
^
o
in
r» cr
r»
^3"
00
3-
a o
•H
(d
N
•H
S
n
Ck
o o
in
•
t
•rt
CM en Ul
CM
cr>
CM
m
O
in
^> CM in
V
'-it
>
•H in
01
N
o
in
•
^ m
•
in cr
in 3-
m
i
a
M
O > M
Q)
m
zr
•
n
•
S3"
5T
IN in
CO
*—
m
o
N
04
o CM
r™ CN
«~ CM
r-
CM
p* CM
P W
r•
M
U
M (0 a s 9
•^
£3
4J
O
•H
4J
4->
4J
P
*
<3
H H W
f-J
U a
4J
T3
(M
T3
(M
—
W
M M
3"
CQ «C
u
u
^^^
W
M M
W
M M
+
-P
H
—
T3
m
M M
H
CO
1-1
M O e
Ui
(V
(0
c
•H
cr
••-»
Q4
u
o
a
o
M
58
This problem has presented an excellent example of
the
necessity
index
carefully select the cost function which will neasure the sjsten's performance. Defining a performance
tc
which
will
weight
the
more
objectionable
that
thej
2
characteristics of a system's response so reduced becomes difficult eliainatec or
ccuplexity ircreases.
are
as the system's
dt,
The performance index, J = /(Err)
was not adegtate in its treatment of the noise and switching
tracsients fcr the above
indexes
used
problem.
The
in
ether
performance
effects upon
functicc
was
were
also
marginal
their
parameter op tinization.
Although
the
cost
be
reduced
to
a
mathematically correct minimum, the system's
achieved. usinc
The
performance was not the best that could
system's
performance was only optimum due tc the definition
The system was
of the cost furcticn.
simulated
the
compensator values determined from the last optimization run 0.5. with the exception of 5 which was increased to «f =
This
siaulaticn was made based upon engineering judgment of Figure 29 shows that this change the effects of varying S . the damping factor produced a system response which was in
nearer the desired response than any of the previous runs.
59
o
•H
o O
W
a;
(0
a
in
o— O u
•
00
<D
a;
0)
o
W
0)
e
•H
<£>
O
Eh
w
>i
o
o
> u
a>
en
O'Ofr
coe
cox o*oz (D9S/UT) Aq.TOOX9A
60
IV.
DISCUSSION^ CONCIUJICNS AND RJCCMMEND A 110 *S
A.
LISCDSS1CN
The
objective
of
this
thesis was to investigate the
feasibility cf developing a computer program which would optimize a variety of control systems' respcnses in the time
domain.
Ihe
program
time
developed
during
but
the
cf
investigation
control system
due
tc
proves
that
dcmain
optimization
responses is net only feasible
readily
desirable
the
interpretable
results of the optimization process.
CADS requires approximately 200K bytes cf computer core when
the high speed printer plot is used for graphical display of
the output arc 230K when outputting calcomp
graphs.
These
cere reguirenents are a maximum and could be reduced by more
careful, professional programming.
The
cenputer
time
reguired
for
CAES to arrive at
a
solution is dependent upon
(1) (2)
the order of the system being simulated, the area cf the search zone determined by
the
upper
and
(3)
and lower bounds en the variable parameters
the
nearness
of
the
starting guess to the
optinuo parameter values. Every trial ?alue cf a parameter selected by EOXPLX reguires a complete simulation of the system in order to evaluate the
system's
response and compare it with the desired response. = number of the T Rut1 The prograi run time is therefore, trials X system simulation time. A high crder system may
61
require twenty seconds of CPU time for simulation.
trials
are
If
300
required
to
determine
the
optimum parameter
values, the total CPU time would be 100 minutes.
Example
Lower
Starting
Guess
XS
Upper Bound
XU
Number
of
CPU Time
Problem
III
b
(pseudo)
Bound
XL
Required
Trials
-300
-34.
3
-3
55
7min03sec
.001
c
.01
.1
.01
0.1 1.0
225 225
24min53sec 24min5 3sec
400
450 450
95 95
500 500
400 d
90
100
100
2,000
•
68min
90
3.4xl0 6 3.5xl0 6
3.5xl0 6
10
40
e
20
30
50
60
400
.3
500
.5
600
.55
230
4hr
1150
1300
1400
Table III
Table
III
presents
a
a
tises reguired to obtain
summary of the search zones and solution for seme of the problems
The amount of CPU time reguired for
considered
excessive.
in this thesis.
for a solution,
especially
problem
III-E,
may
seem
However, there are several considerations which should be mace prior to arriving at this conclusion. The equations of the system do not have to be (1)
written, programmed nor debugged if
the
system's
62
-
component transfer functions are known. lime domain requirements do not have (2) translated
A
to
be
irto
frequency
domain specifications
for system simulation and design.
systematic search is carried cut to obtain the cptinum parameter settiEgs. This assures that
(3)
with
valid
bounds on the variables an acceptable
be
solution
will
obtained
with
the
first
the
cesign
the
optimization attempt.
The time required to perform the above steps in
prccess
by conventional means may result ic many more hours
of CPU tine than if the program
CADS
were
used
from
beginning of the design process.
Several means of reducing the
fcr
computer
by
time
required
the
a
optimization
of
were
previously
outlined in section II. keepinc
should
The most significant reduction is obtained
number
nirimum.
integrations required for system simulation to
Elcck diagram reduction of the system
be
accomplished
whenever
possible.
The
example of the Ward
Lecnard drive system shewn on page 26 can be reduced tc the simple system shown in Figure 30 by block diagram reduction.
tt
J
*s-rr\.
^o
flGUEE 30
K«
rs+i
^6
Z,SH
o—
Elock Reduced Ward-Leonard Drive
63
Section III.E.
cost often
will
showed the effect
of
using
different
is
functions
to
measure
the
quality
cf the optimized
response.
Although the integral error squared
a
criteria
used to judge a system's performance, the user should
carefullj ccrsider how to best define
properly
cost function
(err <
which
penalize
deviations
A
of the system response
1)
from the desired response.
small error
souared
becomes
even smaller.
If all the errors are small, the IES
is a valid cost function but if the system response involves
laroe and small errors
be used.
a
weighted cost function will have to
One nethcd cf arriving at a properly weighted cost
is to sinulate the system using first estimates of
function
over
tte variable parameters and recording the sum of the the different portions of the response.
be used
to
A
errors
ratio cf the
weighting
errors can then
arrive
at
proper
factors for each time section cf the response.
working BOXEIX will continue the optimization process, until it can no longer seventh significant digit the
in
reduce the ccst function.
the
Often an acceptable solution
for
system
parameters
has been found long before th€ ccst
judicious use A function has teen reduced to its minimum. cf CADS nay be made by evaluating the system response after
ten to fifteen minutes of run time to see if
an
acceptable
solution has been found.
B.
CONCICSICKS
cptinization using the CAES program is a straight forward process which dees not require an simple, Economic in-depth analysis cf the system being optimized. times does dictate that intelligent starting use of CPU
lime donain
values and bounds be placed on the variable parameters which
are to b€ optinized.
64
The program is
a
readily
It
is
usable
tool
for
simulation
if
without
cptinization.
to,
competitive
simulation
with,
not
superior
simulating
other
common
routines
when
typical
control systems.
This feature alcne is
expected to bring the program into common usage.
C.
RECCKMOIATIOKS FOR FDTDBE WOHK
All
of
(1)
integral
calculations
performed
during
for
this
investigation nere done in double
accuracy
the system response.
precision
be
increased
to
The possibility of using
single precision
calculations
should
investigated
decrease cere reguirements.
(2)
The ability to begin the optimization
process
at
some
time
greater
than
zero
should
be
provided.
This will
and
necessitate ncdif icaticn of the data input cards
block
eguations so that initial conditions can be entered.
(3)
The program presently
(EBVINs)
reguires
the
by
that
external
which
are
fcrcing
to
functions
within
the
and
variables
Fortran
A
be
optimized be
specified
IV
statements
placed
bedy cf the program.
method of reading these
specifications from data card input should be developed so have to "shuffle" cards in the that the user will not
program deck.
(4)
The craphical output of CADS was all that was necessary
for the
investigations
conducted
in
this
thesis
tut
a
prevision
be
(5)
fci numerical output of selected responses should
provided for detailed analysis.
The
feasibility
of reducing the number of significant
a
digits ECXFLX uses should be studied as
means of
reducing
if
it
optimization time.
reject a system'
s
Also some criteria might be developed to
response before TF is
reached
is
determined tc te unacceptable.
65
(6)
A method of automatically relaxing the boundaries on the
variables
being
optimized
when
they
go
to their limits
should be developed.
The
(7)
standard
cost
function
provided
and
all
user
developed cost functions should be normalized.
permit
a
This
would
more
direct
and easier comparison of a system's
"goodness" when several different integration
step sizes or
run times have been used in the optimization process.
66
APPENDIX A
Block Data Card Format
|1
(11
1
20
|40
P
|60
Z
ELKCCD=EEVV
CC
E
G
EE VV
G
P Z
*
= = = =
=
= =
PC5ITICN NUMBER OF THE BLOCK TYPE OE BLOCK (NUMBEB) INEC1 NCDE NOMBER OUTPUT NODE NUMBER VAIUE OF GAIN VAIUE CF THE POLE* VAIIE CF THE ZERO*
8
**************#*************************«-*****-R-*
co LU
cc
•—
aj
_j
cj»
LU
•
z o 1—
_
•
ex.
2
UL.
CJ LU
a z
X
a.
O
.
LU J> (X -J
LU CO LU
tu
LU
LU
->
kJ
H-4
co
O
•
«i
UL
ulju
a.
Z
i-4
•
o LU
a.
a
I— <l LJ
CC
L_t
O z a
a
1-
Q
o ^
-» LJ CU
's.
Z
—
CC LU cu 2. -i
•^
-r
LU
Q. J-
JC
CU
•
K**
I—
com
HJHH h-lOJ_J 2.LJ
<ltz.
•
LU
(_>
az
LU
LJ
u.
X
-J
a.
Z
•
•
oi•-o
»-_i
-j
LJK-<1 «-*o iC
Z
•
t-ny 1— u
cc
•
CO
I—
or LJ
cj
UJ
LU3
r— .«£ h-
Ui
LU cc LU •cu
_J
LU
»-«
—1 >k
4C
S.
^
4Z.
U O —J
co
cc LU
X
z
^c
L.J
V.
I-«J
UJ
M-
CJ •
ao
LUO
fsJO
rH
z
L_l
£X
(M
LJ ae •a
»--JI-0
X 3a
Da<rm
_l
o aj
>• CU
UJLJ vt^.
OUQ i/izujo
U_
CM
ai^
CULL.
u
<i
a
•JU
UO
1—
u_
o
_i
.£
u
II
UO IJJQ 3 9 X COCC zz CL
II
v»^, _J«a
1— 1— VI
«c
»CO
CC
LU
<
o
<z O CLO h CO ao ^ IUtH LUI— _i<r o MX a 3LCO <i«~* Mi _ cu 1— or*-" <z z CLQ >»CQfNJ
»-M
j
M
>-!»
LJI— —ICO O.LU
1—
<
vC
u_
CO
CO
z o
t-*
LU
3 O M
z <
p«;
•
.i-
I— CJ
LU
CC "V —1
4
z
LU
Z
LJ LJ z: UJ 1— CO
>•
z
LJ
z
•—
cc h-
3
LL
«l
OC
>•<_)
*<S.
1-4
u
•
x
o
OLU
3
—1
LJ LJ —1 CO
1—
1—4
•—
-J a. •
Ol— X-J
-j «a
1—
ttO
LU—
»-CL
o o CL-IZCC o
ace M(J
z<
•1-
cc cu
ox o z
LLI—
LU
LUI—
h-1
CC LU
<. 1—
LU
z
LJ
i—
z
LL.
LU —1 a.
3 a
CJ
cc
U-
CDO
LU
oc 3 1—
CL
CO
z
O
^v O,
CO
£Q.oiocoi— o —lUUQ-O
o.
t— LULU
•-••— a.
CO
_i
1—4
z H-ICO
CCUJ U.CJ
»-«z
oz«-«
<t
DC
t-
o
cc LU CD
UL
<o ZcC
oo.
•—
-JO
3
<uo CJ3
o
z«-«i-
oo »—
cc>
co co LU
.JCJL-S
LU—
LU
CL.
OJ<
<k>— -4CC a:<i
—ICO
<l cc I—
Z z 3 •—
LI•JL
1— LJ
CC LU
cc LU CO 2. -J
<£
o LU
CC
to
cc
OLU
0<X2.JU
<i3a
LLCJO
UJI—
LL
at— JOOO
<cC-l_l
OLU
1
t— —
ZQ-Q-O O •-• LUO
-t
Z a -J z
ii
II
CL
zco H^ZUJ H _J a. ox
1— MCC MMa
CO oaro < MLUO
ZU-<
CL—
3t:<j.z «lQC»-«
ozz z 33 < cc ooo ZCQCD »»-4
to
o o
_l CO
^
LU
CC
z <
cc 1—
2C
LJ
o.>
o
o
—J CO
u.
LU
LL.
CU z.
3
Z
LU
CJ »— LJ <l
LL
CO
>
I—I
t-ccac at lulu <tO-JC
u_
u 33 U.<X
cc LU LU
-i.
1»--i
LU
ii
O
ii
II
LU
j»"*» «>*
CL<
LJ3«1
i— cj
II
i-clo co3_J
ii
ii
a z
I—
l-LU
3LU
o
•—
a.
LULU*£.
z
* * *
—
-"OLL
HII
Sf-M
II
uu
II
<l
ZH
<».
•q.
hLJ _J
u.
i—
3:
i_)<
<l
«i at
1— N
—1
i—
*£
u
M
1—
H w«»
in
— **»—. 3 z
ii—i
II
ii
H*
CC
c->
1— LU
>Jt
Q o o CJ Q O -1 —1 Z o H- O Z cu CO 1— a. z »— u£ > > u< O 1— 3 O o <l CJ LU CO 3 Q. a. LU z UJ LU I— 2. 1— —1 3 a. z 3 a Z > i— O <l CJ < o I— a 1— M
cc
ex.
LL. LL.
II II
QOOOOOOOOOOOOOOOOOOQOOOOOOOOOOOOOOOOOOOOOOOOOOOO
cm cm cm cm cm (\i rvj cm (M cm cm cm cm (\i cm cvj c\j c\j r-j cm cm <\j cm cm cm cm cm cm cm cm cm cm c \j cmcmcmcmcmimcmcmcmcm
II
CC
u.
»~
z • X
1—
•»
m
•-»
LTt
t-t
•b
CM
•»
— in**
•u_
u_ H
•-<I
•
-—*
•
II
V
1—
r-4
CO lu
V*
o I—!• - »—
II
z
»-i
ex UJ
1-4
>•>
•
«—
II
LJ
LU
,
< ^
o
LU
_j _j
LU Q. 1—
oc LU
1-4
4»
LU
«k
CL
u-co
o x CM
r-
»
i
CO «a
i—
l>l
z X.
>•>
<t LU _J
ex
« •
CM
•
•»
A<M
••
X
«a-
•
j» "
*-*
«_
H-t
X
CM
•>
»u>
u^ •
^•"Ol _J—.^H exxLu
LUUA
CL,
•»
—
I—
X o 3 X LU. u. K «— r—
a
iX
-J
•»
in
1-4 __^
•>
O—
9*m
•o
•—
* —* #
—
X. -j Q.
or^o.in
r-iU}
*LU *LU
Ui«fc
o CO
X
>-J
•k
a.
o
»—
MM CL
<_>
•
II
•
II
II
ZUJm
-c
•*
JL
CO
a z «*
tm.
X
-j
a.
•»
1—4
•»
>
H-^
(_>»
tjtw LUUMJJ •-J
cxu. U-CJ
z a
CO »—
•»
ex
a.
•
X
_J a.
<—
»— ;>
m
z
CT»
H-1 a. a.
z »
•>
<t
•
CM
«ft
XCM
••
•>x
—
W) v)
aex
••
z
•»
X
CM
^ hZ
<.
~t
zz
•»
•>
1—
a
m
•tn
«
lu—»i/)V)«a
^ ^ —»ULU zexz
- 1-4
_j <i
a
—1
o
z
1-4
LU••
-J—
<ti-i
<l cx 1—
»/>
•»
X
OQ
i-4
i•>
m
•
il
••
•o
u
CO
QC.^
CL
—ico
Q_
II
Zm »ct
•»
II
O »—
-.
—
2>—>LU
<ll
II
1
-JO Q.O
j^
oo •
•
l_»CJ>
o o <a <->t~ Q. o >xz cO M 1— — Z'-'> LU o < ote a •<I CX o — X OD 2 HH X>TLTt i LUvO-"" LUX
<
z<— •> •V)Z
.,-..->
V*
|-|-CM|-h-»-»— »— H-l— 1— »— ^-»— K <X.<K. -<K.<K.<l<t'a.<lO-<l •<<<«!
X
_JtJ•-•
<l»-4t/>
>l-Z
iii
—J
a.
ti _J t-4_J U_CX«I.L_'U-«LC_ICXO£
U
rLO
O.
2.2.£in2.X2.JLZX^2.2.2.£2.2.2CXCXCX*-' ex a. us. oc uc. ex a. ex jjl ul uc. tx. uc. ac lj LJL-ILJ "A-JLJ l_)LJ(_|l-;t_JL_lOL_)LJL_JCjrj^: U.LI_U_" U. U- UL LL. UU U- U_ U_ U- LL. U. U_ li_ U_ UJ
oooooooooooooooooooooooooooooooooooooooooooooooo o)-r'nvoi^cocr Oi^cNini^tf^^r~a3 0^0'^f>Joi«Tinvui^cucT*Oi-ic\i(^
i
CO »H-l
CJ
<i
0.
2.
H-<
in
CJ <i LU CJ
-» IM
-J
* *
r-4
l>i
^
CJ
«—
z 2
+
1(X
—
CM CJ *—
•"*
a z 3
CJ CD
-J
cj »-
** LU
2T
CO
*
*
z
<l H_J LU
*< CO
o
e>
—» -^ CJ LU
^ IU >
<—i
O) 2.
OI
-^
Q.
—
2.
1—
V
u
CJ
A
Lt_
tX
ex
L_l
Z •
•J LU
•
-^
*
* > — a. m CJ CO s w * w
+
i—t
LU
a
a «•
+
Z <
X
*
«-»
v.
»-i
—
—.—
OIO) 2.2.
•~ Or LU
LU
cx
O
II
•
"^
—
«M
>T QC
1-
Z •
_l (J u.
"""•
m
2.
1—
*'
«a
j_
—
Q.
f»l
«•
r>l
3.
«•
M
o
2:
£ —
—
o»
—><«•
II II
VJ Ul
z
CJ
•~
H-
Z o o t— z
—
+
CO «• >^
+ —
C\J
cy
_
m
<i
»-»
m z
rsi
* *
CO *^ V.
z
Z
>K
cc
O.IM
•a
cy
CO
—
m<—2:
*•(
_ • u
<i.
*
C3
II
ZfOO
O
•»
o
II
CX
X
z
•h
oz— 2. •~—
Of— CO S-CJ2T -*l Ljw
fOOCi. 2.2.
1
X*
CM
X
»-4
II
•>
<
X
*
»» nini a £Z — <<i z mxx 2:»—»—
IX.
K ^ O
CJ
»—
•«*
•
u
Cf.
H~* •^
LJ
u.
m
s.
««•««
•"
•
CJ
•*
z
<a
z <
1—
—'
rsl
r«»
2.
~~
LU
»—
•
H
•H •-•2
+ •
t-c? COLU
<t If—
CJ z •— m <l C4 5. X < CO X f!-« Cl K >* <.-¥ ^
l-«
* -»
<_»
f— CJ
Z t—
«a
_4 CM
00
••
z 1—
<l
LU ->
1-4
UJ-»'—
•—
^>—.-»
H(11P1
IX.
—•
—*\r>
•
1—
X
K
-J
•
"™*
> »
II
1-4
•»
X
X
(X CJ
—
-H OJ
•»
<J
II
Qw^.
2\ 21
in
CJ
a.
•
*"
a.
•-•^-1
1—
^*l
1
;
C-»
».
cov.
a. »ff|i-(^
1•>^
1
coi—
—
II
Z3 CJO
UK.
II
J-H- r»— •-• i-4
zvOO LUO LUX
cut—
<X
us:*
C0»-
11
2:^1- fatNii—^-»
—<I
«..
f*"i
<l
*-*\S)
<lOt>
oij£
X
C\JJ5 ».
ujmO
•
II
«!m
.&.
-^w<t
11
muj
mo
•
11
je—
*-«
1
ZD zo MUh ••2:«-«rCJO c?o LU-4 U.O«l LU<1 X ^ JE — X
•»
OSHOJ »H
•MM>^
•»cn
—
*Q.r>j
+
z
• •
#
*•*
• ar-tt—
CO»-
mt— »— + coO
2:0_Jh-<LU
>-»
H-
ro
•-i
e>— — <imi—
11
• «^- _j
a-t-_>
X
l|M
11
LU»-
—i—"
•-4
w
f—
£.
CJ —1
*:
—*
1
X
V-wO f—
VJ«—
<k
_JU_t_J
1— CO
LU UJ J>
a cj
*a.
CCI— CO
II
»— CO UJ
w u» _j— 1— ^m
cj
»l
.
O U.C0
LU
_J
•
f-4
1
^
mm^ 2.2:»-<
*^"*
11
:•-•</)
1
a.— cjm
11
t—
— <t<i
<4.
11
< -J'jy
tr\i-» «-"
CO^ Z.O CJ-ct LUX
I—
X.
*^ cy
II
Z
-"»•
<j.i—
Xi-i
-.w
h-<
j:
OICOl-l
.2.CU
WMO
U-_J ^*CJ 1— CO
HH ^— <L
Jfc Ul_
Ui^» U._l
-*-CJ
Ocxzf—co —>i CO
ii»
*
VL«J.
UJ.> u__J
UZh> » "^
<
U-CO
LU UJ^» U._l -^CJ t— CO
mxx
1
-4
«»•
— <t
II
11
ZHHI-
»— •CO"—
COCO
LU
ma.i_»
** »4
z:
o-'t— "^1-4
•••iiimi^tiiii i«i^-'»—
j_
t—
CO I-Ht-t
t- CO
J_-*"2. u> cot— •"*•-'
^CJ
t— CO
^»— ^->
*t<l J_-» u_
•->«
<l—
a.u.kL.-£.u>u_i_i h— 11 11 11 11—ni>
LU.> U._i ^-CJ r— »/;
<l O.U.LL..& >— II II
1
1
«JCJCJVJ
CJCJCJCJ
CJOCJCJ
CJCJCJCJ
CJCJ CJCJ
92
•
•
oooooooooooooooooooooooooooooooooooooooooooooooo
cm cm (M cm cm cm cm cm cm cm cm cm (M <\j cm cm cm cm eg cm rg eg rg cm c\j eg cm cm cj cm eg cm cm cm cm eg cm c\j cm cm cm cm cm cm cm cm cm cm
f > >* »— LJ oo O— ft ft CC o pi OP-' o o CP«J'_) X LO o pi LJ • ** 2. Pit— • <l o jr — O Q. » a. X <IIZ ^»f — * t/>0 t— 1— oo ft • oxo LU CM o x«— o •o X z: CM »<i*-s: ho zu -JO ** O-X xou oz
•*•»
t
LJ
»
O
X
•*
Z
•»
CC
«*
LU
•»*^
rvia.
*"•
z K LJ O
X
CC
•*
CM
*
•»
»
CM
I
II
1
•k
1
•»
II
•
f«0
I— u_ LJ
4S.CC
^* >r
X
X
II
•*»p-ii-«
XX
O
LJO
I— CJ »
»—
CJ
II
-£-
I
•
-•WO
P)
JL.
II
1—
i_i
oo ^o 1— 1—
II
o>o mo
1
-»LU»-»0
Pl(_>—»|—
—.--IJJUL
mz
3UJ
ULULO
ZLJ-~LJI
z ZXiU.
<to
•**•
pi LU
II
PI 2.
o
l£
LULO CJLO
_!(/> 96
*"'
P1PILJO
2.2.-J1— <-> «^— JtUJ^.
w\
a o l-«^--z LOZ <_)LU|— UJ_J —i—12. LJ
»-««l
t__i>—iir\i—iouqolu H-LOPJLO-fl-*
#Z
I
—
-J
CC
II
PI
*—
Z
LJ
CC
ii
II
»-
tLJ
l>4
xcc
z
—1-J
2
LJ
imim
1
*;-"*
l_l
^atuL.1—
j-j_x<j«.o
^.o
i>j
«ia
pi
XCCfiLJ
>r
xuaauj
J_£X «— t—
^. 2. LJ 2- <«•'** <* •*_;>—iu^ojujvli LL. LJ rsl CJ Ct CC OC LJ
DIUJZ<JJiU. O o o u. OJ*H/>t— (_!•—•-tU-oOl— LJ a.
CO
ac
xz:couj u-a.ee zqluj
j-iua
H- _l CJ J_ hLU
^ IX
'-»-
o-oc
co
CO -J
M^«l
uuuu
o
uuuuuuouuoouuuuuouu
100
» i
>
*
'
i
—
' >
»
<
1
oooooooooooooooooooooooooooooooooooooooooo oooooo ojrovrm-jjpt^l^^l^sl(^Jc^JC^J(^4c\i^N(\Jn>o^^^^o>al(^lmo^^l^^^r«T«T>r•<T>T•«^^ru^lr^u^u^u^lnt^^Lnu^u^^u^o O O vO "O >o oooooooooooooooooooooooooooooooooooooooooo oooooo xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxx u.aaaaaac(iaaaaaaaaaaaaa.aaaaQ.aaa.aaaaaQ.aaaQ.aa ll u. a. u. u. xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxx CIJaucD^clJaDaJaJaJaJaJcDClJcDaJaJcDaJaJaJaJa2a^cocClljcDCL cucoajaJcQcuajcLiajnjcLajajix.
vt)
a.
,
CD a. oj aj a. cu
I—
• »-«
UJ
OcO
uj
uj
»-CO
LUC J
ctto-s:
•.
to
t
ii-JV)
a:_i uj_'i_) •—• ju_r~«a.»—
—1
•V.LU
|
UJ
02*0
e-cjzcc
acuja.^ujx
<t*-i-XcuH;>Zt-<t—
UJ
•/)MUH.M|uj^ <aa<z
•>
cc CJ
•»
ZUJ
'Z3iC
^tOCJ —
•
a X
z
CJ
•-«
»—•<_!
ot.«a.
U.Z
UJ CJUJ UJUJ
e-cy
«c CJ
1—
<-i
UJ
JC
ujfNia.c_»CJ
t— _Jl-r t— UJf_J •— »<iuj
•+-2.
CJ
1_) _l CJ to •
CJ
—I
—
(X UJ
•—
•»
zts Uj>xm<IZ
Z
>-«UJ
%-i>-*+*v}x to
GCZtJt-»«CJ
h-«c_;uj
V7LJ
uu
i
-Ju.u.a.'a.^cc
djo-
>— •a
s: ac
UJUJLU (-2.CU
a<:A uj oj'-Kiia. •JiiO'"
ZC3
13
CD »Q-UJ
xt—
uiiii tin
hi< j h
I-30DC —>CJ*-»CJC— ^.tr; CJZ O
»o:_jae.z<
Q.f-2-h-(-UJ<
<a
Ul
CJZOT Z<1<«
-J <l
QCtOf-
CJ
u.
:jt
ju
<l
cqcjcx *uj
£ >•
•»
z
-J
X 3 X
Q-
X o x CJOUA »»r~ O h>-LUUJ UL
u.*z-jae; aucJt-tuz
2-»-«*iUJ<l
OOOOOQOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
OOOOOCJOOOOOOOOOOOOOOOOOOOOOOOOOO--<-^»-«t-<«~<^H^^-<-<'^-<f-t»-4-<-^
^;^o^l^l^p•(^f*l^l^^*(^ooa)oow»oooDUJODoou, y, o% ^ru, u, o^u<u^u^oooooooooO'^-, •-, •^'^<
,,
^
X XXX XXX XX XX XXXX XX XXX XX XXX XXX XXX XXX XX XXX XX XXXXX XX
xx**xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 0-aaiiaaaaa.auau.aaaaaaa.iiaiia.Q.aaa.aaaaaaaaa.aQ.aaiio.aacLau.
X
Ltl
CD CD CD CD CD CO 03 CO CD CU CD CD CD DQ 00CO CD CU CD 03 OU CO CD cOcOCOCOCOCOCOCOCQCOCOCUCQCUCOCUCOCO<XJCUCQCOCUCOca
e
«*
o
tto
to UJ
ou.
h»•
1
z
•l
-4Q
II
OC
ac
1
CO lu
1
•O.LU
au
»JJt-c
•—
v>—»^.>-t— OL<iCJ<lOtO
U>
»->o
LU
LU
•
LU
L-IU.:
M<1
LUU.
u.t_>
"-*£«.
U-i_J_J
QCH-UJ
• COOJ OV)Ui
t-rCJUJ
•
cut— to
isxu
•""Ou-ql^uj _JC£->3
LU
-J
»-*LU
or.
JUQ.
-t^ai_;'X LJLJ_JI—<l
_IUCQ>-» <a.zvj
• o -it/)
1
O
t~>
LU •
<1U
X
JX z u_ 0L
->
•a.uLU
X
zd-jhz -i->^J
LUJ_
J-l— t-
<
1
LUI— CJ X»-iuj
»-<_)
LUH-Z^UJ
•-•<
IMQ<J
2U.0LCX
<i>_?
3 z 1— UJl—
»-»l-
O—UJLU (I3XII— CJJKLU
z_i— at
<i
X
i—
lu
lilH-U'h-
3>««X
• «.»— CJ
t-i
2->CD
o
9LULL>.>
• LU
Z
«.
LJ^t
'_>•*
UJ LU
«l_JX«a.
>-
ZOJtO
1-
q
^to»—
lox
.-SZ
z:
acu-x
Ol»~
^1— —ICJLU Z0L<1 -52 + <ll—1COCLH— >- ) o<t
Z> ¥—»— LUO »—
ax^> OLUI— i-«LU •z ^ r-x
1
K-tjtoa.-j
l-H>
12 CO
-£.
O-JCD
•
2".
t—
< ujmuj>
LU ac
••
LULU
uoj
1
< 1
•—l-J-~
•-i^CJ
(— UJLU a.
MX
f—
3E—l>T
*.»-•* <l-6_l
ct
• <r
(— LULU
» ZVJMZ
,
3XLU 3XUUZ aoi- Q.OO «-••-«
Z<I-J
•— UJ<I
••—
COXQ.
>—
»-»-*.>•
<l 3 U-^KX^ mtvM z <T UJ CO
to*<z
ex. tin-
^CLU2.^
>-aL
<y,'L_)
-«*-30to
*-»_ij_ LL.U.I—
i.H-1
uJiUUi
>2_
»-*XCJ JCOLLU
lu
IU»_>i
D-3
OQ2>W X
UI«-«
oo<
-J
•
K
xo
<?U.UJJH _l ojoj^—>a. —iz wo<
-.f—
1-4
e>touj CJLUCt Ct_J U.UJV) <t<l Ll-I CJXCLU
z<ri
*-•
»-«UJtJ-S
•
•— XO ua ZO — UJUJ> UJl— o-c -33 U.I— CUCJ«*-Z ¥-L-i-C cuo 3^ l—J_j-iCt CtLU<l O toi— at 3 i-ulu CJ h<»JJH-H>CJ4i a3 Xm tuz LU— lus: LU IN t-a X» - ct h-uj CJ
•<£ LU- COD.
1— CJ o. u_
<J.'X
>-3
ol
-» ^C
^»
.>
^s:
-!
LU»-
_n/i<: «xlu(_)
LU
23*-*
o l_IUJ
OCX
LIJ
cclu»
LJLU
LU
Z
jxatai
1
»tO
«*LUl_l>-^.
-«—
CJZ
—
CJ !<
-*•»-
UJKU !--JX
ui>ct
acH-
tooLX^jc^a: X— CJDLCX _J
LL.OLUCC
^--iLi-ai—
LU<IZ
• Z»-« •—»— LOCM
U-i-<»
1
_JO
H-
•
J.j^U2 ^<li'
'i
i
LUto_iZ
i
^
•
h-ODZ etf <»-«Z LUUU< ac h-xo> oaatoi— • UUOC > LU
•—«CD
to COO
tocoz<i3
\-
o
<<»-<
tUt-»—
<<<">_j
LUl-4
Z\-h-
LU_£
^ujoKHu-^a JO-J UJ^U.U<1<1 <IZ-I
Z*-» CO hh^»LU LO«-*_j
mZW LU
ZLUt— OQ-t—
I—»LUtO
»->LU
<a
-JCtltO <t
—IOC3I— to
o
»->
••I
<;xa
tootU_t-i
h*
Oh OiJ H
UJZZ
o-
(_)•-* Cfc«-«
UZ LULU
!-«
CO
ac
ZQ • X LUX.
•*
*3 ZO LU <o LUZ LU'V)
_i-J
a.
i—
,
»-u-
H-K
ctoco o «a: IJJ<lULUUJZO
</>Z(—CU_JXCO
Uuuj-cj.
Z>UJ
t—
til
>H-OX_lOLUUJ
^-^
|
.
OUJU->J-X>-iCO
1h-
i'U<XlU
LU<tZH-»—
JLLU
*-«U-t—
OI-IXCJ
o
<I
ZtMOUJ
i-:>z
•o
t-«U.
HU
o.z.-io
DO
—a «LU
<i
ZfH
l-.U.
(— OV)H-«-i
0_J
ZOH-
333^-i
<a:>cc
_JUJQCU.'3
0L33LU
LULU *<X CD3» 2ILUOCLU
Oh-Q.
tOQ»< »zz< CT^-tO LU ZLU> O t/!XH «JOCJO QfLU3 LUO > Jaazz OX-l SZI— <Z XO"-"-« U-»— <1 > a a. KmUJOOZO
<•—
?
>u>o^«o^>0(^^-i^i^i^i^^r"^(^(cooooooiDoocucoooa3u>u^u»u u, tpy, u>(^u, oooooooooo<-'^
,
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx u.aaaaaaaaaaa.aaaaau.aaaaaaaa.au.aau.aaaaaaaau.aa.u.aaaua xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx cu cd od cu cu oj cd cq oj cd ou oj cd oo ou od cq ce ao cq cd cq cd cq oa aa oo cd to go co cu co cu cu cu aj cuaj coco an cu coco
go
XXXXXXXXXXXXXXXXXXXX XXXXXXXXX XXXXXXXXX xxxxxxxxx u.aaaa.aaujaij.Q.u.a.au.aQ.u.a.LLu.Q.aaQ.aaaaLLaLLo.LLuaaaa ixaaaaaaaa xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxx CD CU CO CD CQ CO CD CD CO CD CO OU CQ CO CD CD CO GO CO CD CD CD cucococijcucdcucocdcccxjcucucucucoco CO CJ CU
CD CD CD CD
CD
I
CD
(J
CLLU
»_UJO<ll—
(XilH<i<ico^:.*cuj hi e-»-
UQ
>— <i
•
CO
COCXO culu^
2 CJl-l
iCJuja.^.
3
+•+<. •^H_J_> -•<-» _J»~
I
— >-<:
-
o
»_l
I
I
cji— «^:
LULU*-*
H-<1
<li
aj<t
^^.<i^:'xi
cxujlc(/;cx lu
UJU.LU tvlOlil
«Ui<iu
hhH-
oca
o
U-t— IJJ
<lOI— Ol— «i
•-It-lCX
ZIZ
•cj»H-
<x
OH-CU
4Z7ULULJ LUCX <l cxz xua »cj- ac »-• tp»-i^-jt; -^z
LUI— XCJQC
l-3_IUU
«£-3U<l
LUC)LU_JO
CX
OO
I
<n_i:>ocaLucu
JL12.
LU>-LU_J^.
qcz
LU«-—* coco t-i
LUU-(_Jt— <_>
uaau
x
OCO
<.
xculu
Or-D
^
CJ lucj.£
co'x.^cur— - "ujii/)
t-«CXLU<l
**«/}•
»-«
u.^- 0|— ^c U-LJLUULI h-i CJ H- —I CJ »-• u.
xa
ZLUO.
aci^QX
«1<J.L_>^-CX <l _;x
CXI— XIUJQUJ 3i-«»-«OC <Xh-<X
^
IOLU
^ILUCJCO
^
<l<*.
LU X? <L LU LU LU LU
CJh-LUCJLU
CCUJM
LUI— CU
•
:elui~_jx
<IQrDUJ»-»
lli— t— <.u_ccz <£C02. fH oji— uluj CX LUt—i. 1_<CC •»«•**•! <si->c»; <lCi> CJ
I
u-o
IX
t-OOQtMCCU-IZH
0»~*l_)CJ^ ^JLU LJI— CJ^>LUU->"-»i. K-^H- -J i—. _JLUUJl_l—JO|—
LU CJ >»0<IXH-LUQ.LO CUCJLOC1.H-II— H-UJXJ ao. —» Ouj<3.H-Ct. <l
XXXXXX XX XXX XX XX XXX XXX XX XX XX XX XX XX XXX XXX xxxxxxxxx u.aau.aau.o.aaaaaaaaau.auaaaauaaLLLi.LLau.L.aaa.aaa xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxx cDtnajcQCDaJouoDcQcocuajcDCDcocucDCDajajcocDcDCQUJCUCDcijaDcDajcccucDCDciJciJCDa^ CQcOCDCOCOCOCDCUCU
UJ
JtX UUJ
UJQL CULU
1—
u a z UJ a
UJ
a. l-
COI-
O t— Jt
•-•
1
JW 2.UL
o
to
1
m*
1—4
•>
-^
X
o *
-£.
»—
-Cl-
ctx
<£.
—»UJ
t-Z
rf£LJ
u
(X.
• ^v
UJ LJ
•*
•
UJ
LV
z
U_
Z
1-4
3
•»
1-4
>
->
•»
»(X «t 1co
a.
a.
w Z 3 u_
X
i.
•»
M"J
>—
a \_j <i
>—
<l j£
UJ ecr.
•
»
it
•>
XU.
UJ »a. •
>
QC
UJ
1co LJ
-£
t~
> z > z
c\j
o
a i— o
a.
UJ —1 UL
.>
•"
UjU
•
QC
1—
CO
CO -4
->UJ
a o z X
~>
•«
1-
»-a
»o
> a
»-
M
X
1—
hC-)
•»
t•»
-4
tocc ocu.
UJ
—j
UL.
_l <t
1-4
z
Pi
j»
~3
•»
o
1—
02.
3>-i
•h
UJ LJ
»-
Z »
UJ
UL
—
1-4
•»
Z UJ O o H-
p*
h1-4
z z H*r*UL
u_
a
•»
UJ
UJ
J.
a
e>
->»
a-i
.-iUJ
x>
Lucr UJ
o o
•^
O z »—
(X
UJ it
(X
o
<_j
H•»
*H 1—
F-4UI
—
cut— 1—1-^
r\j
UJ
o z o o Z *
I— a. 1—
~3
*—>-*
:>u_ + >»
.«».—
M
•
1-4
-5-3
o--*Z
LUCO _J •+-
• «-4—
=»21X Z->_>
•*/)«/>
1-4
II
II II
o m — — o o 1— • < w UJ UJ > u. z •
•-* 1—
-«•»
UJ -J
U
—»
X uoz » UJ
•—
1-4
o
»1
h-
**
•»
zoo-
H-
»—
cc
z 3
UL
•
r-lOZ » —»—
4H za HZ
s:
>-•
•»
UI2. 1—
• (/)UJ
>
CJ
UJ
II
G3 »-u.
•
Z
UJ^3
>-J a<i
2Z=»
<
o
»-«
z
UJ
>
UJ 1— CO ac
(jj
UiOZ
Kn a^UJ
-JH- • ujcouj
Z -}
«
-h •
r-4
UUJ
•
ou
it—
••-3Z •-3 II <J»—
UJ
i—lUJ^
uo
coz
U-Z
1-4
UJ.-I—>^UJ
a + z + u.
^IJJLULLi
:£ •
•—
QCZ
Q
LJ
II• -J
-o_J—IUJ
1
I/)UUIL
LX.'-»CJ
"T
-3
a.
ill JJ
CO
—ih-—J
i
no >TT5 »>-»
1—
F
<
— .,,
(^z-<:
<(
II
UJ
ZC
II
~z
H"t»-4
M UJ ^)
o -t
Ul
J-*-!
CM -3
II *" —>*Z
•2
UJ
u.
UU
a.
f-l
QCI—
UJt,>
a or
t—
r—l_JLX_
*-^.
"-»!o uj w UU*£*£
acta _iQ.aj Z
;
OlLMf
l'*
II
1
_3
z
_J
acti—
l_JUJ»
«J UJ
•
_
i*
i._>*H
Z
UJ
U>' •--J
rii
r—
LJ
-J
—UL
1— ^.
joz
uoto
-3
II
• Z3*-*Z -« COt-"U_
1—•'-••"^
Z
ii
au.
a
fc-i_
Z~3
**
»—
<z.
a »— >—
2.
II
UJ t—
—j
2.
•-»
iu«o
1>J
UWIU
l»l
K
>-^»
ut
CJi -**_J
*.->
«_!»-«
«•*
-T
r-
UJ
z
"3C
«o
<aux r^
Lu^t—
uj->»-
iji-imij
fXi
a.—>-j
t-)CJ
—
LJ LJ
LJ
CJLJ
DO
LJLJ
uouu
UUU
CJLJLJ
109
• I
z
J
OOOOQOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOQQOOOOOOO oo u> o cm m >r m cm o >r in o sr m so r» co o cm ro -r in o r- cu o iH cm m «r in >omjo(^ r* uj <o mmm u^ a in in in o^o^o^vo^^oop-p^r-r^r-i^-r-f^-i^-i^-oocoooouuuooao oo Gu<w^cru>cr u o v -r -r >r «r «r >r <r «f «r *r >r <r -r -r >r -r >r >r ^ «r <f <r <r ^r »r >r >r >r >r vr >r »r *r »r >r *r >r xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx u.aaiiaau.aau.aaau.u.aaaau.aciaaa.aaau.uu.aaaaaau.aaaaaaaatLa. xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx fficDajcDajcDajcDiijajajaDcoaJcumaDCLicccL'maDmaJcijaja^
s<j r*a
i-h
i
so
cr>
•-« (\i
»
lt>
«—'
cr>
>
>
%
>
i/>tr»tr«
-j-
•j-
-4" •$•
>*• »3-
>a- >*• -4-
*r
-r
X
-J
o X
CC a.
»-•
iu
a
QC
U)
•-•
h0£
O
IX
I-
Ul
•
>
*-+
2
I-
U
CI cc <t
UJ
CJ uj
CD 2.
2 O
U.I
— -5
••
*£.
UJ
(w)
>
0C
111
oo— cu—
LJH-t
z a
O.
•
Z UJ a
UJ
u_ a.
a
— —
»-h
3 O
•—
(_>••» ocx 2
<LU ^»oor
»— uj
;>
Z> u.
••
l~
h-
sc
•>
*~>
UJ—
I
cccu
<
—
in •
H-
oo—
"SICQ
—
•>
O >
k-
—
i
X
—J
H-
i-.
K
'
+
Z" hJ
cc •• h- t—
k-*-"
Z
h-i
n hu jj O —
I
vo
—.—i\i
••a. —
.
cu
o
3:
-di—
•
•»
_,l/J
->UJ
t— ~>
tuh-i
•«
cm
I—
#—
;>;>
«i cc
<il—
co>
u
I<£
a.i^ 2._j<i ^•st«. 11 ^>u_ v,—.
1 I
Z— oz -•
1— s:
H«/>
'2L
OIO O
—*
CC"
U_ uj cc
O
H-
hh
"*'I— «1
+3
«w
uj
z
2:
— m
«
"f
OCD
2: lu
Z
z
>
Uj
<_)
*
•»
>-x
coi—
:>-» * I— <il—
*— >
X^"-*
i-i
11
1
> U
••
UJ
——
«*
t— ->
«
x_J
uicc I— Xcm CCH-vT uj
——
~3*£
2 *
UJ u-
i_j»-«
v_»
o
H-
> — o •
i_j
uj
a.»-«»H
•--«
»-•<».
-j^^:
«a
2:
o-~
«a-
ii
^ — 2:xuj—
i
mh o— —— hh«> v*2 —IX
Q.<t X2S.
IU
K4<
•-* f-«
.
O
_J Q. a.
t-*
< OJ*UJ»— zz 2:1—
t»
«•-•
z o
•-•
;>oo 2
21—
uji-4
>
+
*
-si—
11
• —
uj 2:
_J
-• -j
*-*
— U_
>
iiit
•
«3
»-
*
>-! >^— X
»-•
->
h-
<i
— m
II
^?^3l_)«-l •-joooo* •
uj<
||
<CJ
||
||
-•
2<-« » •.»-« t-IUJw
II
i-«
•-»«-»
cr»
M wv H «•
I)
tX Z2«— X K >—,-*CDQ. H— UJ 2 a !_»-• -5 ,— u. m —» ^
11
)— uj UJ •
UP
UJ
H-X2
x
•—'
22>»-l*.
•**
11
»-•
•<».
||
»H<CJ»
»-»UJ i-iuj • <r
•-•> < — o >z— 2UU-^I— > »^-> — •>
>-»
— I-
iouokz
t-»UJ
-JH-O »Q-
OO—J—
Z
••
<u_o—
I
I—
rH
11
• uj'—--»
|—
-»|— ucd
(M
iiCCOUuj21
UJ
•»
11
im
->—'—
*-
1
U^—
i.^
/uj»— u_
1
U.t-1
upi/iup *
—;—
'
1—
to->
^
hhvj«
,«4
u 2 "J a. U
l->UO
X.
uuu.
i_i^«-»
—
- CDQ-— h-i-3 o n — • 1—
ivi
z
•-•
* uj
CJ»
-*
uj 13
II
f*wn •
•Q.
f»1 i>j
11
'
uLr-ki--**.
uh-u."
"H
uui/»-»\_>
u^>i>
«-l CVJ
u
cy
<t uc »—
H -J — Z 22 ^ •-it-' —
'
*oi
*^
11
UJUj •CM—' u-ctujin -J •+t— a. »^j_) UjOJCX.—'O
OXX
i—
U2UJ
CD
*"
«^2 uh-u.— «_»
«M IM
1*1 1>J
»h-
lit—
<i I— *_»
>uj
ujuj
U
<_>
4.
u^»->->u
Zu
CC "J
«-^
UJ-
—1— —j
•—•—
*£uju_Lt-*i
V? im
M-ua
1>J»>J
UU
»-«-Su ^ <icc
CJO
<-J
uoo
u
U UUU
u
o uuu
110
»
oooooooooooooooooooooooooooooooooooooooooooooooo 0^*Mrn^uv4jr^co<j^o«-«MiMNru^sur^uju>0'^cM^>ru\vur»cp;j>Of^cMm
OOOOOOOOOOi^«Hf^t^t^t^^H ^^4^Hrsj«\»rvjrvjc\jc\j(\j<Nj(\jr^r«»nKMoimrnnir'in)fr»^->T>T«a'»r-T >r^
in in in in
m in m in m in mm mm mm mm mm mm xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx au.aaaaa.u.au-a.u.aaQ.aaa.aaauaaaau.LLaG.aaaaaaaaa.Q aaaaaa.aa m in in in m in in in in m in m in in in in in in in in in in in
^oo jc •*-«cm _J2 HXCM O CM »zod s: -a » Q2u. o^i^h» hj# ^ o o a — h-u-a. Ma^o o i— z *• 3U.HU LU o ~5 o -»0 O * Z>— 0<ILU *<3— »— O Z3cc:cO U_ — u_j— co _j-3 »x o * ^: 2<l o ni^z X > U_Ow <I + O— Uj mU.f4 »fn I— -huj •COH"
"<r
•»
<;—»_J
o
coz _o—» QU.UIh
••—
«— u.
—•—•— -3U>»-4 * *—
•
* —
— m
•
O z m x
rn
CM
ceo UJO
-iz
~5
o— O Z «-»—
'
COLU cccc
I
o O —
uj
h-i
«— »— ca
+3 >.> * **— t—
_i<tz
+
t— ->
lu
<-«*
jzcc uj
o •
<H-^> x-£-~
a.i—i-1
— »—
-s-c
—'
o
I—
J— cm
UJ :>
j_
*«-•
•
I
o— zz 3 h»
»-•
—»;> —io
2;
o
e>
^-«
-I
Xlilu •
1— t—
COLU
<o*-* •"•!—
2— «s: >Z
^:z3
*+
••u.
h-i<i
<i
I
ii
•.
«—
<
-* 2:
>
+
11
—
•— _j
I—
2T-3
UJ
•
21)—"-'
+><-*
•-•
^-.
LL. «-»•-•
•
OC—
i
*»—
•"
>**—«<
r-KCCJ
II
X
a:>cu
LUUJ
t
—»x
U^»
•>
LU •
lucpxh-u__i
JL
Ul/I<
II
m-IUH
ujin
-J
•»
>>•
Jt>—
•
CC
M
^^tL,
*t
2: >. UJ _J II CC ^Oj 1—iiti^
UJOJ2. ^/)«a<
t—O*** «^:CO
«-•
^_
O
c_»
O ^_
LU t—
II
II
UJ I— I— LU<<
tj.?^
Z ^^.r-t/I i-V)w ^^ u_ ^t^tt a. a.
II
C LL
•MJtMO u.oq.'— >r
Q.
oz
2—
j_
»
-•
H-2.it «z;i-4w LUH-X»-i_J UJ_J**^» cx-^t— oca
xx
«£J>r-»i-
»* •<
I— UJ
9
H-i-3
II
— >Z— 2LU«-"H«^» --<— CJ
•<_>
'
-*•— »-«>
Z
LU
• r—
X
h-
II
h-ilu
CJ>
II
I—
2UJO
1—
«
II
UJ—
r— uI—"
U.
LU^UXUOU>- LU JL -^1—'—J
.j>Jl.
IM 01
II
*
UJ
II
»-*m •
•
OJCl.—•
1>J
<— *
>-• <
«1M *— _l o — ««MI-^ •- — f»_«
tQ.— 7Z
»-*^l
o
_j • *—
2.
UJ
2C
<
2.
I—
w>u
UJ
'
r-u-jt.
c/j<i«si
<_y
<l^u_
o.^j
t>J
»-iuc\-;u.r-ci.uju_uLCj^h- jsu-i-iz »-• *-*-" -«cy^i CJ !*• <i CT»
(jy
_i •**cxr-»-«u-«4 »-»UJV> Ji-«»J
cjt— a-—
oi— "-— 1_»
L_I^>*-«^>V_»
u<-•
CJ^»->>
U'
^
^: <C ci
CJ
cy
CO
•—'UJ
IV)
CX^>UJ h-uJiu
O
«'»
H
IM
,vi
\\t
CM
C)
WO
00
uuuu
00 UUUO
O
CJ
O
O
OCJ
111
.
•
*
z
*
. >
< •
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOQOOOOOOOO co^o^cvjiM^uA^r^cDc^o^cvjoiNrin^or^cou^Ot^cvioi^insu
^u^inu wa in ua Ln in Lnin^>u<>vQ>o>u>o^^u<>i^r^r^r^r»p~r^(—i^r—cuaj cocoes u a in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in
m m XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX a.aaaao.u.aauLaLLu.aaaaaaau.aaao.auLaaaa.ij.u.aLLaQ.aaa.u.aaaaLLaa xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx cucDmcocomcocuajaacocococcajaamcocB
mmuMnmmmmmmmmu>cn
^
m
LU
QC
z w
LU
Z u
CO
-»
—
—1
(-
^3 UJ
co UJ
at
•>
*-
J
•
-J
LU V-
X
a.
x
HI
z Hi O
LU
*L
1—
•-i
X
UJ
w JE Z O
••
I—
uj
:>
t—
a. 3.
Z) u.
•-
UJ
X
•
^c
J-
•
I-
UJ
ex
o ~7 Z •-••.•
UJ UJ _j
ll
O
LU
O
oc
^
•i
«l LU IX
hCO CX
LU r»
t-
m
a
>r
> z LU z UJ
•k
u JE
2:
^
Z
UJ
-JO
<I hm
a:
a
1—
CD
CO LU
uj CX
|
— cm m
••
fC3
z
^>
V)
a
UJ au
o
(X.
2: uj <l _» Q. uj
«»
»—
LU
_J oj
(X uj
w
uj t—
t-i
>C
-»
O •
•
~»
HI
«»
"5
Z *
0—
<ro
•
w U z
•
LU
••
^
-J
—
>r
»-H
in
•«
UJ
*
5t u_
i* CX t- •<_J
O
^
*— »uj
COCXZ
<[
—
+
> a co s
COO
>
z:
'
i-»CC—
>v
— cc o o • o 1— M iO
UJ
lu>> xxa:
«tl— h-
socio q£ujo
uj
-« ^» *: •
-»
»-•
xzz ->_>«-«
+
hU.
-S
It
II
ww
«—
—
2:
ujlulu
**
*IM *• U> ex cm ex -rt— cm »— q.
ZO Z •-»»— uj
il
z
it
z
11
x«a *o I— Q_Z - UJ—
»
II
q.o •• zj-
3£ZJ *v)
»-•
II II
->
v>
II
> + — — 2*2: Z3 *co
*"*
—
»-i
l_JLU
H-Z
z
•*
LL.
U-<lw
l-lf-H(X a: 1-0 LU
O Z X
LU t(X LU
3
«0 —
UJ h1—
a—
>DC£.
I— Q.
•^7
••
oc
—
Q.
•.
SLU —
ens.
3
»_•
CJ
r-
••-*ta CO~7_J co«—uj
oujc_> ocjcuj
?»«->
-•
>^o£ w
«o»-"
»mw
i.
m — z— —3 z o ^»-<
•-«»-«
II
UJ
v
•-*
II
II
ma JZ •—
QL(->
11
•1—
z w Z K
h-
3Z<1
>
_
X
•»— • LU 1—
uz
»z
1— CO CX
l_l
— O •
t-
O
--
»-•
v^ct
Mj^-i-iey^i'cj j. »»• t—
,,,
uuz-£l—
i_>u_lu^. _>
,
i'i—» i.
*•i—
.<£.-£
11
o »—
—
A. _j
CJ
OJ
Q.CIUV i-UV •
UJ
r-
JC
I
COhh cm-—
<_>»/;
Z£ w—
•-•»-•
Q.CJ
z
II
• ^-inia: u.
II
_, -J
i_>
j_h-—
UJ-5UJ
>—
4
Z
•
»—
t— UJ
uc
o>r
>^ uj»*
UJt_ji-»
cio
uco
«»
CJ<_'
-JO
o
«=«"-
>-•
^.~7
UJ ^.
w
»-i»-«
*-i
m
hj
r-uj
im
hum
,
en
i'i
.l.i-ivj'
>r
«r
O
UU
OOO
UU<->«-,t ,lJ^-
^ wwO
i_>u
1
112
•
»
1
—
»
1 <
1
»
« • 1
oooooooooooooooooooooooooooooooooooooooooooooooo
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx cocDcuaDGQcuajaja3CQCucoariaDcocccDCQCi:c^
CO
aaau.aaQ.aaaaaa.aa.aaaaaaaau.tLaaaaaaa.aQ.aaaQ.aaaaaaaa.aa
~+ in
HQC «l— <1C-J<1
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx aaa.aau.aaaaaaaaaaaaaaaaaau.aau.aaa aaaaaaacaaaa xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx tV OQ CD CD CD CD CD CD CO (D CD CD CD CO CO (D CD CO CO CD CO CD CD CD CD CO CD CO CD CD CD CD CD CD CD CD CD CD CD CO
CD CD CO