5.

Interaction and De-coupling of Control Loops

In multivariable control schemes, it is important to consider the interaction
between different loops when the operation of one loop interferes with the
operation of one or more other loops. Examples of such interactions will be
discussed below.
Example (1): The interaction of control loops in a stirred tank heater
Fi, Ti
When the inlet flow rate or the desired
value of the liquid level change, loop (1) Loop (2)
TC LC
manipulates the valve of the effluent
flow. This disturbs the temperature and Steam Loop
thus loop (2) will adjust the value of the (1)
steam flow rate.

F, T
However, if the temperature of the inlet stream or the desired value
of temperature (set point) changes, loop (2) will adjust steam flow
rate to compensate for the changes. This will leave the liquid level
undisturbed.

Therefore, loop (1) affects loop (2) but not the vice versa.
Example (2): The interaction of control loops in a CSTR
If the inlet concentration or F, cAi,Ti
desired output concentration Set Point
TC
(set point) changes, loop (1)
manipulates the feed flow Loop (2)
rate, which in turn disturbs Coolant cA,T CC Set Point
the reactor temperature. Loop (1)
Effluent
Loop (2) attempts to compensate for the change in temperature by varying
the coolant flow rate, which in turn affects the effluent concentration.

If the feed temperature is changed or the desired set point of the reactor
temperature is changed, this causes the effluent concentration to vary. Loop
(1) attempts to compensate for the change in effluent concentration by
varying the feed rate, which disturbs the reactor temperature.

Therefore loop (1) and loop (2) interact in both directions.

Example (3): The interaction of control loops in a distillation column

TC
Example (3): The interaction of control loops in a distillation column
When a distillation column is controlled by conventional control loops, it is
not advisable to specify the composition of both products as controlled
variables, though the design of the column itself is usually based on the a
priori specification of these compositions. This is completely legitimate
from the static point of view.

The dynamics of controlling both top and bottom temperatures, are

however, liable to interference between the two interacting loops resulting
in sustained composition oscillations.

Regulating the top product composition by increasing the reflux rate will
affect the bottom product composition (temperature). The bottom product
composition controller will attempt to remedy the situation by increasing
the boil up rate (steam input rate) which will in turn affect the top product
temperature (composition). This will be reflected as sustained oscillations
with each controller attempting to counter act the effect of the other one.
Block Diagram of Interacting Control Loops
m1(s) +
Consider the shown process with H11(s) y1(s)
+
two controlled outputs and two
manipulated inputs, the input- H21(s)
output relationships could be
given by the following equations: H12(s)
+
m2(s) +
H22(s) y2(s)

Process
y 1 (s)  H 11 (s)  m 1 (s)  H 12 (s)  m 2 (s)
(A)
y 2 (s)  H 21 (s)  m 1 (s)  H 22 (s)  m 2 (s)
It is clear that a change in m1 or m2 will affect both y1 and y2.
Two control loops are formed by coupling m1 with y1 and m2 with y2
as shown in the figure below.
The transfer functions of Loop 1
measuring devices and final + - 1 m1(s)
+ y1(s)
control elements in both y1,SP Gc1(s) H11(s)
+
loops are assumed to be
H21(s)
equal to unity for simplicity.
H12(s)

Gc1(s) and Gc2(s) are the + 2 m2(s)

H22(s) + + y2(s)
y2,SP Gc2(s)
transfer functions of the -
Loop 2
two controllers.

m 1 (s)  G c1 (s)[ y 1,SP (s)  y 1 (s)]

(B)
m 2 (s)  G c 2 (s)[ y 2,SP (s)  y 2 (s)]
 Effect of input changes on outputs when both loops are closed
Initially the process is Loop 1
-  m1(s)
at steady-state with y1,SP + 1
Gc1(s) H11(s) + y1(s)
both outputs at their +
desired values. For a H21(s)
step change in y1,sp
only, hence y2,sp = 0, H12(s)
the following will
+ 2 m2(s)
H22(s) + +
happen: y2,SP Gc2(s) y2(s)
-
Loop 2
Red Lines show indirect effect of loop 2 on loop 1
a. Controller of loop 1 will change m1 so as to bring output y1 to the new
set point value, i.e. direct effect of m1 on y1 through loop 1.
b. m1 will also disturb y2 from its steady state value. The controller of
loop 2 attempts to compensate variations in y2 by changing m2. But
change in m2 changes y1, i.e. indirect effect of m1 on y1 through loop
2.
Design of Non-Interacting Control Loops
In the case of strongly interacting loops the designer should use
Decouplers to cancel the interaction effects between the two loops
and transform them to non-interacting loops.
Reconsider the following figure where m1 is coupled with y1 and m2
with y2: Loop 1
+ - 1 m1(s)
+ y1(s)
y1,SP Gc1(s) H11(s)
+
y1 (s)  H11.m1 (s)  H12 .m 2 (s) H21(s)

y 2 (s)  H 21.m1 (s)  H 22 .m 2 (s) H12(s)

+ 2 m2(s)
H22(s) + + y2(s)
y2,SP Gc2(s)
-
Loop 2

Suppose that a disturbance or a set point change causes the controller

of loop (2) to vary the value of m2 (undesired disturbance for loop (1))
which will cause y1 to deviate from its desired value.
A feed forward control action may be used to reject this disturbance
by changing m1 in the direction to cancel the interaction effect of m2.
H12
This may be done by adding a signal  m2 to the output signal of
H11
Gc1.
It is obvious that the transfer function of this feed forward controller
will be:
H12
D1 (s)  
H11
This element is called a Decoupler.
Decoupler (D1) is used to cancel the effect of loop (2) on loop (1).

Similarly, a decoupler (D2) may be used to eliminate the effect of loop

(1) on loop (2). It is obvious that this decoupler would have the
transfer function:
H 21
D2 (s)  
H 22
 Loop 1
-  m1(s)
y1,SP +
1
Gc1(s) + + y1(s)
H11(s)
+ +
D2
H21(s)

D1 H12(s)

+ 2 + m2(s) +
+
y2,SP Gc2(s) + H22(s) y2(s)
-
Loop 2
H12 H 21
D1 (s)   D2 (s)  
H11 H 22
H12
y1  ( y1,sp  y1 )G c1H11  ( y2,sp  y2 )G c 2 H11  ( y2,sp  y2 )G c 2H12
H11
H
y2  ( y2,sp  y2 )G c 2 H 22  ( y1,sp  y1 )G c1 21 H 22  ( y1,sp  y1 )G c1H 21
H 22
As seen from the above equations, y1 depends only on y1,sp and y2
depends only on y2,sp.
 G c1H11 G c 2H 22
y1  y1,sp and y2  y2,sp
1  G c1H11 1  G c 2H 22

It is clear from the above equations that the modified control system
structure in which the two decouplers (D1) and (D2) have been
incorporated result in complete decoupling. The equations show that
the system dynamics after introduction of the decouplers is thus
reducible to the non-interacting independent control loops shown
below.

Loop 1
-
y1,SP + Gc1 H11 y1(s)

y2,SP + Gc2 H22 y2(s)

-
Loop 2
It is recalled that decouplers are essentially feed forward controllers
the transfer functions of which are not those of standard (PID)
controllers.
In some cases, they may be approximated to “Lag/Lead”
compensation structures. Also, the physical realizability of the
decouplers depends on the form of the transfer functions (H21 and
H12). If an (e+s) term is suggested by the decoupler transfer functions
the system is not physically realizable. In such cases the (e+s) term
and perhaps other dynamic elements (containing “s”) in the
decoupler transfer function may be overlooked and their transfer
functions will reduce to the ratio between the gains of the transfer
functions H11, H12, H21, and H22.

When this is done decoupling is termed “Static Decoupling” which

although unable to completely eliminate the dynamic interaction
between the two loops, is better than no decoupling at all.
 Sheet 2, Continued
4. We like to regulate the composition of distillate and bottom products
from a distillation column, using the reflux flow R and vapour flow in the
column V (or equivalently the steam flow rate) as manipulated variables.
Suggest a decoupler for the column. The following input-output
relationships for the distillation column have been determined
experimentally.
0.6𝑒 −1.1𝑠 0.5𝑒 −𝑠
a. 𝑥𝐷 = 𝑅− 𝑉
(1 + 5𝑠)(1 + 2𝑠) (1 + 6𝑠)(1 + 3𝑠)
0.3𝑒 −1.3𝑠 0.5𝑒 −𝑠
𝑥𝐵 = 𝑅− 𝑉
(1 + 5𝑠)(1 + 𝑠) (1 + 5𝑠)(1 + 𝑠)

0.35𝑒 −𝑠 0.5𝑒 −2.5𝑠

b. 𝑥𝐷 = 𝑅− 𝑉
(1 + 20𝑠)(1 + 25𝑠) (1 + 15𝑠)(1 + 10𝑠)

0.85𝑒 −2𝑠 0.9𝑒 −𝑠

𝑥𝐵 = 𝑅− 𝑉
(1 + 40𝑠)(1 + 50𝑠) (1 + 30𝑠)(1 + 40𝑠)
 0.6𝑒 −1.1𝑠 0.5𝑒 −𝑠
𝑥𝐷 = 𝑅− 𝑉
(1 + 5𝑠)(1 + 2𝑠) (1 + 6𝑠)(1 + 3𝑠)
0.3𝑒 −1.3𝑠 0.5𝑒 −𝑠
𝑥𝐵 = 𝑅− 𝑉
(1 + 5𝑠)(1 + 𝑠) (1 + 5𝑠)(1 + 𝑠)
0.6𝑒 −1.1𝑠 Loop 1
𝐻11 = -  R
(1 + 5𝑠)(1 + 2𝑠) XD, SP 1
Gc1(s) +
H11(s) + XD
+
+ +
−0.5𝑒 −𝑠
𝐻12 = D2
H21(s)
(1 + 6𝑠)(1 + 3𝑠)
0.3𝑒 −1.3𝑠
𝐻21 = D1 H12(s)
(1 + 5𝑠)(1 + 𝑠)
+ 2 ++ V +
+
−0.5𝑒 −𝑠 XB, SP Gc2(s) H22(s) XB
𝐻22 = -
(1 + 5𝑠)(1 + 𝑠) Loop 2
𝐻12 5 +𝑠 (1 + 5𝑠) (1 + 2𝑠)
𝐷1 = − = 𝑒
𝐻11 6 (1 + 3𝑠) (1 + 6𝑠)
𝐻21 3 −0.3𝑠
𝐷2 = − = 𝑒
𝐻22 5
Could be approximated by Pade approximation for dead time
 𝐻12 5 +𝑠 (1 + 5𝑠) (1 + 2𝑠)
𝐷1 = − = 𝑒
𝐻11 6 (1 + 3𝑠) (1 + 6𝑠)
Approximation: to a lag compensator
𝐻12 5 (1 + 𝑏𝑠)
𝐷1 = − =
𝐻11 6 (1 + 𝑎𝑠)

1 1 1 11 1
6 𝑎 5 3𝑏 2
 0.35𝑒 −𝑠 0.5𝑒 −2.5𝑠
𝑥𝐷 = 𝑅− 𝑉
(1 + 20𝑠)(1 + 25𝑠) (1 + 15𝑠)(1 + 10𝑠)

0.85𝑒 −2𝑠 0.9𝑒 −𝑠

𝑥𝐵 = 𝑅− 𝑉
(1 + 40𝑠)(1 + 50𝑠) (1 + 30𝑠)(1 + 40𝑠)
Loop 1
0.35𝑒 −𝑠 - 
𝐻11 = XD, SP 1 + R +
(1 + 20𝑠)(1 + 25𝑠) + Gc1(s) H11(s) XD
+ +
−2.5𝑠
−0.5𝑒 D2
H21(s)
𝐻12 =
(1 + 15𝑠)(1 + 10𝑠)
0.85𝑒 −2𝑠 D1 H12(s)
𝐻21 =
(1 + 40𝑠)(1 + 50𝑠) 2 +
XB, SP + Gc2(s) ++ V
H22(s) +
XB
−0.9𝑒 −𝑠 -
𝐻22 = Loop 2
(1 + 30𝑠)(1 + 40𝑠)
𝐻12 −1.5𝑠
(1 + 20𝑠) (1 + 25𝑠)
𝐷1 = − = −1.43𝑒
𝐻11 (1 + 15𝑠) (1 + 10𝑠)
𝐻21 −𝑠
(1 + 30𝑠)
𝐷2 = − = 0.94𝑒
𝐻22 (1 + 50𝑠)
Could be approximated to a lag compensator
 𝐻12 −1.5𝑠
(1 + 20𝑠) (1 + 25𝑠)
𝐷1 = − = −1.43𝑒
𝐻11 (1 + 15𝑠) (1 + 10𝑠)

Approximation: to a lead compensator

𝐻12 (1 + 𝑎𝑠)
𝐷1 = − = 1.43
𝐻11 (1 + 𝑏𝑠)

1 1 1 1 11
𝑎 25 20 1510𝑏
 7. Adaptive Control
As mentioned previously, process plant parameters are likely to
change during plant operation e.g. non-linearities affect the
system parameters on changing steady-states in addition to other
parameter variations.

This means that the controller settings calculated assuming the

nominal transfer functions of the loop elements will not give good
performance when appreciable variations take place in the
process parameters.

In such cases it is necessary to consider retuning the controller to

compensate for the variation of the process plant parameters in
order to ensure a good control quality e.g. integral square error,
decay ratio, gain and phase margins, etc. This control approach is
called adaptive control.
Adaptive control thus performs two functions:
First, it must sense variations in plant dynamics. This process, which
is called identification, is generally a partial measurement, since
total identification is usually impossible or too complex.
Secondly, the adaptive device must be able to adjust some
controller parameter or function. This step is called actuation. The
basic adaptive control arrangement is shown in the figure.
Identification Unit
 = Generalized Error

Plant Model +
Actuator
Gm -

U
+ E P
R Gc Gv G C
-
B
H

Ordinary Control Loop

 Within these broad requirements there are obviously many possible
schemes two of which will be discussed below.
Self Adaptive Control
In this scheme measures are taken to change the controller parameters
continuously on the basis of process measurements till an optimum response
is detected. Reference Model Output
Outer Loop Model

New values for the Adaptive m +

Controller Parameters Mechanism -

R + E Controller Process C
B-
Inner Loop
An idealized form of what should be the response of the process to a
reference input variation (or a program of such variations in batch
processes) is generated through a model. This is compared with the actual
response and the controller parameters modified in order to minimize:
t
 m
(  ( t )) 2
dt
0
Self Tuning Regulators
The self tuning regulator estimates first the parameters corresponding
to an approximate model transfer function by analyzing the dynamic
relation between an input and an output of the process.

The calculated values of process parameters are used to compute the

optimal controller settings.
Parameter Adjustment Mechanism

New values of Outer Loop

controller m Parameter C
Estimation
parameters
R + E Controller m Process C
B-
Inner Loop
The parameter estimation is usually based on regression analysis from plant
operating data. It is noticed that under proper steady-state conditions, the
variations of (m) and (C) are small which makes the parameter estimation
rather difficult.
 Sheet 2, Continued
7. Sometimes it is simpler to base an adaptive system on an input
measurement than on a variable within the loop. Considering the
figure below, from this point of view, what function would you
suggest for the model?

U
Actuator Model X +
-

R +
Gc Gv G Y
-
B

H
 Original loop dynamics is given when Y=X as: 
U
𝐺 𝐺𝑐 𝐺𝑣 𝐺 Actuator Model
X +
𝑋= 𝑈+ 𝑅 -
1 + 𝐺 𝐻 𝐺𝑐 𝐺𝑣 1 + 𝐺 𝐻 𝐺𝑐 𝐺𝑣
+
The model should mimic this loop and R Gc Gv G Y
-
the difference used to modify 𝐺𝑐 to B

𝐺𝑐′ such that X-Y=0 H

′ ′ ′ ′
𝐺 𝐺𝑐 𝐺𝑣 𝐺
∴𝑌= ′ ′ ′ ′ 𝑈 + ′ ′ ′ ′𝑅
1 + 𝐺 𝐻 𝐺𝑐 𝐺𝑣 1 + 𝐺 𝐻 𝐺𝑐 𝐺𝑣
𝐺𝑐 𝐺𝑣
The actuation function depends on 

the number of parameters that Actuator

U +
+ 𝑮 X+
𝟏+𝑮𝑯𝑮 𝑮
have to be changed in 𝐺𝑐 . 𝒄 𝒗
-

Elaborate computer calculations +

+ +
R ′
𝐺𝑐 𝐺𝑣′ 𝐺′ Y
may be required to compute the -
B
necessary actuation scheme
depending on the precision 𝐻′

required.
Assuming that 𝐺𝑣 and H are not changed and considering reference input changes
only, it can be shown that the actuation should change 𝐺𝑐 to 𝐺𝑐′ such that 𝐺𝑐′
′ 𝐺′
= 𝐺𝑐 𝐺/𝐺 . Considering load changes, 𝐺𝑐 should be modified to such that
𝐺
1+𝐺 ′ 𝐻 ′ 𝐺𝑐′ 𝐺𝑣′
=
1+𝐺 𝐻 𝐺𝑐 𝐺𝑣
 8. Inferential Control
When it is difficult to measure the controlled variable of interest,
another variable affected by the same disturbance and the same
manipulated variable is measured. Based on a reasonable plant model,
the measurements are used to estimate the value of the difficultly
measurable controlled variable to be handled by the control system.
D Unmeasured
Disturbance
Y  G p1 M  G d1 D
Gd1 Gd2

Z  Gp 2 M  Gd 2 D + E M + Unmeasured Control
R Gc Gp1 + Y
- Variable
1 Gp2 Ye + Secondary
D Z M Gp2 + Z
Gd2 Gd2 Measurement
G d1
  G p1  G p2
G d1 Gd

Y  G p1  G p 2 M  1 Z
G d2
 G d2  G d2
  Estimate of unmeasured + G d1
output y + G d2
Estimator
 D Unmeasured
Disturbance

Gd1 Gd2

+ E M + Unmeasured Control
R Gc Gp1 + Variable
Y
-
Ye + Secondary
Gp2 + Z
Measurement

G d1
G p1  G p2
G d2
Estimate of unmeasured + G d1
output y + G d2
Estimator

The controller (Gc) may be tuned on the basis of the loop:

D Gd1
+
+ E Gc M Gp1 + Y
R
-
Ye
 Sheet 2, Continued
8. Construct an inferential control system to regulate the distillate composition of a
distillation column operating under constant feed flow rate when the following
transfer functions are provided:
x D 0.9e 2s x D 1.2e s T1 0.2e 2 s T1 1.0
   
x F 70s  1 R 30s  1 xF 60s  1 R 20s  1

xF is the composition of the feed, xD the distillate composition, R the reflux flow,
and T1 the temperature at the top tray. Assume that concentrations cannot be
measured and that R is the manipulated variable.
•Draw the resulting block diagram and place all relevant transfer functions.
•Since the parameters of the transfer functions above may change, augment
the inferential control system with an adaptation mechanism. This mechanism
can use intermittent measurements of xD and adjust accordingly the parameters
of the inferential controller.
 𝑋𝐷 0.9𝑒 −2𝑠 𝑋𝐷 1.2𝑒 −𝑠 𝑇1 0.2𝑒 −2𝑠
𝐺𝑑1 = = 𝐺𝑝1 = = 𝐺𝑑2 = =
𝑋𝐹 1 + 70 𝑠 𝑅 1 + 30 𝑠 𝑋𝐹 1 + 60 𝑠
𝑇1 1
𝐺𝑝2 = =
𝑋𝐷 = 𝐺𝑑1 𝑋𝐹 +𝐺𝑝1 𝑅 𝑅 1 + 20 𝑠
𝑋𝐹
𝑇1 = 𝐺𝑑2 𝑋𝐹 +𝐺𝑝2 𝑅
1 𝐺𝑝2 Gd1 Gd2
𝑋𝐹 = 𝑇1 − 𝑅 E
𝐺𝑑2 𝐺𝑑2 SP + Gc Gv R Gp1 +
+
𝑋𝐷
-
X
1 𝐺𝑝2 De
Gp2 +
+
𝑇1
𝑋𝐷 = 𝐺𝑑1 ( 𝑇1 − 𝑅) + 𝐺𝑝1 𝑅
𝐺𝑑2 𝐺𝑑2 Hm2 Hm1

𝐺𝑑1 𝐺𝑝2 𝐺𝑑2 𝐺𝑝1 − 𝐺𝑑1 𝐺𝑝2

𝐺𝑑1
𝑋𝐷 = 𝑇1 + (𝐺𝑝1 − )𝑅 𝐻𝑚2 𝐺𝑑2
𝐺𝑑2 𝐺𝑑2 + 𝐺𝑑1
F + 𝐻 𝐺
𝑚1 𝑑2
Estimator
Estimated Measured
÷
𝐺𝑑1 𝐺𝑑2 𝐺𝑝1 − 𝐺𝑑1 𝐺𝑝2
𝑋𝐷 = 𝑇1 + 𝑅
𝐺𝑑2 𝐺𝑑2
 Sheet 3, Continued
3. Construct an inferential control system to regulate the distillate
composition of a distillation column when the following transfer functions
are provided:
𝑋𝐷 10𝑒 −10𝑠 𝑋𝐷 25𝑒 −5𝑠 𝑋𝐷 5𝑒 −0.5𝑠 𝑋𝐷 0.05
= = = =
𝐹 1 + 10 𝑠 𝑋𝐹 1 + 5 𝑠 𝑅 1+ 𝑠 𝑇1 1 + 0.5 𝑠

xF and f are the composition and flow rate of the feed, xD the distillate
composition, r the reflux flow, and t1 the temperature at the top tray.
Assume that concentrations cannot be measured and that r is the
manipulated variable.
•Draw the resulting block diagram and place all relevant transfer
functions.
•Since the parameters of the transfer functions above may change,
augment the inferential control system with an adaptation mechanism.
This mechanism can use intermittent measurements of xD and adjust
accordingly the parameters of the inferential controller.
 𝑋𝐷 10𝑒 −10𝑠 𝑋𝐷 25𝑒 −5𝑠 𝑋𝐷 5𝑒 −0.5𝑠 𝑋𝐷 0.05
𝐺𝑑1 = = 𝐺𝑑3 = = 𝐺𝑝1 = = =
𝐹 1 + 10 𝑠 𝑋𝐹 1 + 5 𝑠 𝑅 1+ 𝑠 𝑇1 1 + 0.5 𝑠
𝐹 𝑋𝐹
𝑋𝐷 = 𝐺𝑑1 𝐹 + 𝐺𝑑3 𝑋𝐹 +𝐺𝑝1 𝑅
𝑇1 = 𝐺𝑑2 𝐹 + 𝐺𝑑4 𝑋𝐹 +𝐺𝑝2 𝑅 Gd1 Gd2 Gd3 Gd4

1 𝐺𝑑2 𝐺𝑝2 R + + + + 𝑋𝐷
𝑋𝐹 = 𝑇 − 𝐹− 𝑅 Gp1
𝐺𝑑4 1 𝐺𝑑4 𝐺𝑑4 + +
Gp2 + + 𝑇1

1 𝐺𝑑2 𝐺𝑝2
𝑋𝐷 = 𝐺𝑑1 𝐹 + 𝐺𝑝1 𝑅 + 𝐺𝑑3 ( 𝑇 − 𝐹− 𝑅)
𝐺𝑑4 1 𝐺𝑑4 𝐺𝑑4
𝐺𝑑3 𝐺𝑑2 𝐺𝑑3 𝐺𝑝2 𝐺𝑑3
𝑋𝐷 = 𝐺𝑑1 𝐹 + 𝐺𝑝1 𝑅 + 𝑇 − 𝐹− 𝑅
𝐺𝑑4 1 𝐺𝑑4 𝐺𝑑4
𝐺𝑑1 𝐺𝑑4 − 𝐺𝑑2 𝐺𝑑3 𝐺𝑑3 𝐺𝑝1 𝐺𝑑4 − 𝐺𝑝2 𝐺𝑑3
𝑋𝐷 = 𝐹+ 𝑇1 + 𝑅
𝐺𝑑4 𝐺𝑑4 𝐺𝑑4
𝑇1 𝑋𝐷 𝑋𝐷 5𝑒 −0.5𝑠 1 + 0.5𝑠 (1 + 0.5𝑠) 𝑒 −0.5𝑠
𝐺𝑝2 = = ÷ = = 100𝑒 −0.5𝑠
𝑅 𝑅 𝑇1 1 + 𝑠 0.05 1+ 𝑠
𝑇1 𝑋𝐷 𝑋𝐷 10𝑒 −10𝑠 1 + 0.5𝑠 (1 + 0.5𝑠)
𝐺𝑑2 = = ÷ = = 200𝑒 −10𝑠
𝐹 𝐹 𝑇1 1 + 10 𝑠 0.05 1 + 10 𝑠

𝑇1 𝑋𝐷 𝑋𝐷 25𝑒 −5𝑠 1 + 0.5𝑠 −5𝑠

(1 + 0.5𝑠)
𝐺𝑑4 = = ÷ = = 500𝑒
𝑋𝐹 𝑋𝐹 𝑇1 1 + 5 𝑠 0.05 1+5𝑠
 𝑮𝒅𝟏 𝑮𝒅𝟒 − 𝑮𝒅𝟐 𝑮𝒅𝟑 𝑮𝒅𝟑 𝑮𝒑𝟏 𝑮𝒅𝟒 − 𝑮𝒅𝟑
𝑿𝑫 = 𝑭+ 𝑻𝟏 + 𝑹
𝑮𝒅𝟒 𝑮𝒅𝟒 𝑮𝒅𝟒

𝐹 𝑋𝐹

Gd1 Gd2 Gd3 Gd4

E
SP + Gc G R Gp1 + + + + 𝑋𝐷
-
v
XDe + +
Gp2 + + 𝑇1

Hm3 Hm2 Hm1

𝐺𝑝1 𝐺𝑑4 − 𝐺𝑑3 𝐺𝑝2 𝐺𝑑1 𝐺𝑑4 − 𝐺𝑑3 𝐺𝑑2

𝐻𝑚3 𝐺𝑑4 𝐻𝑚2 𝐺𝑑4
++ + 𝐺𝑑3
F + 𝐻𝑚1 𝐺𝑑4
Estimator
Estimated Measured
÷