application of the mna design method to a non-linear ... · used the principle of the flyball...
Post on 26-Jun-2018
215 Views
Preview:
TRANSCRIPT
N O T I C E
THIS DOCUMENT HAS BEEN REPRODUCED FROM MICROFICHE. ALTHOUGH IT IS RECOGNIZED THAT
CERTAIN PORTIONS ARE ILLEGIBLE, IT IS BEING RELEASED IN THE INTEREST OF MAKING AVAILABLE AS MUCH
INFORMATION AS POSSIBLE
https://ntrs.nasa.gov/search.jsp?R=19810006486 2018-07-12T20:28:08+00:00Z
Uuclar;29bd9
qp
APPLICATION OF THE MNA VcSIGN METHODTO A NON-LINEAR TURBOFAN ENGINE
by
Or. Gary l*ininy*r
tSA -CK- 1ojd:) t)) APPLICATIUN uY THE MNA
DESIGN METt1UD 10 A NUNLISeAb TUUhUtAt+ tNuiNE
rLUal iie pott ( e'ucdut: Univ.) 1J7 p
IfL Alit,/t!r AU1 CSCL 11EUT
\\AA TI .^
CONTROLCERITER
SCHOOL OF MECHANICAL ENGINEERING
January 1981
eL^ v^
Aftwow^a ^^y7r^
4(44
1
i
II
11
^Lw' -11161- IjJU1
III
PUmDUE UNIVERSITY
WLST LAFAYETTE, INDIANA
47907
fi
r
r
i
F
t
M
1FORWARD
aThis document represents the final report to the
National Aeronautics and Space Administration for work
performed under Grant number NSG-3171. The NASA technical
monitor was Dr. Bruce Lehtinen, NASA, Lewis Research Center,
Cleveland, Ohio.
j
R
1
l
t xy
av i
pm
:. 1
1
i
r
TABLE OF CONTENTS Ij
Page
Chapter I - T^^troduction . . . . . . . . . 1 r
Chapter II -- M"Itivariable. Nyquist Array Method . . . . . 7
` A -Background and Notation . . . . . . . 8
i B - Perspectives. . . . . . . . . .1a
C - Design Procedure. . . . . . . . . . .22 r
D - Summary . . . . . . . . . . . . . . .33
Chapter III - MNA Applied to QCSEE Engine . . . . . . . .34
A - QCSEE Turbofan Engine'. . . . . . . . .35
B - Non-Linear QCSEE Simulationand Linear Models . . . . . . ..41
C - MNA 'Design . . . . . . . . . 46
References. . . . . . . . . . . .96
Appendix A - Linear operating Point Models . . . . . . .99a
1
H
t
1
r
Y $
Y?.3ror.sifliavb.xd;+e^.-.+aw
Wtm+.r+'+^.•.+"..—:.,xx.*-+rv^^^m"...;...«..^:^N.42'.2".^`a"bftmT.d«^>.. .. ....
..
._. '^--^—a...._:
casts . .. r_.r .« __ ., .,_.^ . ..r r,._.. __ rs,twnx_aa`st.-: die. rm tr . frr-x s, ,. r, ar. .x vn ... .,..,. .. _ ... ... .;
F--
I
1 CHAPTER x
INTRODUCTION
To meet the performance demands of future military
and commercial aircraft, propulsion system requirements
will make necessary the design and development of sub-l
stantally more complex engine configurations. These new
propulsion systems will dictate the need for advanced multi-
variable control system design methodologies capable of
' interrelating the design goals with thv: physical :Limitations
of the turbomachiner,y. This means that although steady
state operating line conditionswiTl retain their importance
in the design specifications further emphasis will be placed
on the short term transient performance of the engine. Toi
satisfy these performance requirements, more control vari-
ables will be introduced and more output variables will be }
monitored to evaluate engine performance.
in the early years of jet aircraft propulsion, single-
' spool turbojets were used almost exclusively.. The first
control systems for these engines wre hydromechanical and
used the principle of the flyball governor extensively
E
.n
r
t:
1
^a
r
-2
for fuel-rotor speed control [1). As tubojet engine
performance demands increased, a second compressor
turbine section was added to drive a fan to remove energy
i from the primary jet airstream, thus reducing engine air-
flow velocity and increasing cycle efficiency. The addition
of this second compressor-turbine section enabled greater
flexibility of compressor performance at high discharge
to intake pressure ratios, but required the control of two
mechanically independent turboshafts. Turbofans now
power rrr.7st modern subsonic aircraft and, with afterburners,
many high performance supersonic military aircrafts as well.
The turbofan engines in use today are basically fixed-Y
geometry designs. The control system modulate: one or two ='
variables and is configured as a set of single loop controllers
which are scheduled in an open loop manner according to
K aerodynamic and environmental conditions. The majority of
` the control systems have been designed using single-input
single-output (SZSO) classical frequency response techniques.
The resulting control configurations have typically been
implemented using hydromechanical or electronic analog
components.
As aircraft turbine engines become more complex,
additional requirements for the control system will result.
operational requirements will become more stringent thus
x ^;j'
y
-3-
necessitating increased control system capability in terms
of computational accuracy and static and dynamic response.
As performance requirements increase, more variables must
be controlled and more engine parameters must be monitored.
SISO oontroi system design methods will no longer be
appropriate as cross-coupled system interactions must be
taken into consideration. Open-loop, trajectory-dependent
control designs will yield to closed loop design procedures
as the now-standard, analog/hydromechanical devices yield
to full authority digital electronic control packages.
Some research and development activity for turbofan
engine control systems using modern control theory has been
reported by Michael and Farrar [2,3], Stone and co-workers
[4]', Weinberg (51, and Merrill and Leininger (6]. The most
comprehensive application of modern control theory to air-
craft propulsion system design was performed by Systems Control
Inc. on the Pratt and Whitney Fl00-PW-100 turbofan engine [7-i
111. This LQR control, design was successfully tested in an
altitude chamber at NASA-LeRC by Lehtinen and co-workers
[12,13]. The results of this test demonstrate the effectiveness`
of modern control theory for highly non-linear "sophi-
sticated turbofan application.,; in addition to providing
an evaluation of digital electronic devices for turbofan
engine controlsystems.
-a
3
ter-
l;
N3µ
in 1977, the r-100 sea level static condition was used
Ias the theme problem to explore alternatives for linear
multivariablo control in a symposium sponsored by the National
Engineering Consortium, The results reported in [14) demon-
strate the successfil. "pplicatlon of multivariable frequency
domain control system design methods to the F'1Qb design
problem • i, 1I 1 d
I During the period 1974-1979, the General Electric 4
Corporation, under contract to NASA, designed, developed,
and tested an experimental commercial aircraft engine [15,16).
This now Quiet Clean Shorthaul Experimental Engine (QCSEE)A
incorporates performance and structural characteristics
unlike those in any turbofan engine in production today.
' The QCSEE program introduced high bypass ratios for turbo-
machinery noise suppression, reversible pitch fan blades for
rapid thrust response, total digital electronic engine controls,
and extensive use of composites for drag reduction and weight
considerations. in addition to fuel flow and fan blade pitchG
angle, variable nozzle area 'control is also provided. To
incorporate all of -these characteristics into a single
propulsion system, represents a significant breakthrough
in turbofan engine technology.
During the period of development and testing of the
QCSEE engine, NASA developed a highly non-linear, accurate,
A
i
^E
p
S.
real-time digital simulation of the engine at sea-level
static conditions. This non-linear model was used in ex-
tensive tests at the NASA--Ames in-flight simulator facility r
for test pilot evaluations of integrated engine airframe
combinations [17]
Using the non-- linea r, digital simulation as a repro-r
sentative model of the dynamic operation of the QCSBB
turbofan engine, a feedback control system is designed in
I this report by multivariable frequency domain design tech-
niques. Transfer function matrices are generated at each
a of five power lever settings covering the ranee of operation
3 from approach power to full throttle, i. e., 62.5% to 100
full power. These transfer functions are then used by an
interactive control system design synthesis program to
provide a closed loop feedback control using the mu ltivar iable
Nyquist array (MNA) method [18-201 and the recent extensionsi
to multivariable Bode diagrams (MBD) and Nichols Charts
(MNC) [21-22]
The QCSM design is initially performed holding nozzle
area full open with fixed fan blade pitch angle at the power
setting of 62.5% of full power. This SISO design is evalu-
ated at other power settings and tested in the non-linear
simulation to evaluate engine performance during a powern
slam from 62.5% to full power (100%))
)
I
{The second phase of the MNA design uses fuel flow and
fan blade pitch angle in realizing a two input two output
feedback design, These results are extended to the three
input three output system with nozzle area as the thirdu $
az
input. This variable is used top provide additional control
over inlet-Mach number for noise suppression. In each case,
j the MNA design is compared with the Standard QCSEE control,
F The next chapter details the basic concepts associated
with the Multivariable Nyquist Array Method and introduces
several recent extensions to the theory. Chapter 3 then
uses this design philosophy to determine the closed loop
feedback control system for the QCSBU turbofan engine. The
successful application of the MNA method clearly demonstrates
i the effectiveness and utility of the multivariable classi-
cal frequency design procedure.
r
b
j
C
^,gz
II
F 3
ac_9
ia
^
:
2
e
x
—7—
s
CHAPTER 1ICHAPTERr ,
MULTIVARIABLE NYQUIST ARRAY METHOD s
t The Multivariable Nyqui.st Array (MNA) method is com-
posed of two Nycjuist array design rrocedures one pro-
cedure for the direct polar. plane (ANA) and one for the
inverse polar plane '(INA) Roth methods have identical
design objectives and are founded upon a common mathematical
structure. The principal point of departure is the use and
` interpretation of the multivariable Nyquist stability
criterion in achieving the final system design.s
The INA method was first introduced in 195$ [231 based
upon the premise that compensator selection to achieve the
desired diagonal dominance condition was easier to obtain in
the inverse polar plane. In 1974, a more comprehensive j
description of the INA method became available [,291 which
also provided the forum for the introduction of the ANA method.
Until recently these design methods were considered to be
mutually exclusive in the sense that once the open loop_
design conditionswere met in a specific polar plane the
closed loop design analysis was restricted to that polar plane.
Thus an INA design could not at that time be projected into
^r
i
f
_
t
a
x
- 8-
the more favored direct polar plane. This lod Fisher at al
[251 in a comprehensive application study to indicate a
preference for the DNA method over the INA ;.iethod whereas
the INA method is used most frequently in Burope.
in this chapter the MNA design objective is briefly
reviewed. New theoretical concepts are presented which
eliminate the polar plane dependency and establish a found-
4tion for finite frequency range designs. A new graphical
design procedure is then. introduced which utilizes the
fundamental ideas of Bode and Nichols in multi-loop systems.
I i
A. MNA Background and Rotation
For the system in Figuva la, O(s) represents an mxm
transfer matrix relating the output vector y(s) to the system
input vector U(s). The vector y,, C (s) denotes the reference
output vector. The design objective is to determine the input
vector u(s) such that the system output vector follows the
reference vector with dynam:wcs acceptable to the system
designer. if, for example, y ref(s)
represents a desired set
point operation then y(t) should achieve the desired level
within prespecified rise time, settling time, and percent
overshoot requirements.
ii
77^
5I
To the Oystom of Figure la, three square compensators
are used to close the feedback loop as in Figure lb. For
reasons which will be apparent later in the development
L and F(s) are assumed to be diagonal and K is assumed to be
an arbitrary constant matrix. All matrices Are assumed to
be of rank m. Shifting the constant diagonal matrix L to
the left side of the summing junction Figure lc is obtained.
Re-arranging Figure lc and letting
zref "" L yref(1)
z ( t) = L y (t)
the standard MNA block diagram of Figure 1d is obtained.
Let r(s) equal the identity matrix and define
Q(s) = L G (s)K (2)
Following Rosenbrock (241, K and L must be selected such
that Q(s) and/or
Q(s) K G(s)L (3)
A
is diagonal, dominant by rows or columns for all s on the
standard Nyquist D contour.
A v.c
i
d3C
+ e^ c)
.mac G ( s> e^ t
,FIGURE ,IA-W
OPEN Loo p SYSTEM
p K G (s
FIGURE IB; CLOSED LOOP CONFIGURATION I
a
L
f UUd
F[E csa
a
FIGURE 1C CLOSED Loo p CONFIGURATION
F (s, G cs^ 3 fit) 3,
FIGURE ID; CLOSED LOOP CONF'I'GURATION 3
-11
The concept of diagonal dominant matrices is well
documented (18,23,241 and is defined as follows for the
DNA methods
Definition: Let Q(s) be a complex square matrixof dimension m. Then Q(s) isdiagonal dominant if either or bothof the following conditions exist
f for all seDi
mi e Max Z lgi}(s)l/lgii(s)l < 1.0 (4)` seD j=1
j#i
mMax
6 91
Y seD ^l Iqji(s)I/jgii(s)j < 1.0 (5)
j=
where (4) utilizes the rows of5Q (s) (row dominance) and (5)uses the columns of Q(s) (columndominance)
Clearly if Q(s) in (2) (or Q (s) in (3)) is purelyP '
diagonal, then m feedback controllers (and m loop gains)
tcan be designed independently for the uncoupled system.
To achieve this degree of independence, K and L would of
^. necessity be complicated functions of s and could resultr_
in severe implementation problems. Rosenbrock questioned
the need for a purely diagonal open loop transflr matrix - Q(s)j
x
i ,i,r. "" ^ •. xaaz-^ttw•+wvwt+atcwVUaw.mne+ rrm.—wv+w+J' rca.wtem amw'+.aveaw° .:^'^vo-ax`.:+J?356a_."n'-r—•..,. A.l.{ : .: lW6^G#^A...—:•-ea.°la.ml..M......sty.3.ow.rrL.r-...A.....«^-._».-.-..v _2.zs.^..„e..e.!r,Y
,,'!
..12-
i
and developed a procedure wherein the off-diagoial entries 11
in Q(s) were allowed to be non -zero but of sufficiently }I
small moduli to permit design synthesis by single loop9
methods. By relaxing the diagonality condition of Q(s); v ^
to that of a diagonal dominant Q(s), the compensator matrices#
K and L in most cases remain constant thus significantly;
improving the implementation requirements.
Using the above definition for diagonal dominance,g
Roscabrock was able to extend the classical SISO NyquistY
Criterion to the multivariable case. This development led
to a graphical interpretation of the multi-input multi-outputT
(MIMO) Nyquist design objective where the traditional }
Nyquist curve is replaced by a band of Circles. The radiig
i of the circles (called the Gershgorin circles) consist
of the sum of the moduli of the off -diagonal terms in Q(s)
by rows or by columns. For the DNA method the Gershgorin
radii for ( 4) and ( 5) become
i mdi (s) =1 qij (s) (6)
7=Jai
m'di (s)Z I gj( s ) I ()
7=lj#
s.,. ,
^_.
-13
Hence, the dominance levels can be written as
e^ Max d^(s) /J q, , (s) l11
i
seD1
8y - Max dy(s)/Igii(s)
seD
To determine the elements of K and L, an optimization
procedure was developed by Leininger [18,19] based upon5
the dominance levels in (8) and (9),, rundamenta.ly, theE
optimization objective is to secure the lowest level o f
dominance in the respective rows (or columns) using a
constant compensator structure. This method of achieving
dominance can be used in conjunction with any of the twelve
MNA forms and thus significantly_ reduces the time required
to obtain dominance from several weeks (or months) using
previous trial and error procedures [24] to several CPU
seconds.'
Once the compensator pairs have been selected in
accordance with the above concepts, the respective control
loops can then be designed independently using classical.'
SISO design procedures. The resulting closed loop design
configuration can be further reviewed on an individual
loop basis through the utilization of Ostrowski's theorem [26]
..-«rv.,Fam}^y„^ -... ,_ :. ..—`_-•-----_• .. ..^^...•:.ny.4=^'..y:x^an _-rn*-s^^'+,.._... .. ...<._ ....+.w.a..` ^SmtK:.iIR dfa.3B$rffi:rs`-a:'**#..*..t::»°wLdwrxda.W?•.4....a-v+.cr.+»...t•an f ....: - __..,.•» _ ..«-,...,.
(s)
(9)
-14-
r3
This theorem provides for the reduction of the Gershgorin
radii in loop i based upon the feedback configurations
in loops jai. Thus a more acourate assessment of the respective3
phase margins and gain margins oin be made.
it is useful at this point in the development to di-
gress from the formalities of the MNA design procedure to
closely examine the closed loop structure that has been
created above. This analysis will place the MNA designr
procedure in proper perspective and provide the foundation 1
for subsequent designs using multivariable Bode and Nichols
methods.
B. MNA Perspectives
In most applications of the MNA design method Q(s)
(or Q(s,)) cannot be made diagonal dominant when L is the 3 f
identity matrix and K is constrained to be a constant
matrix. The alternatives are therefore to either insert
a constant postcompensator matrix or allow K to be a dy-
namic function of s To design a K(s) for open loopj
dominance over a large frequency range would be an ex-
tremely cumbersome and tedious task. It is also questionable A
whether the end result (i.e, dominance of Q(s)) could
justify the means by which it was achieved. That is, if y
^^^'^w.-a• '°. ~ _ .a• Y .o ,.: q:. .zw.G. .s ..x 2x s..° sa ti - s
t
-15-
w
dynamics are to be introduced into the feedforward path,
it should be to the benefit of the closed loop design
rather than achieving a particular open loop configuration.
The problem of inserting loop dynamics to meet design
specifications will be addressed in the next subsection;
i Consider the utilization of a postcompensator matrix
L as indicated in Figure 1-C. Note that the shift of matrix
L to the left side of the summing junction completely
changes the characteristics of the design problem unless
L is constrained to be diagonal. For a non-diagonal L
G
the design problem becomes one in which a linear combination
of the system outputs, L y(t), are expected to track a lineard
combination of the reference values, L yref. The error j
vector for this case is thus a linear combination of the
output errors,di
ey(t) L(yref y('t)) (10)
and may become zero for conditions other than
y(t)y (l )ref
However, when L is diagonal the error vector in (10) becomes (
s zero if and only if (11) is satisfied. We therefore, im
pose this constraint on L and utilize the concept of dominance it
a
4
sharing [19,201 to achieve the desired dominance condition -
for 4(s) (or Q(s).)0.
Ai
J 3
Y
a
f
i
-16-
in Figure ld, assume r(s) is the identity matrix,
L is diagonal, and K and L are constant matrices which
have been determined such that Q (s) in (2) is column
diagonal dominant. A common misconception associated with
MNA methods concerns an alleged modification of the rela
tionship between y(t) and u(t). Clear,l^`, this relationship
is always dictated,by G(s) and remains intact regardless
of the selection of K and L. What is dependent upon K and
L, however, is the relationship between the corresponding
entries in z (t) and e z (t)
1
t
ez (t) Zrefz i (t) (13)
x
with cross-coupled error signal (i#j) interactions determined
by the degree of dominance achieved and respective loop gains
(inherent within the elements of Q(s)).
The i th element in (12) can he written as
^m
z,
z. (s) q.. (s) ` a (s) + qj (s) a (s) (14)11
j#1
...Ej{
A
at.lira... .:..e ... _ .n7_.. .. n., .., ^...._d-#^, ' z.. .... [ ^x.^o. .4, _Y,.z:e•.yn.P_.. ri.l. T.>».,e.^ LL,...a _...,__ ...... ,_. ..
z (s) Q(s) ez (5) (12)
When Q(s) is diagonal dominant, z i (s) is primarily dependent
upon the i th error signa l
}
where i 1 1 2 1 ..., m. Note that although we assumed
Q (s) column dominant, it is the row elements of Q (s) which
contribute to the interaction terms in (14). To examine
this anomaly further, let m 2 and postmultiply (2) by
r as in figure ld
C
P''
I Q(s)r. L G (s) K F (15)
r ^
i1
where F its a, diagonal constant gain matrix, since Q(s)
is assumed column dominant, the dominance levels of Q(s)
are unaffected by the moduli of the elements of F. Thus
(12) becomes a4
z (s) gll (s) fl e z l (s) + q12 (s) f2 ez2 (s) (16)
i z2 (s) = q21 () £l ez (S) + q22 (S) f 2 ez (s) (17)i
1 2
r With J `iii (s) ( > J g j i (s) I for all seD it is clear that the
error term e z (s) has proportionally more influence on z(s)X
than on z i ( ), iy^j But note what happens as the gain f2
is increased with f l fixed. The interaction of ez (s) on2_
z (s) increases while the effect of e z (s) on z2 (s) decreases.l
This interaction influence of the gain matrix F is independent
of the respective dominance levels.
Thus low levels of dominance associated with each
diagonal entry in Q(s) are not sufficient to insure low inter-
-is-
action levels unless the impact of the respective loop
gains has been properly addressed. This analysis explains '+
why the DNA (or INA) bands may be very narrow (low dominance
levels) yet the closed loop time domain responses indicate
high levels of cross-coupled interactions.'
To .pursue this analysis further, consider the closed }'j j
loop system of Figure ld and assume that K and L have been jidetermined for column dominance of Q(s). Shift the diagonal {
L matrix around the loop as in Figure 2 recognizing than this
is the most probable implementation configuration since aour interest is in the y(t) vector and not necessarily z(t)
where i
i z (t) - ,L y(t) (18)
From Figure ld, the closed loop relationship isr
Z (s) = (l + LGKF) -1 LGKF Zref (s) (19)
For
Z (s) P (s) Zref (s) (20)'
with
P (s) (I + LGKF) -1 LGKF(21)
(I + Q(s)F)-1 Q(s)F
`k
f
Y
{{ ^
c
V
LA LLY ..
X
'ujF i
N
FEI O
J
C Wcn 1
(V,W
T)wOOLL
^-^,3x
1
/ 4
+
j
e
L....... a . b ^...,. n a .. .. ^ _ .... ,..., .^.u?. a ....»^.. x ^ ..a ^ ^........n....^, .i ..t ^ ^.^^" k.> . ...5 ,...^... r..+. w.y ,.. ,r ...... ... .,.. .. ._
i
y(s) (I + G(s) KFL) G (s) KFL xref (s) (22)
or
y(s) H(s) YreE (s) (23)
H(s) _ (I * Lr 1 Q(s) FL)-1
L-1 Q (s) FL (24)
Comparing (21) with (24)t
if follows that
f
H(s) -- L (s) L (25)
or
P (s)17^?1
p12 (s) . m P (s)lm
P (s)21 p22 (s`)
Z Pm l ( S)
. . . . . .p
(s). . mmm
{ 4
i
1
]171
-.. 3
per,
...21w
1
4
rt
i
Hence, the diagonal elements of the closed loop transfer
function matrix are unaffected by the compensator shift.
Further, if each control loop is designed for a small
steady state error condition (i.e., high loop gains),
then the of-diagonal terms in (26) tend to zero and y(t)
tracks yref' If the gain in loop i decreases, then the
degree of interaction in loop i from the remaining high
gain loops will increase. This interaction effect is not
easily detected from the Nyquist array diagrams as it is
independent of the dominance levels when using a DNA
column or an INA row dominant design objective
The concerns associated with closed loop interaction
levels and system dominance can be resolved simultaneously
when using a DNA row and/or an INA column dominanceobjective.
For these N.NA design configuration the m2 elements of K.
! m elements of L, and the m dynamic compensators of F(s)
must be simultaneous) selected. In this wa , system domi-Y Y Y
trance, loop gains, and response shaping for all loops are
addressed simultaneously since they are highly interconnected.
As a result of the compl,uxity of these design formats;
they have never been seriously considered as viable design
procedures.
in the following section recent extensions to the MNA'
1 design technique are briefly described,.
r
i
3
r-3
i
a
-22-
G. New Design procedures
The MNA method as originally conceived by Rosenbrock
was composed of two design procedures. Each design pro-
cedure was ,relegated to the specific complex plane in
which the diagonal dominance conditions was obtained.
Stability theorems based upon the extended Nyquist cri-
terion were then developed for each design procedure:
In a typical application of the MNA design methods,
Q (s) and Q (s) are not, in general, simultaneously diagonal
dominant in the Gershgorin sense for all s on the Nyqust
D contour. This is due to the fact that
qij (s) ^ qij (s) (27)
for i t j, and s on D. Hence it was thought that the MNA
forms were not compatible, i.e., DNA compensation methods
could not be used when INA dominance was obtained and visa
versa. This limitation of the ,MNA design philosophies
is !removed by introducing a conformally mapped image process.
The details, of the imaging process are contained in
[201 and will b4it 'oriefly reviewed,
Assume Q(s) is Gershgorin dominant in the inverse
polar plane Z where the origin is exterior to the respective
Gershgorin bands. LetF j
W 1/Z (28)
e23-
i
j
a
t I
represent the conformal mapping from the inverse polar jI
plane to the direct polar plane W. Thus under (28) there
4 exists a one-to-one correspondence between the paints
z(s) - x(s) +j y(s) in the Z plane and the points w(s)
IU(s) + J y(s) in the W plane. Since (28) is a conformal i
mapping circles in the Z plane map to circles in the W !
plane,
Let d(s) represent the radii of the circles in the Z
plane and r (s) represent the radii of the image circles
in the W plane. Then the image center point and image
radii are given bX
U(s) X(s) (z9)x (s) + y (s) - d (s)
V(s) Y(s) (30)r x2 (s) y, (s2^) — d ^
d( s)r (s) = - 2 ------^------ 2 (31)x (s) + Y (s) - d (s)
Since the origin is not an interior point of the Z plane
circles, the interior of these circles map to the interior
of the image circles. Furthermore, the relationship be
tween respective circles in the W plane will be maintained
xz.
-24-
I
since. (28) is conformal. Hence, any gain and/or phase
margins measured with respect to the Gershgorin circles I
(or envelope) in the Z plane will be retained in the imagei
plane.
This mapping procedure works equally as well when
Q(s) is diagonal dominant and the Gershgorin bands are
projected into the inverse plane. Note that the imaging
process is independent, of Gershgorin dominance levels in
the image mode as the imaging process retains the dominance
levels from the projected domain.
Stability theorems for the image bands as well asifurther insights into the imaging process 'are contained in
^P
Appendix A. The significance of this theoretical development
will become more apparent when the Bode and Nichols methods
are discussed for M1MO systems later in this section.
t
A second limitation of the MNA design methods as pro-
posed by Rosenbrock was the requirement of diagonal dominance
for Q(s) or Q(s) by rows or by columns over the entiref
Nyquist n contour. This condition was imposed to ensure that
the stability theorems of Rosenbrock would remain valid. For
if the system lost diagonal dominance, then the theorems
neither confirm nor deny stability.t
x
-25-
Since all zeal systems are band limited (i.e , strictly um
proper in Rosenbrock's terms), the search for dominance
only needs to be checked as w proceeds up the imaginary
axis on the D contour. Although this simplifies the search r
to the customary ranee of 0 < w < - it imposes an excessive`
burden on the system designer if he knows that frequencies
beyond some finite value, wo, are not likely to contribute
to the system response.
In response to this concern, a new theoretical develop- fment was introduced wherein the contribution to the MNA or
image bands for all frequencies beyond wo could be properly
accounted for. The specific mechanism for finite frequency
domain design evaluation and related stability theorems are
detailed in - [201a
Fundamentally, the contribution to the dynamic response
for frequencies beyond some wo are contained within a circle o
centered about: the origin in the complex plane. The radius
of this circle is determined by the moduli of the elements
in Q (s) or Q (s) at s = jwo . As wo increases the corresponding
circle radii de , rease for the DNA method and increase for the
INA method. The feedback gains are then selected in accordance
with the appropriate array bands and origin centered circles.
,_ y
".
..26-
The origin-centered circles associated with a finite
frequency range evaluation can be conformally mapped toN
the image plane. In this case the origin is contained within ►the circle, hence under the mapping principal the interior
of the circle in one domain maps to the exterior of the
circle in the image domain.
The two now theoretical developments discussed abover
provide the basis for the equivalence between the multi-
variable Nyquist array design procedures. As a result,
closed loop compensator design analysis and synthesis is
no longer restricted to the polar plane in which the desired
diagonal dominance condition was obtained. Thus, DNA
designs can be used for INA dominance conditions and the
desirable INA minor loop synthesis procedures can be used
when DNA dominance has been achieved. In addit$.oe, the
theory is now complete for finite frequency dominance.I
With this equivalence principal we are now in a position
to extend the design synthesis procedure to the magnitude/
frequency and phase/frequency planes. The resulting graphi-
cal display will allow for design analysis and compensator
{
synthesis using standard Bode techniques. Using the
magnitude/phase angle plane, a graphical display for each
control loop can be reviewed by the familiar Nichols chart
methods.
i
.-27-
I ^ 1
IX 1ViaK
Nx^
A
I
• EoI o ^•
p allww
: I ^
I
,
1^q'M
t uOO Q W W.................... ^„ • ZLYa:... .. ..: A 1r.^ _.7
'ter
Ox^ ••^^y
I ^^•:^ sst.NZ . K N it sar ss sts^ iR r Illiii• ^►K
MAW
o
Q-
.0
WGas
w^c
scC
J1 _
I'
MOCR
p 1! W
M
d ; I N^ a, : I _NQ'Coo
^--• ; i 1-nm 2 1 1-,^a a
Q. ^`^ r • ^^ ¢ j • SAS. a
n ,p
= I~
• w
_'Nil
i Ia
:'i^ I • Z S
roa _ vw
1
0
w. -
-28-
The fundamental concept of multivariable Bode Plots
and Nichols Charts to be presented below are described in
more detail in [21). Reference [221 describes a computer
aided design suite for closed loop system design and includes
all of the graphical. procedures referred to in this chapter.
The multivariable Bode diagrams are constructedidirectly from the data used to calculate a direct polar
f
plane Nyquist diagram. If the OLTF matrix is DNA dominant,
the envelope curves [27] of the corresponding Gershgorin
radii are each converted to Polar form and independently
drafted as a function of frequency. When system dominance
is obtained in the inverse polar plane, it is necessary
to first form the equivalent set of image bands in the r
direct polar plane. The envelope of the image bands are i
then used to construct the multivariable Bode diagrams.
Figure 3a represents an_INA diagram which is dominant
over the finite frequency range 0 < w < 3 0. Figure 3b
contains the corresponding image band for Figure 3a with
} Figures 3c and 3d displaying the Bode information.
s
Since the mapping between the respective complex polar
planes is conformal, it is not necessary to actually compute
and/or display the image circles. The envelope curves for
the image domain can be computed directly from the envelope,
curves in the domain in which dominance was achieved.
d
1
t
aw+w^+ww yrrac'<aew.w.n++^•++sa._ i ,..^.,-.-.-..w_. -.. .,.. .:.
Let c1(w) and c 2 (w) ,represent the envelope curves
for an MA diagram. The envelope curves, C 1 (w)and C
2 (w),
to the corresponding image circles in the image plane are
g iven by,
Cl ( w) = Lcl (w) ] -1 and C2 (w) = [c 2 (w) ] 1
(32)
Proof •
The relationships in (32) follow directly from the
conformality of the mapping W l/Z between the respective
polar planes,
Q.E.D.`
Thus for an INA system dominant condition, the Bode
diagrams can be calculated based upon the inverse of the
INA Gershgorin bands. The multivariable Nichols chart
is then obtained directly from the data used to construct
the Bode diagrams and is presented in Figure ,4 for the
conditions detailed in Figure 3.
With -the graphical analysis procedures described above,
system output response shaping can be considered by any of
the standard SISO methods. Since INA and DNA design methods
are described elsewhere, the remainder of this section will
be :..restricted to the Bode: design form.
f
f
a
-31_
Consider a DNA column dominant condition for Q(s)
E in (2), and assume the system is type o Since the
loop gain determined at w = 0 may be low, a reasonably
large steady state error could be expected. If the
design specification will accept an error no greater
than 10%, we may compute the required gain increase
using
te5S = 1 + f i P -ToT (33
where P i (0) is the value of the i th Gershgorin envelope
(or image DNA envelope) closest to the origin at w = 0.
usina this value in (33) ensures that the steady state
error will always be less than the required minimum level.
Thus
_ 1 - e
f i e SS) P i (0) i l l 2, ..., m (34)
ss
t Similar relationships can be obtained for type 1 or
higher systems.
Inserting (34) into the system will only shift the
magnitude curve of the Bode diagram. As the curve is )
shifted, the closed loop bandwidth of the system (-3 dby
point on magnitude curve) will be simultaneously affected.
1
It is also possible the system dominance may be lost
as a result of the gain change. This is apparent, forthose cases in which the 180 0 phase angle informationcorresponding to the new zero db crossover point is
contained within the phase envelope curves.
This condition, however, does not present anyserious complications since we can easily synthesize anappropriate compensator to shift the -180 0 point outside
the phase angle bands at the crossover frequency. This
can be accomplished using standard lead, lag, or lead-lag
methods. The details of the compensationprocedure
y ' are described in [21, 221. Similar procedures apply for
type n systems.i It is now apparent that the multivariable Bode diagrams
and Nichols chart can play an important role in the effective xdesign of multivariabl.e closed loop systems. In, addition, f
it provides an effective measure of the frequency range '.over which dominance-was obtained. If the desired closed
loop bandwidth is beyond the range in which dominance wasobtained, new efforts are required to compensate for this
9 fitideficiency. At the other end of the spectrum, if the zero
z db crossover frequency is well within the dynamic range,_
this could reflect too conservative of a frequency range
5
-33-
(i.e., frequency range too large) for dominance acquisi-
tion. New design efforts over a reduced range may provide
less interaction levels in the bandwidth regions of interest
and could improve the overall design significantly.
D. Summary.
In this chapter the MNA design procedure was brieflyi
reviewed and new design tactics were introduced. The Imagef
bands play an important role in establishing the equivalence
of the INA and DNA design methods and provide the theoreticali
foundation for viewing the INA Gershgorin envelope curves
in the Bode and Nichols domain. The finite frequency de-
velopment provides the theory needed to reduce the com-
putational burden for the diagonal dominance search.. The
multivariable Bode diagrams provide a new view of the system
design where SISV compensator design procedures can be
used effectively to secure dominance and simultaneously4
achieve the desired closed loop performance. The correspond-
ing multivariable Nichols chart displays the Bode information
in a familiar form which again is suitable for design
analysis and control system synthesis.
The next chapter utilizes the design procedures and
theoretical developments of this chapter to design a closed
loop control for the General Electric QCSEE`turbofan engine.
k
CHAPTER III
MNA APPLIED TO QCSEE ENGINE
The turbofan engine selected for the MNA control
design study was the quiet clean shorthaul experimental i
engine ( QCSEE) [16). The QCSEE engine program was a NASA
sponsored effort to design, build, and test two experimentalengines which would serve as test beds for advancsd
propulsion technology evaluation. one engine was designed 3
for under--the-wing (UTw) mounting while the second designwas developed for an over-the-wing (OTW) mounting. Thisstudy was limited to the UTW design.
iTurbofan engines of the QCSEE design are under con-
sideration to power shorthaul aircraft of the STOI con-r figuration_. These aircraft would operate from small,hir
ports located near major urban communities. In this wayi
large transports could land at metropolitan airports many
miles from the urban center with passengers and cargo
transported to the civic centers via low _noise and high
'k reliability STOL aircraft.fl
To provide these services, primary propulsion re-
quirements for the aircraft are those of low noise levels
k
-35-
and exhaust emissions at all operating levels, rapid
thrust response capability including reverse thrust and
low weight. The design challenges were met by the
General ;Electric Company.
In the next section the specific design for the QCSEE
UTW engine is detailed with principal emphasis placed on
the control system design requirements. The following
section describes the non-linear digital simulation program
developed by NASA from which the linear operating point
models for the MNA design are obtained. The last section
discusses the results of the control system design study
performed using the multivariable frequency domain design
procedure.
A. QCSEE Turbofan Engine
►
A cross-section of the under-the-wing QCSEE turbofan
engine design is detailed in Figure 5. The design in-ir corporates performance and structural characteristics
iunlike those in any engine in production today and includes;
i1. An extremely high by-pass ratio (12:1)
and a high throat Mach number inlet~ fornoise suppression.
2. Reversible pitch fan blades for rapid 1`thrust response.
a
j
"36—
}
LN
4
Oz
^a
4A LiJ
V i
VA
Y 5 `n
x
/oo
k t ^y nm m^ /
a ^ i
i r ^
V I
C^ t
tq !
1 z r u
m U-. G^
t
>v C7 ';
t
1
f © ^ aW
I,E
.-..z.^ ._. • u., _. .an a v. + .. ..N.. ^^c. siw^F.ww^F , .,., «- .w ...va.w. a.n .,,..wvw .,. >^,.v.-^a.m3 x.. ^ ^^..,,... .».. >~...+..« q,.i _ .. ^l .^
3. Geared turbine/fan combination for lowfan tip speeds with a high thrust rating.
4. Digital electronic engine controls.
S. Extensive use of composif;es for dragreduction and weight conziderati,ons.
The engine also incorporates a variable area fan duct
exhaust and variable compressor stator blades. The fan
I
is driven by the low pressure turbine through reduction
gears to achieve an efficient low pressure ratio, low
tip speed operation. The variable pitch mechanism is
capable of moving the blades from a forward thrust position
to a full, reverse thrust position in less than one second.
In the reverse thrust position, the hydraulically actuated
fan duct exhaust nozzle is opened to a flare position
to serve as an efficient inlet,.
The power requirements of a shorthaul aircraft system
dictated the design specifications which had to be met by
the QCSEE control, The primary specifications were:
1. to allow the pilot to command percentrated thrust,
2i to assure engine operation at all timeswithin safe engine limits,
3. to control inlet Mach number to meetnoise suppression levels,
4. to achieve rapid thrust response (inboth forward and reverse modes):
(
kLLe.3a ._ via >:^KSYNi+ air ^ wS? rY "ms n v+e saa,cuvUy,p f. xxx crcn.ie. a-+w.x«.a
-38-
These requirements all relate mainly to safety and noise
emission considerations. The first requirement resulted
from the desire to decrease pilot work load in a STOL
environment. This can be achieved by closing the feedback
loops using a parameter closely related to thrust (such
as engine pressure ratio) instead of using a parameter
indirectly related to thrust (such as spool speed). This
reduces the time the pilot must spend adjusting the
throttle to achieve the desired thrust level.
The second requirement meant that logic must be.
incorporated to prevent compressor and fan overspeqds
and turbine overtemperature. To meet noise suppression
specifications f a high inlet throat Mach number is required
to significantly reduce noise radiation from the inlet
duct. The requirement for rapid thrust response arises
from the need to insure safe STOL operation. in particular,
to have a safe go-around capability, thrust changes from
62% (approach power) to 95% of full power must be achieved
within one second. This is two to four times as fast as
is possible with current production engines.
The above control system requirements were met by the
engine manufacturer by using an engine mounted digital
control system shown in Figure 5. All primary control
-7 1 9-
functions are programmed on the digital control computer.
Digital command signals are sent directly to the three
input actuators:I
1. fuel flow actuator,ri
2. fan pitch actuator,
3. nozzle area actuator.
A hydromechanical control provides fuel flow limiting and
is also used to provide the command to the compressor
variable stator actuators.
An important consideration in the manufacturer's
design process was the control mode selection. This is
the process of pairing the four control variables with
four output variables. The mode selection was made such
that at key, selected steady-state operating conditions,
primary (unmeasurable) ;performance variables such as thrust, :p
specific fuel consumption, and stall margins are relatively
insensitive to small sensor inaccuracies. The result of3
the mode selection process for the QCSEE UTW engine control
was to pair fan speed with fan pitch angle, engine pressure
ratio (P /P ) with fuel flow and inlet Mach number withS3 TO Y
t fan duct nozzle area. The core stators were controlled
essentially independently, being scheduled (according to
standard practice) as a function of corrected core compressor
'speed.
t
R
-40
a
The steady state operation of the QCSEE engine is
determined by a set of demand schedules. These schedules
command optimum values of engine pressure ratio (EPR),
inlet Mach number and fan speed as functions of the pilot's
power request and .ambient conditions. In particular, the {
EPR schedule is a monotonic increasing function of the
power request. The inlet Mach number is scheduled to be
a constant 0.79; however, at low power (less than 70t of
maximum), the nozzle is full open and the Mach number then
varies with power request. The fan speed is scheduled1
as a constant value for power requests above 555. Below
that value, fan pitch is at its minimum allowable setting j
and hence fan speed varies with power, request. Keeping
,I
fan speed constant jeans that thrust can be rapidly mod-
ulated without the necessity of waiting for the rotor
speeds to change. This is especially important when
attempting to meet the stringent thrust response requirement r
of one second between a 62% to 95% power request change.c
,,
C
in addition to the steady _state schedules, a number
of limit schedules are incorporated; for engine protection.
.Fuel flow is limited by maximum fan speed, core speed, and .r
fan turbine inlet temperature schedules. Also fan pitch and F
nozzle area controls have maximum and minimum limit schedules.
And finally, closed loop proportional plus integral control
,_ _.. .....'. L..»^:a-x z^< e_„vsv __ ..zvm+ttn`a..+x sc -gym ry ^t.._
-41-
was implemented on EPRI fan speed and inlet static pressure
(inlet Mach number) to ,insure that these three outputs
follow their respective demand schedules.
The QCSEE control designed using the Multivariable
Nyquist Array (MNA) method (described in Chapter 2) makes
use of much of the control logic described above. Ini
particular, demand schedules and limit logic designed by
the engine manufacturer are used with the MNA design as
are the max and min limit logic. hfowever? the MNA design
replaces the three independently designed uncoupled servo
loops on EPR, fan speed, and Mach number with an MNA
designed pre-compensator matrix, which provides cross-
coupling between the command schedules for optimum overall a
control. performance.
In the next section the non-linear digital simulation
for the QCSEE engine is described. This simulation is then
used to obtain linear models for the MNA design method,e1
B. Non-Linear QCSEE Simulation and Linear Models
During the hardware test of the UTW engine at the NASA
Lewis Research Center, NASA developed a highly non-linear
accurate real-time digital simulation of the QCSEE engine
at sea-level static conditions. This non-linear model Was y
used in extensive tests at the NASA-Ames in-flight simulator l
facility for test pilot evaluations of integrated engine air-
frame STOL configurations (17)
a _ _ y __._.
-42-
The digital simulation of the QCSEE UTW engine, shown
in Figure 7, was based on both steady state, flow-pressure-
temperature relationships (maps) for each component (fan, 'Icompressor, ,combustor, etc.) and on dynamic relationships.
These relationships are based upon continuity considerations
for mass, momentum, and energy.
Maps of pressure and temperature ratio as functions
of corrected rotor speed and weight flow are used to model
fan tip and also high pressure compressor performance.
Compressor variable stators were assumed to be on schedule
at all times. Fan hub performance is a polynomial function
of fan tip performance. The inlet throat and duct are
modeled by static pressure loss relationships. The combustor
is represented by a heat addition and a pressure drop, with
combustor volume dynamics neglected. Both high and low
r
pressure turbines are represented by maps of flow and
enthalpy drop which are functions of pressure ratios across
i the turbines and corrected speeds. Compressor speed is
i
computed by forming a torque balance between torque generated
by the turbine and absorbed by the high pressure compressor.
The same procedure is followed to compute fan speed.4
The complete non-linear simulation is programmed in
Fortran and uses a modified Euler method for integration.
It runs much faster than real-time, having a problem time to
CPU time ratio of approximately 25 to one on a UNIVAC 1100-40.
-43-
The MNA control system design procedure requires
linear operating point models. These models can be
generated directly from the non-linear digital simulation
through the use of a perturbation technique. This procedure
(also used to obtain the F100 linear models) (28] producesmatrices A, B, C, and D for the standard state variable
form model:
x Ax + Bu (35)
y = Cx + Du (36)
where x is the system state vector of dimension eight, u
is a third order control vector, and y is an eighth order
output vector.
For the QCSEE UTW engine the physical variables in-
dicated in Table S were used to define the linear state .ymodels. Note that several of the variables in the y(t)
vector are not measurable. {
To obtain 'linear models, the following procedure was
followed;
1. Run the simulation to steady state
y operating point conditions.
2. Disconnect integrator inputs and 1perturb each state variable x j (t) in
turn (j = 1 ..., 8,) by a small amountAx(t) from its steady state level'.
f
yi
F
-44-
5
3. Note the corresponding change in thestate derivatives (Ax j (t)) and outputs
Ayj (t). Compute elements of A and C
using
Aij = Axi/Axe
Cij = Ayi/Axe
M
4. Compute steady state gain relationshipsbetween u and x i and yi by perturbing
uj in turn (j = 1, 2, 3) and allowing
the simulation to reach steady state.Note the steady state relationships as
Ax O = K Au(-)
Ay(-) = K Au(-)
5. Determine the B and D matrices as
B _ -AKX
1d
D _ K CKy X
Following this procedure linear models were obtained
for five power demand values, i.e., 62.5, 70, 80, 90, and r
100 percent power. The matrices obtained are indicated
in Appendix A
A
- J
' S
\^f
` -45-
F
t}n
t
TABLE
' VARIABLES QCSEE-UTW LINEAR MODELS
State Variables
P12 - Fan inlet duct total pressure
P13 - Fan discharge total pressure
P 4 - Compressor discharge pressure
P8- Core nozzle total pressure
NL - Fan speed
NH - Compressor speed
T3 - Compressor discharge temperature
T4 - Combustor discharge temperature
Input°Variables
xMV - Fuel metering valve position
X18 - Fan nozzle area actuator position
Theta I - Fan pitch mechanism drive motor positiont
_x
Output Variables
PS1 1 -
Engine inlet static pressure-I
-P13 - Fan discharge total pressure f
P4 - Combustor pressure
P 8 - Core nozzle total pressure
NL - Fan rotor speed
NH - Compressor speedi
T41C - Calculated turbine temperature
FN - Net thrusti
3
_tYu..«... \.... .. w a. ... ... ..l... .. .. .s.....n..._..YM.,.., r2EE'v i:tt _x s..,...., ..... ....t_. _.. ..wvG..^: .J.. .. r. ... . ......... ... ? ..^... ..._.. . ., .... < .r.
i
-46-
C. MNA Design
The MNA control systems designs to be discussed in this
section utilize the state equations of Appendix _A aug-
mented by one state variable for each integrated input variable.
This integral control action eliminates steady state offset
and conforms to the standard QCSEE control, form. Since all
control loops utilize integral control action, the entries
in the transfer function matrix, G(s), were modified appro-
priately. Thus, the resulting MNA designs could be mple-mented directly in the non-linear simulation for design
evaluation.
The First MNA design for the QCSEE-UTW engine assumes
'
nozzle area to be fixed at the full open position and fan
pitch angle fixed at the 62.5% of full power position. With
these two control inputs fixed at their respective values,
the only input variable to be determined then is fuel
metering valve position. For this SISO condition, combustor
exit pressure (P 4;) was selected as the output variable to
be controlled as it is most representative of cjystem thrust
response.
From the linear models of Appendix A, transfer functions
augmented with integral control action for P 4 /xmv at each
'F
operating point are obtained. For the 62.5% condition, the
Nyquist, inverse,Nyquist, and Bode diagrams are indicated
v
-4.7
in Figure 6 for a feedback gain of 0.04. The phase margin
is 29.90 degrees with an infinite gain margin. The closed
loop bandwidth at this design point is approximately
10. radians with the feedback control
Xmv,,,04 P4 (37)
To examine the effect of this control at each of the
fother four operating points (70%, 80%, 90%, and 100%) the
composite Nichols chart of Figure 7 may be used. Here
each ct.lrve represents the respective Nichols plot for each
P4/Xm transfer function with a gain of 0.04. The 62.5%w
power curve and the 70% power curve coincide. As the power
level increases, the corresponding curves move to the right.
This signifies an improvement in phase margin (to 50 0 at
100% power) with a corresponding decrease in loop bandwidth.
The transient responses of the nonlinear simulation to
the SISO design discussed above for a power change from 62%
to 70% are indicated in Figures 8A-8D. For referenceI
purposes the transient responses of the simulation using the full
GE QCSEE control are also indicated in the figures.
Over the power range 62.5-80% power, the full GE control
consists of a SISO Xmv control and sufficient control logic1
(i.e., constraints) to impede movement of X1 8 (nozzle area)
-gg.. •
N
I ^ v ( ^ y
s s
n ~ E': CL
I ` = aN ^ ^ o ^^^ 3
".. r..
_ I tIUWWI. •.......................
. •.• .6HU...• ."ma, x n .................................................. I ^^^Auuurrr•y ^^.w^^ • Alr l^. `N ..^T^.^^^^T .^rf • :Ably.
f►.rM W
N + LU^ p j
L w a
s I Zh N
^ I ^ _:►•^•
COn I n! a
zLJJ :' • ii w i i i• 3
i NW.
j
s ^ ^^ _ s s
a
a
_q9--
NICHO.S CHART
3•r 0! 0
P
G
II
to
N'
D12
E B-iD
IBE -^LS
-f•
-_+
SF4+70 ••"•••_.....AF449i ..«...
^'N R r9swG6FT1144 s. +.-•
pi-A
zel000,
-2+o -2!• --180 >150 -120 -90 -i* -30 0
PHASE ANGLE IN DEGREES
FIGURE 7: NICHOLS CHART FOR SISO DESIGN AS AFUNCTION OF POWER SETTING
i
i
"so-
I ^,v
#
wCA {W ^N Nw L= •
'^ oQ N
`^Rl
1
^
L#.
Ad 0
Ln
# ~ ^. •,..I. #•.. R•....• .......... ......1•... `. iN
LLJ
•^ • M rw11^
IW
• '^= 77
I WN
M ».0-i Ic
i ►^ (0-9 ^ t
I z 0 t) a ..
I I I ^ I' _^ ^ W
9
IL s„^
n.. •s.r• ,.# .....•. ......................... • „~,. `: ........................................ • w.
I
yi
0 Qr`"•
S I WMWCOI
0w
a w ,Iz c^ ( w
I—
A t .t /....i t'!t kitt.11.. etf...}. t.ttr.r.!!ttlt. Y ^ 1 H•....^^..^..^.^.t.l.,.....r. t..Itl A
IN4
y^ZiO I
A EN
c
_(A : 1 1►. I I =ow
N SIL
c U) ;
tZ 0
iit
^ • ......... •.....• .............•............. t M1 1 •
fff.IJ w
sZW1WWl.^1Nw`
z All
^O
M ! rm r
iyo
0 W `sM 4Q W ('
a !I
w
1__ S
t
x
ti
{ it Ii
1- J I •,
H
=
'c W ^ L
$ W;^^ •
^i SN S I •+Q
_ O =I _s
r r ^` i I O tacNv ...Y.u....•1..lf.••,. .......................
...................... . ...
..•.. rI• y
K
►.n
WW
cW
JJl
I
Z
0..CIO
co
W
to
LL.
.51—
.. Y air v • W"~ M• O
N ry M O
i
-52—
! s
A :• ....... Y••.#.t
............... ••• • ^. ... r.... ♦ ......................•i!f[}•.••Y.•..1• r0 A
w
Iii
iiJ C
N w 'i5
Z U ta
^N_ Q
«Z.r
to
a Ln
wrw. L Z•' ►r J uj
.F RNR#ftR#.l!# .. ..####t ...............
Ii
Cb
uj
We,
M
r
• W.^,yy
{yyy^^^ ^ a 'r /-9`
r c M,^ 1c, : ^. Gam. ••
CA Icou
0w
;
;!_
W
s, L: wW w
0W N
a en , a in^! IL
= v I E o I
I
^^T
MIL
f
i
s11
-53
^_ s 0!'
r
wife
O F t^
2 CisH Q
oil
W • HN^( lie
4n w CA w
W (_ a N Z "
cvI"I
b' co {^" dt 'f
I i I : H Y
Cdr
uj
I .._7 -7 • tr : N.tlt/r .c rr rt: • ..r • .a. .. .^..•r ...••.a. r..•..rr.
_•.•cnU
IL 10x rZ LLI
•_ ._ Li
HI ^ ^ ^..^ ; + 'QNo. Noy
NJ
. r _ r ^ gin az N ( = N NN W
ttI ^ N W
V I=F = V ;+ r. 00'•
jZ
I I^ ^I 3 I!;L.
Nc=.7
;f 4 :I _^
LR Wb
~
i t S W"..^.... ^ I p S• 1 J
;............................................... . .4 ...a ....... ....... ....... ..N
N
it
w .. •- • N A N ^ • F
-59-
and Theta I (fan pitch angle). When the environmental
conditions near 80% power are realized, the.GE control 3
3
logic permits Theta x to change, at or near 90% power
conditions X18 is permitted to vary. It is not surprising', {
therefore, that the steady state conditions for the SISO
MNA and full QCSHE control responses are identical over
62.5-70% full power,
what is surprising, however, is the improved responses
for Xmv , thrust (FN), and fan speed (NL) of Figure 8G. Since
total fuel consumption is the area under the X mv curve,
it appears that the MNA control of (37) is more fuel
efficient at this operating condition.
To examine the impact of the constant gain control of
(37`), in light of the Nichols chart of Figure 7, a non-linear
simulation run was made a power slam situation, i.e.,
r 62.5% to 100% full power. Here the engine command setting
of 100% full power was requested from the 62.5% operating
I
point condition. The corresponding transient responses for
the MNA-SISO and full GE controls are indicated in Figures
T 9A- 9D.-
From these responses it can be seen that fan speed (NL)
and inlet duct pressure (P12 ) are significantly higher for
the MNA-SISO design than for the full GE QCSEE control.I
`However, engine thrust (FN) and total fuel consumption (area
t
1:^ ..s_,. .... __. _..<.__ _, ._ ,:.: r..x," T,•n,..^_^.»,,.n ..;-ry .v. n_s:^cr..yr".::et. .x. a,..... ._.—. _» ..,_ __. ,.
-55-
WN j N_IWsr^ 0 c , r ,4
Z SI I cc I "
M I I^.^ H I f^.
H W
"^ W. N
I ~_1
_ W W _ •
O W j ^'^' W = y`
a M a N
=O I # _°
r ;L N /^ =
1' ~ I
NLO
cn
I
I
J
1
4 ., ........ ..............•...............• ^. ,. „ ............................................ I M,
tl lH...• « 0 • ti W
4L
I
I (^ I
1
01)
a^ Ipo a^ I "
MW (^ . NW {(_ so Wcr
0
_ tO =O i I t1..
CEI
I '
. .
s^ -^.aM
...... w....w.......1.. +.. M . ..................................... .....Igo
t t ^C _ s t !Iwo
kr
•I
iWfu
e I..Q.
Z Hi
!-^ J =;r^ IX
AlQ=
i= H
H Ni1 M
^^N ^Z VIC v
'i
0
^I=
s;i d^ J
wIV
N ^.
^iN
H Q
J ?)E W ; Mr~
N W/.
Z 0 ,fit_0 ;i"' a1 ^i^ ^
J _f,rL=
_:
M
N
i
.^, c........................................... _ ►-^ M
N
Z M
^ H 4w
N W i ~ 13
O^
Ic
LJ............ ......... ... ^._ w
vM ^ • 1 t ^ ^ lJ, i
i
anhLL
v!= i i
CD_ ; Q
CIOt NN W _ I'
W ( .,
rnc o1 ^i ^
11Ja
=i
i ; W
z
(^.................................
NI xN
rM
4A WC 9A
^ = o I ^ e^I a
^•` "'M^ ,'' fCJ1
A . .. emn. ..• ....•/ ................. ............ f .. ^' CL
LU
•7
T ^ • ^l
V
o~ 4I=
a^^ fN »
W ' Ncvf
Ir '
!jj M V 1 N •
^I •i I
v
^I
-57-
a , A 1^'1 a . 1'i
a^ ( a^ f »0.1^
c . ( 1f• •^ s c'•.. » . E to = _ 0
N W ^L S
H W =
aH
a cn = ( . a to = ^"= o f ' c v ,
^ ^= f ^ M I ' i f ^ N
=r = I
e^ = I ^. .JJ
., ................. .......................... ^ri „~„ .► . ... .......•.....o. o ... ......... f. «.. a
N N .r w ^ ! ^ V ♦ M •g
3
r
0
F •a.1
iN W
= 0
-58- m
Q O
J EIcW MX0w ,I=
L CJ i t^ O ;
I
i
L. w
;I
i
fit
WN( x^r r
r zi
M_
1
N
Un
U-1
f^= w. ............................................. • = c.^..lei
.. • u..r
E Ii= ^ I e= ;Q
t
.
\^ 1 M 3c M
c~Aw ~= w
0wLs
;(~ Z^N S
`'"J la_.9
'"\
= 0 S I
^^..^ S W
•
(^ O i ( ^
w^
..
Irk
rte'" I ;9
............................. . .
J,_.
^,... ... ..•.. ... .. ....... .....•. ..... •^ wQ.V..r
.. _._
A
•
rt ^. .. -. ^ -+u^imw-wires..+
-59-
aa
under XmV curve) for the MNA SISO could be consideredt
an improvement over the corresponding full GE control'
responses. it appears, th%refore, that both X 18 and
Theta I (fan pitch angle are used for noise suppressionI
in the full GE control and not (as stated in (15)) j
"primarily to provide rapid thrust response." Except i
for the obvious reduction in NL the impact of Theta I
of Figure 9C is unclear.
For a more meaningful comparison, the GE controlsi
on nozzle area (X1 8 ) and fan pitch angle (Theta I) were
held fixed at the 62.5% full power position. This condition
is then equivalent to the MNA STSO design and,is presepted
in figures 1OA-10D for a transient power slam to 100
power. The two designs are comparable in transient and
steady state performance. The control logic to implement
_(37) is considerably simpler than the logic in [15, 161.
As indicated in Figures 8A-8D, good system performance
i
is obtained at the 62.5% power point by a SISO control design.4
The design point at which two controls are simultaneously used
by the GE QCSEE control occurs at the 80'% power point. Using,
the linearized model corresponding to 80t power, a two input
two output condition was configured for an MNA design
attempt. The inputs selected were xm v and Theta I with
_ J
iI N
^I^^-z w
H Q i h Nr=
W M I ..
L................ •
r : NM
! XN
z° i =i of- r"_ 64 _ i
4A w L w
z 0 I ^'c o I
;
0-L
' Il` M
$w._^_,.T.-
... ........... •
.....^.._r_I-J
qso-
d
t0
O !~- r=
61E
I Igo
0 zvt0
a
I` I
u
Lf1
I
VW
C_t.
' I I
W
................. .................. to
Cn
I
t
•
iwow ^=
_:I ol_
i14 Q_ ^^
.y
c We oC _~toM
=1 = w
1
~Q H= O
i s
^i a.► .........w ............... ...................i
N• ^ i T w ^ w w • ! • •
r^
r
-61-
Wes
^^Ij
Q O ^1Z NO !" ( M I r+
s I ~_ o.W I ^^pq
Yf WW 11 ^I
; {I ^
I ^!L^ ^ ^ `', ^ 1.11fI
^ `_ ; It l LJJ g
................................... ..................... N A.
, p1
. I0 1- o F- ,
f d' 0 _
w= o c aI a
^_
,'_(t
^I J 1
,,
'^-........................
........... ... ^^ w ,, ... ..... .... .......»»....... ' a .,
7 ^• • •
w .:-A r N Ali' 3
•.62..
I ^^ o^t
M
!'r. M
NA^ I ^^
Np
^.r-•
WIZ
~J Ic { I N
1{~A W 1 d t V
aco I z^^ g o
_
` 2 t00
II CW
♦♦I^
• ..}•.................... .. }r ry Vi.a ........ .. }... +
J • s Cs+
I a4
_ ._ « I •
H¢ ► (_ H Q s!! C") a
W LlN W
N w ^^s( zy I
LS
L=
io
....1- ►-•1.1 ..
................ .•... . ................
s
^111 ^ ^ M ^ r M ^ 3
3
j
id
'ALlif7 xv. +v _ s. ^^ .r _ r z m. r _.z*i . s:.w3i?`.^ P.er r ._ ,.. -.,.., .. _
63.
^N
.M,r
Z
t=^a rzH J
Era.) )N►_ N ~
_W IliI a
w
1 i,
M J •' Z
y^ M M • 111
ION
r^ _ifti 3c ;^
C V1M i tW w
;^~z VIc v ,
gi)
...................... ..•.. .. ^.a
•a
1` 11z aH d t
I Mej
1CII ^^
H WNN
(1r
C)
W
» wA Iy •i ^ C7
Cn
V){^ I ui N 1
z ° =jt
_F J IxJ_ ,^ to«t y>L i ^ ^_A Wz 0 ^~ L1J
IC
A wI(z ^ ^
f ^+
: Y^YYY...^•^^.•Y.•Y. • •.YYYY•.YYYYY•Y •Y1•Y^ ^. -
Ww
,
^w
-64-
outputs P and NLf4
NL Gl,(s) G (S) X12 Vmv
P G (S) Theta I4 21(s) G2200
liere nozzle area (X was fixed at its full open18 position
(62.5% power)
Application of the MNA design suite described in [223
to the corresponding transfer function matrix at 80% power
yielded system dominance over the range 0.1 to 50 radians.
The resulting constant compensators are
1.1 80163.5K (38)723361.1 -2.1263(10
0.5163. 0.0L (39)
0.0 13.8743
3.0(10 5 0.0F
-80.0 i.i(10(40)
The DNA f INA, Bode, and Nichols plots appears as Figures
11-14. Transient responses to a power step from 80% to 90%
•d ►.. r•.••. ►.^^* ► . ►►►►s►►►►^►•••• ►►w^►^ ►►►a• L as ►.••••iu^••h•.• .^^•..►►.►.r•• ►i^►►s►►}qpq^. L..'
r ry
i ^ « hr I Ith
t' M w o
v
.. ........... ..... ...................... ....IRS.t I : _. • ^.
r ..... ................ ................ ►........... ^... o
•
. I s „^ I ^ aN
LZn jj^
IA `" A I ni
AS
t
-66-
^r ! Ia
I
I R
r^ w
I tul fi
I► p°i
!
^.
IJL
O KWW tl
ice.
Nj97
^.^iiiiilii^aiTi>13•^••••J-_
a IC)U.
J
1 ~I cn
1,
L $ ^^
L $ "o
^;i
!
I
II ..
I ^w
b i
N n Y' r tV ^
ui1
a I amo
1_ ^^. i 1e' ": r I.i .J.
o os
'
IE
^/w i l ^^^ ^^ I~g`+W^
^
Iw
..^ .. sat su lt.rsaar sit .....,ilr t IH IN .. ......• •• .....
SO^Irwsat+} Tarsattaai O ........ ................. } S
t
r
n
y,
e r
s
X
-67
r;h
I-lk
•
TwWOCO
Vb
IL
..0
1 a,
-j 0¢ CQQ Z Q00 ( QWy^^pWJY! ax
aa x ;V
OWS4=S^h•
M . IJJ
0 .. W1-- 0a
►~IU—^r
^ aSdcZ^iZ4Za• uWi
F^CQ.1WOO^ ►•
i00 ^C+uW ^ ^•JE ACY
^ZO1W1 4¢RVryWy^^WS.lW^IYWZOf Wt1t^rGGKK
^
#; k
{3a
e
Q'=iC
Q' ZE
HZ
i
WWOC
A
^ W
Z
S ¢-^
CLo^ N J
a.^. CL
HZ0
e
OU- j
Nwe u 1 .^i t M^ e Y c[
i J0CAL oc"Im mom 1 OWoNt"W ei ga v 'JOV
AW^^^^VV
Wf
^ ^ zr.2
^` Z W I
.^a
LL 3
o ^a
Q auc. w000cJIKQOCIC2W1C da aaWicsz.`'.eatnze.. • • a riaaaa°la
th a
UwAL
,, r
9
t
t _
i
r
j_
-69-
of full power are indicated in Figures 15A-15D with the
two input GE control superimposed. Nozzle area was
held fixed at the full open position which corresponds
to the 62.5 power setting. The peak overshoots are signi-
ficantly lower for the MNA design with no degradation
in performance. Also total fuel consumption (area under
Xmv) is much less for the MNA design.
►
For a full poorer slam of 62.5% to 100%, the 2x2 MNA
design is compared with the two input GE control in
Figures 16A-16D. In every figure, the MNA design performs
very well with respect to the two input QCSEE control,
A comparison of this 2x2 MNA design with the full GE QCSEE
design is indicated in Figures 17A-17D. The only signi-
ficant deviation is, as expected, in P 12 and XM11. These
variables are directly effected by nozzle area, X18.
There is, however; a significant difference in imple-
mentation Logic. This is apparent when considering the im-
plementation of the constants in(38)-(40) versus. the lbgic
outlined in [15, 16).
To regulate 'inlet Mach number, a third controlled variable,
inlet duct pressure (P 12 ), is introduced. Using nozzle
area as the third input, a three input three output design
condition exists„ Although time did not permit -a full scale );s
three input three output MNA design over the full operating
k
4
-70-
^^gu m
l ^ I s
9fu
s Z owl;r o 1
HQ t(•.
1~-.a.1 0.4Wo
•
I I= .N IL'^
N W-'ru_ (A W : 1
1 dW ' L 1 _ V) 'I
LLi a •)
I
'
(M ol 1 a.. ,^ :.000 1
`^
I l ^. - JI IL
^_ + t• w
.............. ....1
__ ; J^^ ; Qr N............. w
Lit I
!W • Z dW
i •tQ..
Z NI X ;,
10
I ^N
c
12 WAAc
a ;isW^„a'r,.;r^
n►.^W I
c 0: I Na x
H ..Imo'
• I cv•
Nrf W r^^, N°
aW i
N
W ^ I
= O I N Q N ;_ aZ ti: : I ^I '' r i
I ^ n a
ct
N ^.^^ L 8 eCD
U-^.. M ^^L
• ;a.
r v'' I y Iii ^•
a.
r•^+ I J
1.... .......................... ........ .• n .•.... ........ ......... M ,
W 3 ^ E 8 a= Z _ $
M
c^
-71-
^r
s
s ;I$^^ ►1Q =I$^a
^ H a I ^s =1A ^) ^d ;:t
gas
1^ ►.^ x I ~^' A W
t~ l
if
_^ I= Wf
N N
' 1 N L = Q j) N
(A W( t .,.+
Q N( ( z iL
cn
I
i 1 S I M ?) Oco
Wopo
-
IL
rr
3• +r_ a. tom• r i w i
• .. Lu . . 4 „1....,.....,..... j. a........................................... ...
CL
Iz
r
.1 0I u+a cto l L O ;I A! 4
^"^ I= vim = H1 ;I=uocO 1-- W W i O F- a CV
!- JN; Mi
IA W N W rW ni W ; I N :5 j
I Q Na
Q No
1 ZC ICO ;(
oLLLe
^Lo IL :Y
J s ,I
3 ^
r ... ....... ......a ....... n..... ....... ..... .. .. j G
=A
44AA
.. .... .......
IL IL
w Tg $ $
•^.»^..;^..,. .W^ . ^. rte, ,^:,^^ .
-72-
^
F
1 ^I ^^j WN
i X.l•
1
I Mp= nx Z H OC 12
It. O Oft
i (. Z Q. I% .off'
9^.^
'` H tip ^° ,aj 'AcM4
v st^ E'
^-+EN
N W
H ,W ,
• K
fN W
na !Z V 1
N d
l 1 Q33ii
,t`,`, ^••: H ••.•. . • ... • .. ... ........• ..... • . ^: N ui
dN ^ ^i ry p1N .^ • N ^
{{
CL
^ (1 ..t O
^=W^u :O1" Qa p F.
r a-3 ~ a_W X_^ (• r' ^( ca► N
CA It MW
QN)ti W ;.^^+ 111
j1 z 0 S _UO
41( W
^N.... ........ ........... ...
MW-----
Z--r`-^ r _..^ .. wAN---------f----^J
_
r
iA V t N ^ N N w ^
iE
9
..73.,
i tU
^I ^ {It
S u.a IA
O {^'=u«O
,Ig W
Z M ^ ^ ^w
Z M^ ^^^
:i^, O^ ; r = oa• r _a
ct ; tLsIN S «+ N 3
SIf W , a to W{ I
a
w;I ,^ W SI ., IQ N
=v I z0'C O
CD
I^ { I O
l I 3_
O 64...•............................. • M ru •.. t. .......•. a .• a...... .......•s,u......- L.^ ^' .i 9N
LLJoff ^•i N -+ i
d
i M
Z ;1 s z I a
O ; ^ o^ O { I Yf3 0 '
~ Z M{ ^ Q
; I _ ^u of O 1-` I '^' Z
^I OJ_
F 14
W I
N
i Z V 4 Z V
1 ^ o it s O {I
I
c . (w, i
•• ,y...... ...........................r
q ^N .i, i •
-
t
-
r
CL
-79-
s •y
^
wev,
oIYy ZI t
UNz 0.4 r^ _ = rA wl►-^ a
a= G }
J C..Y
J C ' +^ 1
z N ;I_^ •
NW NW^ti d
a ' i,Ln
LoIL
l ? I g ^,,,
I
a't ~ r S W W
..y•terr•r...• .••••• ....•. • ... ...•• . .•r•.• ' F` U-i
VO
^'.+ ^ ^ ^ ^' v ^ N N 111 •
a
Z 9
On N
Q6-1 I'
104 A
Io• 3
pqQ
r F 4 5 `°
W : ^. N ^ r1 1
'ILaN i ILw
z v 3s I r o "'
I, rrLN ^~ L" r, I ,
L^ _ ^
s ^ '
_ o, . ..... ....... ... o
•A ^ 1 IY f7 N w w • • y
4
^y
s
as
if!s n ...:. .-. t ^ .. -:: , s_ vr• ..,z. a .. - f z .r. n w. .n^ a .... _..
_ . ^ . __ ^ ":i.:4 ,ww^.^ _ ._ .^ ^. r:.of ♦ a r a r 4 v. ^.r4n a +
-75—
All
A'/I
Y W
JIIV '^. H '7 uWilrmga^ =1 Wo Wo,I .,.N.
Z N= to ;I _ ^ ^1 ^
4h W IL ZN Ws (i.N
SL
SI ^IZ V •_ _ VI= Q , I it ^ Q } INi.
it d LOI M I
1 .............•...t....... •. .. •..........•• ^q • N ^ mom INfr. r.• x •f..1i• 4flf^. •.....•'
O N^.^
N ,.. .r ..^ i^..._ ^.5 ... ,. ^ 5 .., .^ W SJ "...rJ Z
ri
^ 1
Owal
^_ W SA yrr i; • W ©^ ^"^
_( °S
wi
wa Q r^ZIL
We e
r=i N = HN ,Iw=N W ; i w N W
IL
J
LO
1-1d
_. V i. ^ ^_ ^.... S (.
WQ i I Q I O
I^ i ^ ^. 1 i ^, . y. ►ft
t`^ wZ i Z ^^ ^
.. .............. ................................ ^. wA• .' ...... ..... . .............. ^. op
,a ,' g E 8 8' ! N efu
'_^+ryt.v ,_w .. _.,. f -.., ^. ex •, ,. .:. _w ^ ^.sw.:wrf= m.^Yi'..u^M>c. ,a.^.w ..+-.Sir ..zu ....,: ^ .w'0.1.-,faLi- .c, a, _..1^^ _. s _aw_^.. fnr a_. . _ .... _ -
-76—
J Ic
FNWQ 0Ic a
iI
....................................... ... .. ..
IL
J = ^"'
•(A ww
Ic a
UPCD71Ln
J..... ........ ................. ..........
a.wow ri
z0z
z
2 4., CK
cNJcM
Ic49Ic 0 r. 0 Xc.,j
0 ww
V) w i= -^w
z 0 IL z 0. LOc a i c jsn
^( ILJJ
r^I i I ^w
........... .............ti,loj
............................. ... ..........li
UN^^1
2
b = P-3Ii
0 W^ ~ IL
a vii
!
~ mss ^' .
,) a
i ► ^^^
wM, .
^i • 1'1 .r j
rp.
li !
Z r ^o
H;.. Q. ^ (t No
NJ Ic
`" s lN W ^_
Q #A ! " a
Q f _.w
1 ^Z
N
W ^O •0
............................................M
Wit• NF
0 w ^' • •
...77»
L
or^ ziWe
Ic4A €iW IL w
z C.)i
,to
r
d
r Qir ^^.. ^. .......,r^i. a.e, •........................ : 64
ir • r • N •
iqfuI
p
Z Hcu
H¢ Elul, M
N W Z
N AQ rr4r W
ai)
Mj Ir
I ^J
aA .............................................. ~
1—^^ H
!-i
Lt1
NlU
WwnE"
0..Z1-+
3
X,^C M
0
.d t
M(A 'W
W ! ^'
E ^ M
t^..
W' L cm
.^^ t s
A .•..r Y•.•.••..... r••••r.^••I•r•.•Y!••,.•..
^r• H^..
'Rai C^
Z ^ I •^r
c 0J S u(_rrm N
N...,
LUl W =_
t^^L V
lr--1c o ^ ^ a^_
L ^
I}
LLE
t l
IA
r
^^ I 4W '^^^ r i
Y. s :..........r ......... ....r.....•...........rear w
-7$.=
1
I •^ry
:3J_ U)
NW(_
^^' I •tt
i
A. • Y.•Y.•Ii.Y •.•l..1. •. •.I.•!!• .^4r.• r ^ MN
IE ^ = FQ1I x
^a
iU)WI^
Q
N :•.t
Lt .
tS} ^ ^(
I NA
r ^
r
1
N
-.
w
^•
! 1 I a
t %I J
prr^ r
Nn.• ............•...................•.•
r W Yr W. Ih;-,p 9 G '^ ^ N N ry ^ w • •
r
-79-
1+
a Iy0
,
r ;I I sQzI^
^oz M }
I
zky
c01 9: t M M
.. .
z _jN
d„ L^Yyy1
Cl. ^^^ j L M CD
Ln
a c0
^yI.Y
...i R•.J.r i}. ♦....l ll ea•IIR1..e r. tt a.•,, 4i.
M
W.
MA
}^..i}fs..e.s............... r. .. }r r. •.r . ♦ R^ N I juj
a. i^• ii+ i yl ^ ^ F N IV 7w • M R. J
ItH ^ t^ ^ rI ^ `E
t r-^ ;IiH Qa«^ 6I = x_'^}! __r^
i HNs^• c(A 364
1 0 W ^ ^ _ N W I( ~! '
c0 !^ w CO ^^'wuj
1
z ^I
01),,^ ,L• ^, 1 L.. In
n •
!J;^. • ►' 3
I J 14
V+N.................................... ......... 2
..rM ..................... . . . ..... ................J
t L L G ♦ L a
-so-
"ati
ZIL
b.
oF- c19-^ ^ ►^ !a^Q j ( N W ( =.1 E ws
s ^Y .NW;^NQ NZ t)C O r-.^ I
rHL ^ ^
`jj^,. w L!J ^. v,
LnI-•
,
...,.,...., ..... ..........: , ' „ ..... j .............. .................... ^.J
:w r T I_ T _,. _.^ ►,
^cY1 ._ ^ _ _ ^.-...^,.._r,.._ .._ Tr .^ ^
a _ _ (Z
i Vol M 4 r N N i .^. ,
^ Jm
zm
906 pi
;I a ^' s L
M H !M
' j = • WNW ;^__ ^ vfWW ^N 3
Q ^ 4Z ^ ti ^ ^ Z O $ 1.1_
,
IL
w .00 w
IL
j,
.•. •... .... • N IL •. .... . ............................. .... ... h
.
9
}
Ea'
MW
7 A
9i
I
11!
•
i=
WNxhZ
O ,i y
JFI ^
of I~EW ^
'1= •Z C^ ;^N
S O '^ Jt
LI• M
V
L.
i
.,
•i
• sJaM
•ri. . .... . .. . . ..^
p..
^--Z^^T"` .T-1
--1-IM
w
^Op M V N •
i a• i
y1N
• XZ= vha
O ^^
~aI B^
r"
alee
WI;''^
tl
,
Z/111 1 ^ i V'jjj
•
lzl
.^ •7 .. r^L^ ^•.T T .^ ... ^J • r
t
au,
LO
wcn
J
i
i
)Z % ^ X^AI ^
Cj
N1 J Z i t` ^" '"^ ,;
Eul , N I ~1N W
'I C•_ ¢i! Q
Win CA w
ss WN
wZV
aa
W v JM
'
L aNI
oo t
S •
•
4 W I^^^r.►^^ .i- G F. ,:. w
nw_T`__r"T'--1J
Wi•.•z
-T r '- T '^ J ^+n'
•o
lyq::
-82-
range, some analysis and simulation were performed. The
remainder of this section is directed to this condition.
Since the GE nozzle area control was designed using
SISO methods, the SISO MNA and the 2X2 MNA designs were
independently augmented with the GE nozzle area control.
The first case represented in Figures 18A-18D consists
of the SISO MNA with the GE X 1 control while Theta I
is fixed aL the 62.5 power position. Note the rapid
thrust response (same as in Figure 11C) and the respective
decreases in P12 and XMll from their counterparts in Figures
9A and 9D. Execept for fan speed, the Theta I curve for the
full GE control has little influence on the remaining
variables.1
Augmenting the GE nozzle area control to the 2x2 MNAa
control design, results in the transient performancei
indicated in 'Figures 19A-19D. As expected, the dynamics
are unchanged from the previous case except for the improved
performance of P12 and XMll. The MNA control performs very
well with the GE nozzle area control fornoise suppression.
Preliminary design using MNA concepts for the three
input three output case was performed using the 90% power
point. The input vector u(t) and output vector y(t) were
x
'I;i
Le
e4eie n .... e. ... . .._..ve, i .. : r .. ..__ .xd r ._ :.-- v rrhxaa F .._ a+m^ _fPf4 ^u..4 _+J'--•++.. :sx ti ra..r e. vif xu+cC2e .f ,., krx .:. a. vz, ar . ir:^. rr 4... .. .., . _ .. _
i
—83—
^ W tY !
I
SANI
O^I $ z ^^I ^
a m ; I v.•^ t 1^. a i 1. C5W0 ~_ i^N W
iI Ln
QNs^
N #AWL V_0 11 .+ Wav' I '
w
I I i w
U-
.. ...................................• ^_J
.. .. ... _ ............................•
JOVol v
} ! ^4D CO
i
; I
p ^ ^if1r
vF LiJv
J = 3J =^-^ r^
s
= vs lie «.! W W
I
(N W
W s(
^_~i .^
•3
r+J
IL
Q
uj
b ^,^ ( ^—•^ `^^^ i C7Mme.. I ,
f $. M
,
, V
w ....• ..•...w•..... . . . . .......... .o... I•e = .'
• ..w • .^ •^^M
J
r w •fv
-1
1
-84—
I .. I •!
WN ^N
a i h^0.3
Q 0 Q 0I
= N _I _^ I t o
Vf W _ 'a W sLr)
N _ QWf1 = NaI z ° , I C
co _I zo =I
LU
ulCIO
s s
„ ......................................... . .•........ ............................. ^._ aJ ^ w NI'1 • _
QU0 ;ry
L X~ co
uj
°f'ww' sr ^ aIc 1r ^0 i_ « i. M
0 W I 1^^ J t i I ^'! 3_$A
$A W sPI _
N S bi =( = V
I CA)
;L = av :I^/ r I
00
CD
.r
^•.^Y
....»................................................ .............. ,. .................... ....... .......«.... ..,
=i1
k
Iris
-85—
i=N
; Y
Z sls^yro
w ^ I
;I t
Ic(n _^^ s
H-W I .^
z v slC a =1
I
I^:
I
S
I
^
.,
nay
. ...... ....... ............. •. . ............ ^: M.^
N
Uyyyy^^^^ .
vM
L a
Z^
c a I^.°a ItWQ 0 r^ s
N W I= _
M
O0
NCDco ( iw
OctL.L;
cwerr+
_t_J
LLdO
I.. r..•v ................. .....r ....
dFa-
h ^ Y NL.i
S 1JNS I iS^ii IWN^ • 7FC^j!
• Q ^ ! `^I
I •f— J I
S(• _ 1.+ ^^
Q^ ..
•
i'. N~
.I
I_ A
O^N
^w NWS } •
ILI II
N W t^ Z VWz I ..
I 1 `` =v1 ;I
^Ir E a
I ^^ ar^^ 1 ^^ •
i s
^ e ^_
II K C7
I1 •...................................., ^...: ,, ~................ ...... ^Y
.+ L iAN • ^i
N ^M w w • • r i p •
-86-
II •_ I •_
Is ItsM^1 Q
r r^
I We = N 1 I^
^ N iI_~ H N
N LnW ( #A W _
aCA r" a0z
{ Lo
o^ v
i w
LE
it ^1
. I ^, s I J
H .
A ..........................
^.. N. ^ ..... r........1.... ♦ ... ^.. I-
0 .^_,, __ T_. r^^^^ _r. T..—sue.------^^ •
y~j M Y' • N • PI 111 .^OC.J
00.
a . c c . I « ^-
WW c!N H`W i N `
1 ZU t^ ZVI_ I _ ! v A
00
ul_J 1
........ .................................................. I-.».» ......... ........
CKX
-87
t{ ^_ iI rQ
r N _ ^ 2 jar
a rm to o ^ i r'"04 cdc m
Na
C N =^ oi
L i^'N NW fl^i 3W o LC1
c VA i^
MV Z tom,}i It
{ ILLU
' ♦ .............Yi...• ..{.•A..{.{•{,{a.. {. a.a.• ' F. O M. . ___ .................
.,..1111. ^.. N Y
p ^ ri i tii r: N .. ; •, t iQ
a
p I a p ,) ^a
o ZJ
Na (^ HQ+ u Nwoe ^.c c fir^ w9-1 ELD
1 W 0I
^H N O 1
6 NW NW
z ^_ zIL
N
L°= L=
L .............................................. ..... • M 0 .,. ^"; ' jH
} 4" S ►
3,a
fpp ee
- —89 ►" ..
Aji • ^^ ^*
I vz ^^I O' I., CCt
^^ Z I '^
r jrAsi E lie si Ic
N ^I= = MN ) ^= OSAW vy W I . r-iW N ` W ^F
}tj
^ 1
Cl- ^.^ , L^ ,^ a• •^; L=
N W
rt^^``
rI
•
^^^~//
ri O
a •^ • (_>
` wr : N ;
N
low —._^^__—^......_^...—._^a .. WW -;''---T--r--r--^^N ......... •.......... rd
p WM h ^. N
IA
wt W+^
I
N_ h ± r
LU
z ^^ a _ ,I ^^J
lO Qr ^= y^j^ ^ Q j I= ^^ c^0.^.
NJs r + _ W1
Qo
W IL =•' N' W ': I Z V
I
L ;_ L4
^' t``r'
^ Z ^,` t I • Ql
w y
fit
4 ,.................................... ........... ••r
47 ...... ............... ...............4r q.....
.......
r aW-T--.—r..^^..T.^T^IJ ., w S
4 sc 3 X _ N a
y
r1
sak iKS+iswe.._.__.
4 _ _s...:^'E^tkecavvr 'ies.-^.rnv:^-a. s•:^•v.^;r -exu s..^s re^^.w. .... s. .. ^...:: a. ,.. _ ..:
-89-9
i!
:a
i
c `I g " i-''
a= ^I C^ i ♦MIN ( ^! ^ ^ iV1 W Q
Ln
^ N ^ I •• f W I ^'^ '-1 ^,
CV1 _ ! aN -^ 1
NW ^N t0 IN ` N
aul 'I ` ! I
z v ^ I .. «-• i. zo (
- I , ^ L . Y► i ^, z
^^^ I • ; , yr 1 1 V W
.......... st` ..v « ... .r1 t ^. M^.. .. .• ....eMr.•,...1.........1.i...r ..11.f.. w./
Ll
y^_ s F
{ WM iI N J
t
'{ z ;Lw
z ^+ ^+ z ^-+ { `^
Q^ ^ I C• Q^ I I Lt' H
X11 W(
WW
Nf Q
R i► C a
L; ^► rn` ; I • ^^., ^
LUz
I ^ ;
^' ' J i l C7 ', itfr ; s
r.4 n LLnn .,, ..... .. . .«....... . ........,iqu
wl
CLf4p,
4, N w G co*
N YI '1f IY r
.y
'
q.
{
^r
ii t,` „4 r
CD
NW S -(AWW ^''N S.^
O
• Lt
L' r ^ Vryry; r^^ . Y1 ^ d
♦ t^
L .J
4.. ..^ r
IL
WIL 1 j
ui
tog 4z
ry
O^ ^ n
tZ ^`:
oI
a
.as M a^
(a w ' ( H
....
L:Q ► ^N 3 W $Z Z O a Q dff ^.
C!1_ _ x
j ; ( r
fV
i
a ^^ ( 4 N
i I •
r :1 d -,W S .^ W :
00 L
uiIL
N ...«.......... .......................«..... ♦ _ r ..^ ..................... .......... IL r
yr v • ^ s a • r; r y
aria T
u
-91—
chosen as }
a.Xmv NL
f
U(t) Xl$ y(t) P12
Theta I P4
Ixwhere Xla is nozzle area and P12 is inlet duct pressure.iFor this case, system dominance was easily obtained
lfor each linearized operating point in Appendix A. The
CAD procedure of [18, 22) resulted in the following
compensators when 01 t W C 40 radians at the 90% power
point
i.2032E-4 -.01856 0.00157
K .0300 7.9359 -.02023 1
0289 103,, 67 .001195
L Diag{0.0415, 71.455, 0.45231
F Biag{10.0, 0.1, 20.0}
The corresponding DNA, Image, and Bode diagrams for this
operating point are presented in Figures 20-22.
,4I
f-
-9z-
II
f
rI
R I ^ I ^u
Is
I•
Ag A
wlLij
I p^ I a
f I wY^ I ^-
I ' /I Q^^ I
p
^ fz' .T r•.:e:•_;to• kl« Ito :^^:^:^s:tt^sss^:^t^st^t.•^s::ts^:rsrJ- ^ J
PJ
7• z
ia I CIDz
.. s I w
p
f•l
^.1 ..•........•.. R ..•...••.••.•...r..r•rr.•
pS HM Z i NtlOUM
tn 1°w ^ i h_n^ap
rr1 1 1
? :G
^^ war f;
silig
' ^ 7i c...k..l .....;.:.-....,v .... ....-.„r^,^.. _..w .0 s....44n.e+,r^.r.^^ti... ..w «,.._.,_.. .... .,..
-93_
L4~N
MI
IJJ
^ R
I A - +^M
thE Bg
;.^ r"r".' S'. •....... .^.a•^..t......e +^r.^.i`—•;o`.^.'..^'{^11)r3.i5i^ir 1► . j
hM?G
CNf rN
. w••r•r• ' .. • •.}real. ^ ..^ ^ {{ W
i ...» ....................................... •e... ^.^ H J
j_
I N^^a ' ^ Nor►
. ^
^pq ^.^ r,N ^^IGQQ
t WOW ,.imp +U1.4r ^, ( W..
~t ^
• ppCl N
1 w
.. o ^+ a r v ^ h •
r94-
I
fv
v
Usu ^ ' •
u c
1.
i {Cy
MLla
I0
I
W^ 0 V
off Q-/ u '^ } W^ J...V : (
,, z::^ s ss: s. f s cis :r .h ^ Atm.w
to
•I
I _ •
w
O •rrr ... ..... I
• fnn
mom
ro MOMI r.^o1DO ~ wadi
I• Wes• I W.sW.,
M
QWW `' ~rte Y
•►+ >>
I ua^a*w
r jr--.itI
swratAg
MW a a sON
s
-95-
F The MNA designs discussed above demonstrate the utility
of the design concept in achieving good turbofan engine
performance. The :Full potential of the Bode and Nichols
concepts for multivariable plants were not fully utilized
in this design application, however, as dynamic compensators
were not required. in each case, appropriately determined
constant compensators provided adequate closed loop system ^f
performance and feedback gain scheduling was not needed. 1
This desirable condition is due principally to the well be-
haved and easily controlled dynamic system structure.f
Y
t
r
t
S
14
4
k
t
_ YPvSlf^n r,vwrffiYw+neMww k..wr--ifl Wa. ^xa.. i4rr w. 4Y-,as.x-h,.a,an°#Utdd^:au.^s4rw.ei^..f s. anFsLzthe....ucQa Y.x.a :.iJi3r,:ate+.u.,G.•,v..:i_..ks..w v.t.nx> .. .. W .. _.... _
-96
REFERENCES
1. Sobey, A. and Suggs, A., Control of Aircraft andMissile Powerplants, Wiley and Son, Inc., 1963.
2 Michael, G. and Farrar, F., "Development of OptimalControl Modes for Advanced Technology PropulsionSystems,,"' UARL Report M9-11620-1, 1973.
3. Michael, G. and Farrar, F., "Identification of Multi-variable Gas Turbine Dynamics from StochasticInput-Output Data," UARL Report R 941620-3.
4. Stone, C., Miller, N., Ward, M., and Schmidt, R.,"Turbine Engine Control Synthesis," AFAPL ReportTR75, March 1975.
5. Weinberg, M., "A Multivariable Control for the F100Engine Operating at Sea Level Static,"ASD-QTR-75-28 November 1975.
6. Merrill., W. and Leinginer, G., "Identification and y
Dual Adaptive Control of a Turbojet Engine,"IFAC Symposium on Identification and SystemParameter Estimation, Darmstadt, September 1979.
{7. DeHoff, R. and Hall, W., "Multivariable Control Design {
Principles With Application to the F100 TurbofanEngine," JACC, Purdue, 1976.
4
84 DeHoff, R. and Hall, W., "Design of MultivariableCon roller for an Advanced Turbofan Engine,'!1976 CDC, Clearwater Beach, Florida.
9. DeHoff, R. and Hall, W., "Multivariable Design Proceduresfor the F100 Turbofan Engine," Final ReportContract Number F33615-75-C-2053, Systems Control, Inc.,Palo Alto, California, January 1977.
10. DeHoff, R. and Hall, W., "A Turbofan Engine ControllerUtilizing Multivariable Feedback,"_IFAC MVTSSymposium, Fredericton, Canada, 1977
t
i
ij
W
-97-
11. DeHoff, R. and Hall, W., "System IdentificationPrinciples Applied to the Control and FaultDiagnosis of the F100 Engine," 1977 JACC,
FSan....rancLSCO.
12. Lehtinen, B., Soeder, J, Costakis, W., and Sel ,dner, K."Altitude Tests of a Multivariable Control forthe F100 Turbofan Engine," NASA Technical Reportin preparation.
13. Lehtinen, B., DeBoff, R., and Hackney, R., "Multi-variable Control Altitude Demonstration on theF100 Turbofan Engine," Joint Propulsion Control
I Conference, AIAA 79-1204, June 1979.i
14. Sain, M., Peczkowsk , J., and Melsa, J., Alternativesfor Linear Multivariable Control, NEC. Inc., 1977.
i 15. QCSEE-UTW Engine Simulation Report, General. ElectricCompany, NASA CR-134914 1 July 1977.
16. QCSEE-UTW Digital Control. System Design F.eport, GeneralElectric Company, NASA CR-134920; January 1978.
17 Mihaloew, J. and Hart, C., "Rea; Time Digital PropulsionSystem Simulation for Manned Flight Simulators,"NASA TM-78958, July 1978,
18.. Leininger, G;., "Diagonal Dominance for MultivariableNyquist Array Methods Using Function Minimization,"Automatica, Vol.. 15, pp. 339-345, 1979.
19. Leininger, G., "The Multivariable Nyquist Array - TheConcept of Dominance Sharing," in Alternatives forLinear Multivariable Control, Sain et a1 eds., 1977.
20., Leininger, G., "New Dominance Characteristics for theMultivariable Nyquist Array Method," Int. J. Control,Vol. 30, pp. 459-475, 1979.
21. Leinginer, G. "Multivariable Compensator Design UsingBode Diagrams and Nichols Charts," IFAC CAD ofControl Systems Symposium, Zurich, 1979.
t ^
x,.yk y
^..tt.. 't{.iSf"Ys1^,5^+i'^ -- 3-•.S.LNYtYu.tiJ.v"fa[_.vhd[e....cL a m..y,i.11.>Jlm _w.i"W^rc •N
f
-98-
22: Leininger, C., "An Interactive Design Suite for the 1Multivariable Nyqua._st Array Method," IFACCADCS Symposium, Zurich, 1979,
23. Rosenbrock., H, "Design of Multivariable ControlSystems Using the Inverse N uist Array,"Y 9 Y4Proc. IEF`, Vol. 116, 1968.
24. Rosenbrock, H., Computer Aided Design o f Control Systems,{ Academics Press, 19 74 .
25. Fisher, D. G., and Kuon J. and Ex Per mental^ ^ " ComparisonEvaluation of Multivariable Frequency Domain jDesign Techniques," IFAC MVTS Symposium, FFredericton, Canada, 1977.
26. Ostrowski, A., "Note on Bounds for Determinants withDominant Principal Diagonal," Proc. Am Math. Soc. ,Vol. 3, pp. 26-30, 1952.
27. Crossley, T. R., "Envelope Curves to Inverse NyquistArray Diagrams," Int. J. Control,, Vol. 72,, 1975.
t 28 Daniele, C. and Krosel, S., "Generation of LinearDynamic Models from a Digital Nonlinear Simulation,"NASA Technical Paper 1388 0 February 1979.
3
i
rw
"
. - a
G 1r
i
r
"
t i
K-99-
k
5
APPENDIX A
LINEAR MODELS FOR} THE QCSEE ENGINE
1. 62.5% power level - Model includes actuator states.
G 2. 70% power without actuator state s.i
3. 80% power without actuator states.
4. 90t power without actuator states.
5. 100% power without actuator states.3
i
t
X
{
ti
-100—
NW..jCA av a o
.. ar I .- ..^' Q z Na O WIlk U FW W Qo Q a N
^ arN N V fY
^ W Z Qm N m N C pw 0 w m m m 2 H !-a> • m m in 0 m m m m m O Ii` 2 U+ I + I ♦ I + ♦ ♦ ILL QW W W W W W W W W W W W ^+CD A LA IA 00 ♦ m m m N Om m f'1 .+ m 00 .+ m m m m W Wm - ♦ IA A W W+ N m m m V J O
m m m m m m m m m m m Z /- ..1 I ^ ►•1 Q
fU.4^+mm IVmfa.-4CJ4DN0m0N000d00 a ►•m m m m • o m 4 m m m m m m ®• m m m m m m !- U OI f 1 ♦ + I O ►♦ 1 ♦ ♦ 1 ♦ I ♦ + ♦ ♦ + ♦ ♦ Z6-6WWLJWIAJWWWWWWWWWWWWWWWWW W 0'+ N 1-Nm1DtrJm -+ mm^c'1mAmNmNmmmmmm = O N UM W N V 0Am ♦ —A0r-tD V m00mmmmmm a Q Q^..f'f^ mr'1mTNNm•+m^"1m('lmmmmmm
. . . . . . . . . . . . . . ... . . .
II 1 1 1 1
m(v)m`4m(U-o4A(1m(U@mmmmmmmmmmm• m•• m m m^ m m m m m m m m m t^ m m m+ I+ ♦ ♦ 1 + 1 ♦ ♦ + ♦ ♦ ♦ ♦ ♦ + ♦ ♦ ♦ ♦ ♦W W W W W W W W W W W W W W W W W W W W W Wmf'1mA0MMW V —(' W0%D0 V mmmmmmmmm•mlJiAf'1 ♦ f` OOmmAm^+mmmmmm
. . . . . . . . . . . . . . . . . . . . . .m^ m m <O m m m m m m@ m an d m@ m m m m m
1 I 1 I I 1m r+m ^'N r -+ N^+1nRJU1 @ N^^ mm@@mm mmmmm@mmmm^@ m m m m m m m m m m m m m m m m m m m m m O m m m m m m m m m m+ 1+ ♦ + + I .... . .. . .. . . . .. ♦ + ♦+ + ♦ + + + + +W
III W W W W W W W W W W W W W W W W W W W W W W W W W W W W W W W
m00mf'1lJ'tAm.-r00 r+f'1Q (L• N^f^mmmmmm mmmmmmalmmm@@mmf^NA^+`r (^mf'1f^^11f'1mN mmmmmm m@@mm^9mmm@mmcD.3> +•- -+vNm -+ mmmmmm mmmm0V4D W M 4.4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .m@ m m m m@ m m m mm m m mm @ m m m m R? m m m m OD* mmm 6> m m
I 1 I 1 1 Im@(Um -^ M-44) dm@@mmm mm @ mmmmmm @m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m m+ + + + + I I I .......... . ... ♦++♦ ♦+♦ ++♦♦W W W W W W W W W W W W W W W W W W W W W W W W W W W W W W W W WN^DAm00000000AAAf'1f-m ♦ Ammmmmm mmmmmmmmmmm
• AmtDm0000mtDAAmft^ mf'^^mmmmmm • m@mmmmmmmmmQNm r+mlllf'1[` 'rt^ ^AlDr+mf'1lDmmmmmmGOmmmmmmmmm+m
. . . . . . . . . . . . . . . . . . . . . .X@@@@@ m m m m m m m m m m m m m m m m m X m@@ m m@ m@ m d mOC^1J@N@N-+ ►+ f'1(^mN^+^+mN-+mmmmmmOCQ`@mm@mmm •+@mF- m@ m m m@ m m m m@@ m m m m m m m m m 4D m m m m m m m m m m mQ + + I + + I + 1 ... . ... I .. . . .. Q + ♦+++++♦+++= W W W W W W W W W W W W W W W W W W W W W W r W W W W W W W W W W W
lAmr^mm-»N('1ll1111 mdlA00Nmmmmmm mmmmmmm@mmm
H @@mmLAmm00Am f'1r-Mr- @@m@mdl- mmmmmmmmmmmO —m00V-VWWr- r-N AmmmmmmDmmmm@@mm mm
IL . a
Z m m m m m m m m m m m m m m m m m Qie9 m m m Z m ..........
m@ (9 m m m m d m m
►^ 1 I I I I rr
4 )lil(;[NAI, PAGE Iy)F' POOR (QUALln'
N
N(O
QJa
-101-
0r
a
m fn m N m m v v'o m m m m m m m
+ I + I + 1 1W W W W W W W W
m .-4 .4 m r- N NCD O1 — N 00 -+ 00 Qt
m 01 v- 00 N N U)
m m m m m d m mI I
N .-. m CD .-0 m m mm m m W m m m m1 1 + t +
W W W W W W W WN ^ m m m m m m
m N m m mm N m m m
m m m m m m m m1 I
m m m — m N m mm m m m m m m mW W W W W W W Li,m m m V — f- m m(D m m m LA U) m mm m m M .-4 .4 m mm m m m m m m m
1 1 I
m m N .-^ .-. N m -+
OW(Dm m m w m m m m m m m m m m m m+ + + I + + + + I + 1 I + 1 + +W W W W W W W W W W W W W W W Wm m W m Ln m m m VM^M WU) MV
m m N v LA m N N m lD lJl 00 m l!1(D V1
m (p + v N (D U) (D NvNNm rl mNm m m m m m m m m m m m m m m (v
I I 1 I 1 Imm(Um--mf')(V)mCUW (D—m M - (DW - --4MNm m m m m m m m m m m m m m m m m m m m m m m m+ + + + + 1 + 1 + + + + + t + + I + + I + + + +W W W W W W UIW W W W W W W W W W W W W W W W W^ (!D a) (3) 00 CU M Cu qT M V00(Ummr+ m00V(U—MWCY1[-mLn(D00(!D-4 - L).4 NID—mmin n Lor-Wr-VMMW. . . . . . . . . . . . . . . .
X m m m m m m m m m m m m m m m m X m m m m m m N m•--^ I 1 1 ^--^ 1 1 1 1 1 IMNm-+mN c M<SoW (S)NmQ` M N vHmmmmmmmmmmmmmmS ► mf-mmmmmmmmQ + + + + + 1 + I + + + + + + + + Q I + + + + + + +E:WWWWWWWWWWWWWWWLi1EWWWWWWWW
LAMM(p.- LACU (7)mmN (DNmNN 171M - r- (av(DNf-mm0)m00 •- —CUf-N(7)r- -01 IWm1-f- OS) mLn0(r) MC1Q-...a ..Z m m m m m m m m m m m m m m al m Z m m m m m m 9 m
^--+ I I I 1 1 ►-+ I
-102-
+ 1 + ( + + I I
@ (D In m (D v N mm V N r+ v (D (D N@ (M m In 00 1`- m 1-
1N m m
Ir+ @ @ m
Ca @ m m m m 6 mI 1 + + + + + +
(D m CD m In @ m m(D @ m m m m m
m @ m @ CD @ @ @11
9 m W r+ M •-^ @ @Ca m
mm m m @ m
W W W W W W W Wm @ Ca V► tD N m m@ @ CD M (D m m
m m Ca @I
® m m m1
m CD N ^ .1 Cu m + .^@Nf NCD@@m m m m m m m m mCflmm(DmmCa+ + + 1 + + + + I + I I ++++ W W W !s! W W W W WWLAJWI►JWWW@ m (D m v VI N IV M •-+ In r P) 00 (D 0)C'v m N N to 1- N m (n 1` V N v--q V (D V9 m qr •-• (D In (D N v m v m w4 In v. .m 0
.m
.C9
. . . . . . . . . . .m CD m m (!D (S)
II I I I I@@NCD(U •-4 OWM MCfl (XI C9W-4CD -+Cad (T)•-4mN--4Nru@CDw(9wwm mmmm mmm mm mmmCamm(am+ + + + + 1 + I ........ 1++I++++W W W W W W W W W W W W W W W W W W W W W W W W.-4mm<p MMWLAVLn(D(rl.4WMN lnv(p mv-q )mw'?mQ@NNNOOIDfrflnm^m^m n WMv-(nullnWW
A: N@- @^4 V U) M N W r- W OCD ")(-AOM W WNt-MMN.. . . . . . . . . . . . . . . . .X m m m m@ m@@ m m m m m m m m X m m m@@@ m m►-^ 1 1 I ►-^ I I I 1 I I
o MW4CD WCLI M V CD M <V W W M --4 M V W M M M V00 1`- @ mm mmmmmmmmmmmmmf-mmmmm mm4CD
Q ..... 1 + 1 ........ Q I .......11 E W W W W W W W W W W W W W W W WX: W W W W W W W W
ln mNmV (DV>V, ruru -r- (Dmmm (AD V1wIVV mN(VI)HmsmCD('')mru@ — v-W00(7, —t-.-+1-'TQC900•-+V,mN
a :3--+@00mMV-•W- -^N--^-+ WM .-4.q =v NNN •-4 M-4 Ina. .. ... . .. . .... . ..azmmmmmmmmmmmmmmmmzmmmmmmmm►^ I I I I I r-+ 1
-103-
m m m m m m m m+ 1 + I + + 1 1
0 MN ^m 111 r M m
4M 000 7 V0x0
N ^ Nm m m m m m
I
m mI
N -^ m m --4 m m mW m m m m m mmW W W W W W W W(D (^ m m N m (4 9O1 U1 m S1 ^ m m m-+ N 9 m -+ m m mm m m m m m
Im m
Im m m -4 m N m mm m m m m m m m
m m m m N U) m m^ -r f^U mm m m 4T
m m m m m 91
m mm m N
i
--^ --^ N (S> + r-mN^Nmmmm m m m m m m m mmmmmCDsDM+ ♦ + I + + + + 1 + I I +++W W W W W W W W WWWWWWWWm m V -+ 00 v w N M y r- N N w mmm m N m — W 1~ N MMmNm 1 m00CD (P U) N f- W M v M In N N N U). .m m m
.m
. .m m
. . . . . . .m CD mmmmmmmm
1 1 1mmNmN--^mf`1^mMm.-
I 1 1^m.-+m Mw<DNMwNNm m m m m m m m m m m m m m m m m m m m m m m m+ + ♦ + + 1 + I + + + + + + + + I + + I + + + +
WWWWWWWWWWWWWWi ,.AiW WWWWWWWWN m V, m In 00 --+ 1n t- 00 N — M m •4 00 (n V-4-4 ( r) 4 x 0)1.1
a N(S>Nml!) (nfnU)WLnw-+ MNm(`)T a M MMCDWMMWQNm-- (D- IVCUW - Cu —. . . . . r- -- CD(Ur-c»U)--+(UMM-+VM. . .........
o X m m m m m m m m d m m.
m <D m m m X m m m m m m m mO^ r-^ 1 I I ►-^ I I 1 1
^Nm--+mN--^--+M^'mM-.NmNm^-•m^fNU)U)M^I^-mmmmmmmmmma^a^mmmml--mmmmmmmm
Q ♦ ♦ t ♦ t I + I ♦ + ♦K: W W W W W W W W W W W
♦ ♦ + + + Q I + + + + + + +W W W W W>E W W W W W W W W
G. f-mfUmO1N00r-W(Df- U)-+vdf- MU)f-mM Lill NCDF-mmO1m00U)V"t^N(0(001--+M-+•- i.- MV 4W—(UTNf'D V •-4(D-+N - ^--^-+MN-+^MOONM •-4 fv) LI)a .a.zmmmmmmmmmmmmmmmmzmmmmmmmm" I I I 1 1 9-. 1 1
-104-
• M • '. • Q t w• • m • m 4+ V •♦ I ♦ 1 ♦ • I I
s oo a t^ ^ + N 1c• 00 N + f` OD Q l0
• • m m m m m m
1 IN o m m m• m • m m m m o
1 1 ♦ ♦ ♦ % $.W W W W41 W w WLA N0 m (^ m m b.♦ ♦ m m m m m ma f m m N m m mm m m0 m m m m
I ^
Zm m .-^ t'1 N m mm o m m m m m
W W W W W W W Ww TI
m m 00 A t1 t^/ mm S 6 U .^ 111 o mm m m m m m m m1 Im m N N (D mmfU+(UbmmR+ m e o m m o o m o o m o m m m
♦ A I ♦ ♦ ♦ ♦ ♦ ♦ I I
i. W W W W W W W W W W W W W W W
d = .-i '7 M 1'1 111 + — N Fl en• m N N ♦ r N -, 111+N(, —+Nm m Ul) 00 IA a (o N1+wm+mr. . . . . . . . . . . . . . .o m 0` m
. m m m br m a1 m m m 4D* m m
I I 1 1 I 1.+mNmN^tJf`1('1mf'I^^a10 +m mNmN-+NNNo•ommmmmmmmmmmom mmammmme♦ ♦ ♦ ♦ ♦ I ♦ I ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ <• ♦ 1 1 ♦ ♦W W W W W W W W W W W W W W W W W W W W W W W WV WW0r` a(D-4 0Nf-Mw0Nm f'1('1+U1f`1mwmNm1/16r++(^N+ANM+mCUC1 • f- ('1mfu00 -y.4+
p Q - (D -- (D - IV fur- 00N-4r-WaD"f-m.-..4mvNf'1f-{AlCD-4 X m m m m (ID* m m m m m m b m m m m X m m m m m m m mH I I I^ 1 1 I I I I
II •NmNar ('1-^-+f'1f'1m('1 -+NmNm^ r+m+Nllllllf'1^'F-ommommmommmmmmmmb-mommmmmm
q Q ♦ ♦ ♦ ♦ ♦ 1 ♦ I ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ Q• 7 It W W W W W W W W W W W W W W W W= W W 4 I W W W W WLL OOmmm00 +(°1LA ••.-rf^NANf'1+ .af^ 001111^O100t^
H NmNmm^Of^ ^ IAm+mf'1111SON1~- oNb('1('1+f'1(O••m—(9--'!.+1-fTfnNN-4(nN•^700--^N(7)--.M--111
IL . . . . . . a . . . . . .CD m m m m m m m m m m m m m m .m Z m N m m m m m mZ " 1 I I I I ";
top related