EG55P7 - Tutorial 6

Cascade control

Problem 1.
For the paper drying process shown in the figure below, the following information is available: The
flow control loop (FIC47) can be represented by a first-order lag with a gain of 4 (m3 h-1)/%CO and a
time constant of 0.1 min. The transfer function of the air heater outlet temperature to the fuel flow
is a second-order lag with time constants of 2 min and 0.8 min. A change in fuel flow of 1 m3 h-1
causes a change of 2°C in the outlet air temperature. The drier can be represented by a first-order
lag with a time constant of 5 min. A change of 1°C in inlet air temperature causes a change in outlet
moisture of 0.5 mass percent. The moisture transmitter (MT47) has a range of 0 to 6 mass percent
and a negligible time constant.

(a) For the control scheme shown in the figure, draw the block diagram of the moisture control loop
showing the transfer functions. Decide on the fail position of the control valve and the controller
action, and make sure that the signs in your block diagram correspond to your decisions.

(b) Consider a cascade control scheme using an outlet temperature sensor with a range of 120 to
170°C installed on the air line from the heater to the drier and a temperature controller to
manipulate the fuel flow set point; the output of the moisture controller (MC47) sets the set point of
the outlet air temperature controller. Draw the instrumentation diagram and the block diagram of
the cascade control scheme. Show the defined transfer functions on the block diagram, and specify
the action of each controller.
Solutions
(a) This is a cascade control scheme. The flow control loop can be represented by a first-order lag
K
process, so the transfer function will follow the generic equation G  s   , where K is the gain
 s 1
and τ the time constant. This process refers essentially to the dynamic response of the valve
manipulating the flow. The valve receives a signal from the FC-47 controller (%CO-controller output)
and affects accordingly the flow (m3 h-1). Hence we have:

GV 
KV

4
 V s  1 0.1s  1

m3h 1 / %CO 
The air heater can be represented by a second-order process, so the transfer function will follow the
K
generic equation G  s   , where K is the gain and τ the natural period of oscillation
 s  2 s  1
2 2

and ζ the damping factor. As is known from process control theory, this second order transfer
function can be further proven to be equal to the product of two first order processes
K
G s  with τ1 and τ2 being the equivalent time constants. Moreover, the gain
1s  1 2 s  1
specifies essentially how much the output changes per unit change in input, or how much the input
affects the output. That is, the gain defines the sensitivity relating the output and input variables,
which can be defined mathematically as follows:

𝛥 𝑜𝑢𝑡𝑝𝑢𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒
𝐾=
𝛥 𝑖𝑛𝑝𝑢𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒

It is given that a change in fuel flow of 1 m3 h-1 causes a change of 2 °C in the outlet air temperature,
so the gain of the air heater transfer function will be equal to 2 °C/(m3 h-1). Hence we have:

GH 
KH

2
 H ,1s  1 H ,2 s  1  2s  1 0.8s  1
 o
C / m3h1 
The drier can be represented by a first-order lag process, so the transfer function will again follow
K
the generic equation G  s   . It is given that a change of 1°C in inlet air temperature causes a
 s 1
change in outlet moisture of 0.5 mass percent. Here is needs to be also considered that an increase
in air temperature leads to a decrease in outlet moisture and vice versa. So the gain of the drier
transfer function will be equal to -0.5 %Moist./°C. Hence we have:

0.5
GD 
KD

 D s  1 5s  1

% Moist. / o C 
It is also given that the moisture transmitter has a range of 0 to 6 mass percent and a negligible time
constant. This means it can be represented as a zero-order transfer function, essentially just a gain.
The transmitter receives an input signal of moisture mass percent (ranging from 0 to 6%) and
generates an output signal (TO-transmitter output) that ranges from 0 to 100%. As such, its gain is
equal to:

GMT  s   K MT 
100  0  %TO  16.67 %TO / %Moist.
 
 6  0  %Moist.
For the flow transmitter no data are given. Assuming that it also has very fast dynamics so that it can
be represented as a zero-order transfer function and that it has a measuring range from 0 to
100 m3 h-1, its gain will be:

100  0  %TO  1 %TO / m3h1

GFT  s   K FT 
100  0  m3h1
 
The respective block diagram considering the above is given below:

MC-47 FC-47 Valve Heater Drier

R(s) + E(s) MM(s) + EF(s) MF(s) 4 F(s) 2 TH(s) 0.5 X(s)

GCM GCF
%TO - % %CO - % %COF 0.1s  1 m3/h  2s  1 0.8s  1 o
C 5s  1 % Moist.

FT-47
TOF(s)
1
%

MT-47
TO(s)
16.67
%

Regarding the failure position of the valve, considering it refers to fuel delivery in a combustion
chamber it is logical to assume that it must be a Fail-Close (FC) valve. This means that a 0% signal
from the respective controller (FC-47) corresponds to a 0% valve opening, hence no flow.
Conversely, a 100% signal from the controller corresponds to a fully open valve, hence maximum
flow.

Regarding the actions of the controllers, we have:

 FC-47: Flow higher than Set-point  Valve opening must reduce  Signal to FC valve must
reduce  Reverse action
 MC-47: Moisture higher than Set-point  Drier temperature must increase  Fuel flow to
heater must increase  Set-point to FC-47 must increase  Direct action

No data are given for the types of the controllers, but given the actions above the gain (Kc) of FC-47
will be positive, while that of MC-47 will be negative. For example for FC-47, if the flow is higher than
the set-point, then EF will be negative. At the same time we saw that the valve must close, so the
controller’s output signal must reduce, meaning that in deviation variables it will also be negative.
Hence, the gain of the controller must be positive, since both its input and output have the same
sign.
(b) This is a three-level cascade control scheme, enhancing the operation of the previous one. The
updated instrumentation diagram is shown below:

Wet paper Paper product

Drier

MT Moist. MC
SP
47 47
Hot
air Tset

TT T TC
47 47

Ambient
air Hot
Heater
gases Fset

vp FIC
47

FT
47 F
Combustion
air Fuel

For the new temperature transmitter again it is assumed that it has very fast dynamics, hence it can
be represented as a zero-order transfer function. It has a measuring range from 120 to 170 oC, so its
gain will be:

100  0  %TO  2 %TO / oC

GTT  s   KTT 
170  120  oC
 
The updated block diagram considering the above is given below:

MC-47 TC-47 FC-47 Valve Heater Drier

R(s) + E(s) MM(s) + ET(s) MT(s) + EF(s) MF(s) 4 F(s) 2 TH(s) 0.5 X(s)
GCM GCT GCF
%TO - % %CO - % %COT - % %COF 0.1s  1 m3/h  2s  1 0.8s  1 o
C 5s  1 % Moist.

FT-47
TOF(s)
1
%

TT-47
TOT(s)
2
%

MT-47
TO(s)
16.67
%

The failure position of the valve, considering it still refers to the same manipulated variable, remains
Fail-Close (FC).
For the actions of the controllers, we now have:

 FC-47: Flow higher than Set-point  Valve opening must reduce  Signal to FC valve must
reduce  Reverse action
 TC-47: Temperature higher than Set-point  Heater temperature must decrease  Fuel flow to
heater must decrease  Set-point to FC-47 must decrease  Reverse action
 MC-47: Moisture higher than Set-point  Drier temperature must increase  Hot air
temperature from heater must increase  Set-point to TC-47 must increase  Direct action

Again, given the actions above the gains of FC-47 and TC-47 will be positive, while that of MC-47 will
be negative.

Problem 2.
Consider the jacketed continuous stirred tank reactor (CSTR) sketched in the figure below. The
following information is obtained from testing the reactor and its control system. The transfer
function of the reactor temperature to the jacket temperature is a first-order lag with a gain of
0.6 oC/oC and a time constant of 13 min. The transfer function of the jacket temperature to the
coolant flow is a first-order lag with a gain of - 2.0 oC/(kg s-1) and a time constant of 2.5 min. The
control valve is linear with constant pressure drop and is sized to pass 12 kg s-1 when fully opened.
Its time constant is negligible. The reactor temperature transmitter is calibrated for a range of 50 to
100 °C and has a time constant of 1 min. The jacket temperature transmitter is calibrated for a range
of 0 to 100 °C, and its time constant is negligible.

(a) Decide on the proper fail position of the control valve and the action of the controller for a simple
feedback control loop with the reactor temperature controller manipulating the position of the
coolant valve. Draw the block diagram showing all transfer functions and write the closed-loop
transfer function of the reactor temperature to its set point. Pay particular attention to the signs,
which must correspond to the fail position of the valve and the controller action.

(b) Write the characteristic equation for the single feedback loop. Using the Routh-Hurwitz stability
criterion examine its stability characteristics and calculate the gain at the critical stability limit.

(c) Design a cascade control system for the reactor temperature with the jacket temperature as the
intermediate process variable, specifying the action of both controllers. Draw the complete block
diagram for the cascade control system showing all transfer functions and their signs.

(d) Assuming a P slave controller with a gain of 2%CO/%TO, write the transfer function for the jacket
temperature loop and redraw the block diagram with the jacket temperature loop as a single block.

(e) Using the simplified block diagram from part (d), write the characteristic equation of the reactor
temperature loop in the cascade control system and use again the Routh-Hurwitz stability criterion
to examine its stability characteristics and calculate the gain at the critical stability limit.

Solutions
(a) Regarding the failure position of the valve, considering it refers to coolant delivery in a reactor,
the safe position to assume is Fail-Open (FO). This means that a 0% signal from the respective
controller (TC) corresponds to a 100% valve opening, hence maximum flow. Conversely, a 100%
signal from the controller corresponds to a fully closed valve, hence no flow.

Regarding the action of the controller, we have:

 Temperature higher than Set-point  Coolant flowrate must increase  Valve opening must
increase  Signal to FO valve must decrease  Reverse action  Gain positive (error negative)

The transfer function of the reactor temperature to the jacket temperature can be represented by a
K
first-order lag process, so the transfer function will follow the generic equation G  s   . In
 s 1
the given data, it can be seen that the gain is 0.6 oC/oC, logically a positive number, since an increase
in the temperature of the coolant (and the jacket) would lead to an increase in the temperature of
the reactor. Hence we have:

T s
GR  s  
KR
 R 
0.6 o o
 R s  1 TJ  s  13s  1

C/ C 
The transfer function of the jacket temperature to the coolant flow can also be represented by a
first-order lag process. In the given data, it can be seen that the gain is - 2.0 oC/(kg s-1), logically a
negative number, since an increase in the flowrate of the coolant would lead to a decrease in the
temperature of the jacket. Hence we have:

T s 2
GJ  s  
KJ
 J 
 J s  1 FC  s  2.5s  1
o

C / kg s 1 
It is also given that the valve is sized to pass 12 kg s-1 when fully open and has a negligible time
constant. This means it can be represented as a zero-order transfer function. The valve, since it is FO,
receives a signal from the controller ranging from 0 to 100% and manipulates the flow from 12 to
0 kg s-1. As such, its gain is equal to:

FC  s 
12  0  kg s 1
GV  s   KV    0.12  kg s 1 / %CO 
M  s   0  100  %CO

The reactor temperature transmitter from the data given can be represented by a first-order lag
process. It has a range from 50 to 100 °, so its gain will be:

 0  100  %TO  2 %TO / oC and its transfer function will be:

K MR 
 50  100  oC
 

TO  s 
GMR 
K MR
 
2
 MR s  1 TR  s  s  1

%TO / oC 
The respective block diagram considering the above is given below:

TC Valve Jacket Reactor

R(s) + E(s) M(s) FC(s) 2 TJ(s) 0.6 TR(s)

GC 0.12
%TO - % %CO kg/s 2.5s  1 o
C 13s  1 o
C

TT
TO(s) 2
% s 1

(b) The characteristic equation of this loop is (following the rule: 1 + “product of blocks in the loop”):

1 + 𝐺𝑐 (𝑠)𝐺𝑣 (𝑠)𝐺𝐽 (𝑠)𝐺𝑅 (𝑠)𝐺𝑀𝑅 (𝑠) = 0

Assuming that the temperature controller is Proportional only, since no data are given, the above
can be expressed in polynomial form to apply the Routh-Hurwitz stability criterion:

−2 0.6 2
1 + 𝐾𝑐 (−0.12) ( ) = 0 ⇒ 32.5𝑠 3 + 48𝑠 2 + 16.5𝑠 + 1 + 0.288𝐾𝑐 = 0
2.5𝑠 + 1 13𝑠 + 1 𝑠 + 1

1 32.5 16.5
2 48 1+0.288Kc
3 A1 0
4 B1 0

At the critical stability limit all elements of row n=3 are equal to zero, which means:

48.5 ∙ 16.5 − 32.5 ∙ (1 + 0.288𝐾𝑐 )

𝐴1 = 0 ⇒ =0⇒
48
 48.5 ∙ 16.5 = 32.5 ∙ (1 + 0.288𝐾𝑐 ) ⇒ 𝐾𝑐 = 81.14 (%𝐶𝑂/%𝑇𝑂)

The gain at the critical stability limit is typically called the ultimate gain (𝐾𝑐𝑢 ).

(c) The updated diagram of implementing the cascade control scheme is shown below:

TR TRset
1
TJset
1 TC
2
2

vp

The jacket temperature transmitter has a range of 0 to 100 °C and its time constant is negligible, so
its gain is equal to:

100  0  %TO  1 %TO / oC

GMJ  s   K MJ 
100  0  oC
 
The updated block diagram of the cascade control scheme is given below:

TC-1 TC-2 Valve Jacket Reactor

R(s) + E1(s) M1(s) + E2(s) M2(s) FC(s) 2 TJ(s) 0.6 TR(s)

GC1 GC2 0.12
%TO - % %CO1 - % %CO2 kg/s 2.5s  1 o
C 13s  1 o
C

TT2
TO2(s)
%
1

TT1
TO1(s) 2
% s 1

The valve obviously remains Fail-Open (FO). Regarding the action of the controllers, we have:

 Jacket temperature controller (TC-2): Temperature TJ higher than Set-point  Coolant flowrate
must increase  Valve opening must increase  Signal to FO valve must decrease  Reverse
action  Gain positive
 Reactor temperature controller (TC-1): Temperature TR higher than Set-point  Coolant
temperature must decrease  Set-point to TC-2 must decrease  Reverse action  Gain
positive
 (d) The transfer function for the jacket temperature loop can be found with normal procedures
(following the rules: numerator = “product of blocks between input and output”, denominator = 1 +
“product of blocks in the loop”), where the secondary loop is treated as an independent system
(highlighted below):

TC-1 TC-2 Valve Jacket Reactor

R(s) + E1(s) M1(s) + E2(s) M2(s) FC(s) 2 TJ(s) 0.6 TR(s)

GC1 GC2 0.12
%TO - % %CO1 - % %CO2 kg/s 2.5s  1 o
C 13s  1 o
C

TT2
TO2(s)
%
1

TT1
TO1(s) 2
% s 1

−2
2(−0.12) ( ) 0.48 0.324
𝑇𝐽 (𝑠) = 2.5𝑠 +1 𝑀1 (𝑠) = 𝑀1 (𝑠) = 𝑀 (𝑠)
1 + 2(−0.12) (
−2
)1 2.5𝑠 + 1 + 0.48 1.69𝑠 + 1 1
2.5𝑠 + 1
TC-1 TC-2 TC-1
Valve TC-2 Valve TC-2
TC-1
Jacket Jacket
Valve
Reactor TC-2 Reactor
Jacket
Valve Reactor
Jacket Reactor
So the block diagram can be redrawn as follows:
) E
+1(s) +1(s) +M
E2(s) R(s) M E21(s) E2(s)R(s) + M1(s)
FEM+ (s) E2(s)2 M1(s) F
(s) T (s) 0.62
0.324
M2(s)E2(s)
+JC(s) TRT
FJ(s)
(s)
M (s)
2(s) 0.6 2 TCR(s)
F TJ(s) 2
0.6 TTJ(s)
R(s) 0.6
G0.12 GC 0.12 GC2 o 0.12 0.12
C
GC GC2 GC2 C1 2 GC2
1 %
- %TO
%TC-11
%CO - % 2 % TC-1
- 1 %CO
TC-2 %TO
Valve TC-2
C
- %COkg/s
%CO
1%
1
- 2 %2.5 s  %CO
Valve
TC-1
Jacket
1
-C%CO21.69
1TC-21 kg/s %132.5
sJacket
Reactor TC-2
s11
sValve 1 Ckg/s
%COC 2 o o
13  1s  1
s2.5
Reactor
Jacket
Valve
o o
CC
kg/s 13
2.5ss 11
Reactor
Jacket
oo
CC 13 s 1
Reactor
Reduced secondary
) E
+1(s) +1(s) +M
E2(s) R(s) M E21(s) E2(s)R(s)
TT2+ M1(s)
FEM + (s) E2(s)TT
(s) 2 2 M1(s) F
T (s) 0.6loop
0.324
M (s)E2(s)
+ (s) 2 2 TTFJ(s)
(s)
M (s)
C(s) TT22
0.6 TCR(s)
F TJ(s) 2
0.6 TTJ(s)
R(s) 0.6
GC1 GC2 G0.12 GC2 C1 2 GC10.12 GC2 oJC 2 cascade 0.12
TT
GC2 oR o 2 0.12
1 %
- % %TO%CO - 1TO -2%
%CO(s) 2 % %TO
C1
- %CO 1%CO
%
kg/s -
TO2(s) 2 %2.5 s  1
%CO 1 kg/s
-C%CO
TO2(s) 1.69
2%132.5
s s s1
1 1 C kg/s
%COC
TO2(s) 2  1s  1
13s2.5 o o
CC
kg/s 13ss 11
2.5 oo
CC 13s  1
1 %
1 %
Reduced secondary1 1
% %
TT2 TT2 cascade loop TT2 TT2
TO2(s) TO2(s) TO2(s) TO2(s)
TT11 TT11 TT11 TT11
% % % %
TO1(s) 2 TO1(s) 2 TO1(s) 2 TO1(s) 2
% s 1 % s 1 % s 1 % s 1
TT1 TT1 TT1 TT1
TO1(s) 2 TO1(s) 2 TO1(s) 2 TO1(s) 2
(e) Assuming
%
again
s 1
a P temperature
%
controller,%the characteristic
s 1 s 1 %
equations of
1
this loop is now:

0.324 0.6 2
1 + 𝐾𝑐 = 0 ⇒ 21.97𝑠 3 + 36.66𝑠 2 + 15.69𝑠 + 1 + 0.39𝐾𝑐 = 0
1.69𝑠 + 1 13𝑠 + 1 𝑠 + 1

Applying the Routh-Hurwitz stability criterion:

1 21.97 15.69
2 36.66 1+0.39Kc
3 A1 0
4 B1 0
At the critical stability limit all elements of row n=3 are equal to zero, which means:

36.66 ∙ 15.69 − 21.97 ∙ (1 + 0.39𝐾𝑐 )

𝐴1 = 0 ⇒ =0⇒
36.66

36.66 ∙ 15.69 − 21.97 ∙ (1 + 0.39𝐾𝑐 ) ⇒ 𝐾𝑐 = 64.57 (%𝐶𝑂/%𝑇𝑂)

The cascade system is actually less stable, since it ultimate gain is smaller than (b). This effect
however is correlated to the probably poor choice of the value of the gain of the inner loop.