Appa Corrected
Appa Corrected
Objectives
The objective of this appendix is to review the basic concepts usually covered in an under-
graduate course in process control. Concepts reviewed include:
503
504 Review Of Basic Concepts Appendix A
The block diagram notation is extensively used to describe and analyze control systems. The
simplest form is shown in Figure A.1. The diagram shows the cause (a change in the input
variable) and effect (a change in the output variable) relationship between the input and out-
put variables. We define input and output variables as deviations from an initial nominal
steady-state of the system. The box represents the system and the relationship between the
input and output. Arrows connecting blocks represent information flow. The block diagram
notation is used to represent the control strategy implemented in a process.
Temperature
Temperature
Transmitter
Controller
TT TC
Remotely
101 101 Located in
Electronic
Control Room
Tranmission Line
4-20 mA current
Feed, In
FE
101
Pneumatic Line
3-15 psig
Automatic Control
Valve
Cooling Water, In
Manual Valve
Reactor Product, Out
ACTUATOR PROCESS
ERROR
TEMPERATURE
SETPOINT CONTROLLER SIGNAL TO CONTROL TANK
Σ VALVE
Tset + VALVE
- WATER
FLOW
Tm (MEASURED VARIABLE)
TRANSMITTER
P: Pressure
T: Temperature
F: Flow
C: Concentration
L: Level
D: Density
A: Analyzer
I: Indicator
H: Hand-operated
V: Valve
C: Controller
A.2 Laplace Transform and Transfer Functions 507
The Laplace transform is often used to represent the dynamic relationship contained in a
block diagram. It is defined as follows: Given a function of time f(t), t ≥ 0, we can define a
new function F (s ) as
∞
F ( s ) = ∫ f (t )e − st dt. (A.1)
o
This represents a unique one-to-one mapping. F (s) is called the Laplace transform of
f(t), and f(t) is the inverse Laplace transform of F (s). This is represented using the notation
L[ f (t )] = F ( s), (A.2a)
L−1[F ( s )] = f (t ). (A.2b)
s is called the Laplace variable. Table A.2 gives the Laplace transform of some com-
mon functions. Some useful properties of the Laplace transform are given in Table A.3.
Property Description
Linearity Property L[ f1 (t ) + f 2 (t )] = L[ f1 (t )] + L[ f 2 (t )]
L[k f1 (t )] = k L[ f1 (t )]
where k = constant
Time delay L[ f (t − θ )] = e −θs L[ f (t )]
= e −θs F ( s )
Differentiation d f (t )
L[] = s F ( s ) − f (0)
dt
Integration t 1
L {∫ f (t )dt} = F ( s)
o s
Final value theorem lim f (t ) = lim s F ( s ) provided LHS limit exists.
t →∞ s →0
where u(t) is the input variable and x(t) is the output variable. Notice that we have
taken the initial condition to be zero, since both x and u are taken as deviations from
the nominal initial steady-state values. Taking Laplace transform of both sides (we
will informally use x(s) to denote Laplace transform of x(t)),
Since x(0) = 0 ,
b1
x( s) = ( )u ( s )
a1 s + a 2 (A.4)
K
= ⋅ u ( s ),
τs + 1
where K = b1 a2 is called the gain of the system and τ = a1 a 2 is called the time constant
of the system. The function
K
G (s) = (A.5)
τs + 1
A.2 Laplace Transform and Transfer Functions 509
is called the transfer function relating x(s) and u(s). The transfer function captures the dy-
namic relationship between the input and output variable in a compact, convenient, and alge-
braic (as opposed to the differential equation) form. G (s ) is an example of a first-order
transfer function, since the denominator is first-order in s. Other commonly encountered
transfer functions are given below. In general, G (s ) can be expressed as a ratio of
polynomials:
N (s)
G (s) = , (A.6a)
D( s )
x( s ) = G ( s )u ( s ). (A.6b)
If the transfer function given by Eq. (A.6a) is stable, then the gain of the transfer function is
defined as the steady-state change in x(t) for a unit step change in u(t), where x(t) and u(t) are
the inverse Laplace transforms of x(s) and u(s). The final value theorem applied to Eq.
(A.6b) shows that, by the foregoing definition, the gain of G(s) is G(0).
d 2x dx
τ2 2
+ 2τξ + x = Ku , x(0) = 0, x' (0) = 0 (A.7a)
dt dt
can be represented using the transfer function
κ
x( s) = ⋅ u ( s ), (A.7b)
τ 2 s 2 + 2τξs + 1
where τ = time constant and ξ = damping coefficient. If we can factor the denominator
τ 2 s 2 + 2τξ s + 1 = (τ 1 s + 1) (τ 2 s + 1),
where τ 1 and τ 2 are both real (this can be done if ξ > 1 ), then we may write
K 1
x( s ) = ⋅( ) u ( s ). (A.8)
(τ 1 s + 1) τ 2 s + 1
That is, the second-order system can be viewed as two first-order systems in series. Alterna-
tively, we can think of the second-order differential equation being written as two equivalent,
coupled, first-order differential equations.
510 Review Of Basic Concepts Appendix A
3. Lead-Lag
(τ1 s + 1)
x( s) = K u (s) (A.10)
(τ 2 s + 1)
Ke −θ s u ( s )
x( s ) = . (A.11)
τ s +1
This is perhaps the most frequently encountered transfer function in process control. Many
process systems are approximated using this transfer function. Note the abbreviation
FOPDT, used frequently in this text.
In this section we will discuss the transfer functions associated with commonly used control-
lers. These transfer functions have some parameters associated with them, which must be
determined to yield good control system performance. The latter process of setting these
controller constants is called controller tuning.
In the Laplace domain (we omit the bias term since we are dealing with deviation
variables),
m( s) = K c ⋅ e( s ), (A.13)
A.3 P, PI, and PID Controller Transfer Functions 511
where m is the controller output and e(t) is the error in the control defined by
e(t ) = setpoint − measurement. (A.14)
K c is the controller gain. Note the bias term m (the steady-state controller output). This is
set by the operator or initialized to the current output value when the controller is turned on.
One difficulty with this controller is the choice of m . Since this is not known a priori, it is
usually not possible to drive e(t ) to zero to eliminate steady-state offset in the controlled
variable.
The integral or reset action allows the controller to eliminate steady-state offset. There is no
need for a bias term, as the controller output will change as needed to drive e (t ) to zero. The
controller transfer function is given by
1
m( s) = K c [1 + ]e( s ). (A.16)
τIs
τ I is called the integral time constant and determines the weight given to the integral action
relative to the proportional part.
3. Proportional-Integral-Derivative (PID) Control
The ideal form of the three-mode PID controller is
1 de
m(t ) = K c [e(t ) +
τI ∫ e dt + τ D
dt
]. (A.17)
This adds another parameter τ D , called the derivative time constant. The transfer function of
the ideal PID controller is given by
1
m( s ) = Kc (1 + + τ D s )e( s ). (A.18a)
τI s
However, because of the difficulty with differentiating noisy signals, actual industrial con-
troller implementations vary from the ideal form. This is done by adding a noise filter (typi-
cally a first-order lag transfer function) before derivative action is applied. A typical imple-
mentation of a real PID controller would be as follows:
1 τ Ds
m( s ) = K c (1 + + )e( s), (A.18b)
τ I s ατ D s + 1
512 Review Of Basic Concepts Appendix A
where α is typically chosen as a number between .05 and 0.2, depending on the manufacturer
of the controller. In some digital control systems the user has control over the value of α.
Larger values of α lead to stronger filtering of the derivative action of the controller.
Another form of PID controller still found in older industrial applications is the industrial
form of a PID controller, so called because prior to digital control computers it was the most
common form of PID controller. The industrial form of a PID controller is given by
1 (τ D* s + 1)
m( s ) = K c (1 + ) )e ( s ) (A.18c)
τ I*s (ατ D* s + 1)
The controller given by Eq. (A.18c) is not as general as that given by Eq. (A.18b) since the
zeros of Eq. (A.18c) are necessarily real, while those of Eq. (A.18b) may be either real or
complex. In the case where the zeros of Eq. (A.18b) are real, then the relationship between
the parameters of equations (A.18b) and (A.18c) are
τ I τ D = τ I*τ *D , (A.18d)
τ I = τ I* + τ *D . (A.18e)
From the equation series (A.18), changing either τ I* or τ D* changes both τ I and τ D . There-
fore, the controller given by Eq. (A.18e) is often called an interacting form of the PID con-
troller.
There are many other versions for the real PID controllers, but we will not discuss
them here. All of the previously mentioned controllers focus on controlling a single output
variable using a single manipulated input variable. Hence they are known as SISO control
systems.
The addition of a feedback controller to a process can alter the stability characteristics of the
system. We define a system as stable if the output of the system is bounded (finite) for
bounded inputs. Since stability is an important characteristic of control systems, we need
some tools to analyze the stability of control systems and to evaluate the effect of controller
parameters on stability.
Consider a system containing the transfer function
A.4 Stability of Systems 513
K
y (s) = x( s ). (A.19)
( s − p)
If the system is forced by an impulse input given by the Dirac Delta function
x(t ) = δ (t ), x( s ) = 1, (A.20)
then
K
y (s) = ,
s− p
y (t ) = Ke pt . (A.21)
Let us consider the behavior of y(t) as a function of p, the pole of the system. In general, p
can be a complex number given by (see Fig. A.4)
p = a + ib,
p = a + ib
= re iφ
r = a2 + b2
r
b
φ
a
Real Axis
Complex poles always occur in conjugate pairs in a real system. The response y(t) is given by
y = Ke pt ,
(A.22)
= Ke at (cos bt + i sin bt ).
514 Review Of Basic Concepts Appendix A
The imaginary part of the output will be cancelled by the conjugate pole. Let us examine the
asymptotic behavior of y (t ) as t → ∞ .
Systems whose outputs grow unbounded as t → ∞ for any bounded input-forcing function
are called unstable. A necessary condition for stability is that all the poles of the system have
negative real parts. A typical transfer function can be expressed as a ratio of polynomials,
N ( s)
G ( s) = , (A.23)
D( s)
D( s) = ( s − p1 )( s − p 2 )...( s − p n ), (A.24)
where p1 , p 2 ,..., p n are the roots of the polynomial D(s). These are called the poles of the
system. We can expand G(s) in a partial fraction expansion as
N (s)
G (s) = ,
D( s )
N (s)
= , (A.25)
( s − p1 )(s − p 2 ),.......( s − p n )
K1 K2 Kn
= + + ....... + ,
s − p1 s − p 2 s − pn
provided p1 , p 2 , and pn are all unique. Otherwise, the expansion must be modified slightly.
If this process is perturbed by an impulse input, then
y ( s ) = G ( s ) ⋅ x( s ),
= G ( s ),
K1 Kn
= + .... + .
s − p1 s − pn
Inverting,
Thus the behavior of y (t ) is characterized by its poles. If any pole has a real part that is posi-
tive, the output will grow unbounded. This leads to the following conclusion:
For a transfer function to be stable, all its poles must lie to the left of the imaginary
axis in the complex plane; that is, in the LHP.
Figure A.5 shows the behavior of the solution at different pole locations in the complex
plane. Note that all poles on the RHP lead to exponential growth. If the pole has an imagi-
nary component, we get oscillations. A pole on the LHP leads to exponential decay (with
oscillations if the pole has an imaginary component, we get oscillatory decay). A pole on the
imaginary axis leads to sustained oscillatory response (except at s = 0).
Imaginary Axis
⊗ Sustained Oscillations
⊗
⊗ Oscillatory
Growth
⊗ ⊗
Exponential Decay
Sustained Oscillations
⊗ ⊗
Oscillatory Decay ⊗
Oscillatory
Growth
In order to analyze the stability of closed-loop systems, we need to develop the transfer func-
tions associated with the closed-loop system. Block diagrams offer a convenient way to do
this. The block diagram of a feedback control, shown in Figure A.6 shows four blocks. Let us
define the following transfer functions:
y( s)
Process: G p (s) = . (A.27)
q(s)
y m (s)
Transmitter: GT ( s ) = . (A.28)
y( s)
q( s)
Actuator: Gv (s) = . (A.29)
m( s )
m( s )
Controller: Gc (s) = . (A.30)
e( s )
Measurement
Transmitter
ym ( s )
GT ( s )
Solving for y,
A.5 Stability of Closed-Loop Systems 517
G p Gv Gc
y= yset , (A.31)
1 + G p Gv Gc GT
= G CL ( s ) ⋅ y set , (A.32)
where GCL (s ) is called the closed-loop transfer function. Note the following:
If there is a load or disturbance variable that can also affect the output, as shown in Figure
A.7 the transfer function will be
Gd ( s )
y = d ( s ). (A.34)
1 + G p Gv Gc GT
d(s)
Disturbance G d (s )
y set
G p (s )
+
Σ Gc (s ) Gv (s ) +
Σ y
+ -
GT (s )
y K c K p (τ 1 s + 1)
GCL ( s ) = = , (A.35)
y set (τ 1τs 2 + τ 1 (1 + K c K p ) s + K c K p
with a steady-state gain GCL (0) = 1. Hence no steady-state error (offset) will be present for
a step change in the setpoint.
♦
y Kc
= . (A.37)
y set ( s + 1)(s + 2)( s + 3) + K c
It is interesting to study the effect of changing K c on the poles of the closed-loop sys-
tem. Table A.4 shows how the zeros of the characteristic equation (system poles) change with
the controller gain K c . These poles are plotted in Figure A.8 As K c changes, these poles
trace a path on the complex plane to generate the root-locus.
A.5 Stability of Closed-Loop Systems 519
Since the dynamic behavior of the system is characterized by its poles, we can con-
clude the following:
1. The system response is nonoscillatory for values of K c < 0.39, since the roots
are all real.
2. As K c approaches 60, the poles moves closer to the RHP, and at K c = 60, the
system has two poles on the imaginary axis leading to sustained oscillatory
response.
3. The system has two poles on the RHP for K c > 60 and is hence unstable for these
controller gains.
K c = 60 is called the stability limit of the system, also called the ultimate gain, K u . The
frequency of oscillation at this gain is called the ultimate frequency. From the calculations
shown in Table A.4, the ultimate frequency is given by the pole on the imaginary axis
(ω u = 3.32 sec) . The ultimate period of oscillation is Pu = 2π ω u .
6
4
Imaginary axis
-2
-4
-6
-6 -4 -2 0 2 4 6
Real axis
PID controllers are used in the majority of SISO feedback controllers because of their effec-
tiveness and simplicity. The transfer function of a typical PID controller is given by
1 1
m( s ) = K c (1 + + τ D s)e( s). (A.39)
τIs (αs + 1)
A number of authors have published empirical correlations for tuning PID controllers based
on the selected performance criteria and assumed process model structure. The FOPDT
model is used most often. Generally, correlations are provided in terms of the process gain,
the time constant, and time delay in the FOPDT model. Separate correlations are provided
for servo control and regulatory control. Different criteria are used to evaluate the perform-
ance, including
1. Integral Squared Error (ISE)
∞
ISE = ∫
0
e 2 dt (A.40)
A.6 Controller Tuning 521
This criterion is widely used in optimal control theory. It is also the criterion of choice
in many industrial implementations of model-predictive control (MPC).
2. Integral Absolute Error (IAE)
∞
IAE = ∫o e dt (A.41)
Ke −θ s
G (s) =
τs + 1 (A.43)
the Ziegler-Nichols tuning constants are listed in Table A.5.
Kc τ1 τD
1 τ
P
K θ
0.9 τ
PI 3.33θ
K θ
1.2 τ
PID 2.00θ 0.5θ
K θ
Kc τI τD
1 τ 1 θ
P 1 +
K θ 3 τ
1 τ 1 θ 30 + 3 θ τ
PI 0.9 + θ
K θ 12 τ 9 + 20 θ τ
1 τ 4 1 θ 32 + 6 θ τ 4
PID + θ θ
K θ 3 4 τ 13 + 8 θ τ 11 + 2 θ τ
Typically the tuning methods presented in this appendix attempt to optimize the closed-
loop response to a unit step change in setpoint (see Figure A.9). This response is very similar
to the unit step response of a second-order underdamped system modeled by the transfer
function
1
G( s) = 2 2
. (A.44)
τ s + 2τξs + 1
For such a system,
1− ξ2
Overshoot = e −πξ
1 −ξ2
Decay Ratio = e − 2πξ (A.45)
Period = 2πτ 1 − ξ 2
Hence τ tells us how fast the process responds and ξ tells us how fast the oscillations decay
and how large the overshoot will be.
A.6 Controller Tuning 523
Rise time, T r
1.4 Overshoot, A
1.2 B
0.8
0.6
e(t), Error in control ISE = ∫ e 2 dt
0.4
Decay Ratio = B / A
0.2
0
0 1 2 3 4 5 6
+ 1
Σ Kc(1+ ) K/ s y
τIs
-
y τ I s +1
= 2
. (A.47)
yset (τ I s / KK c ) + τ I s + 1
ξ = 1 2 KK cτ I . (A.48)
For values of ξ < 1 (i.e., small values of K c and τ 1 ) it will be oscillating. Typically, we want
to keep ξ close to 0.5 (to get about 20% overshoot). This will yield
KK cτ I = 1. (A.49)
A reasonable choice for K c is 5 %/% where the percentages are taken with respect to the
span of the variables
5 % change in controller output
Kc =
1% change in measurement
(A.50)
controller output span
= 5×
measurement span
τ I = 1/( KK c ). (A.51)
A.7 Regulatory Issues Introduced by Constraints 525
In this section we examine regulatory issues that can be derived from the process model. The
focus is on steady-state suppression of disturbances for SISO systems. We examine the dy-
namic suppression issues in Appendix B. For convenience in the analysis we consider first
the representation of models using dimensionless variables.
where y max − y min is the span of the measuring instrument (a field adjustable quantity). The
smaller the span, the greater the accuracy of the instrument. During startup and shutdown
phases, a greater range of the measurement is expected and the span may be kept large.
ymeas is the normal steady-state value of the variable. Ideally, this should be set to the mid-
dle of the span to yield a range of
− 0 .5 ≤ y ≤ 0 .5 .
A similar definition is used to scale manipulated input variables. With control valves,
the nominal steady-state valve is kept closer to 1 (say in the 65% to 75% open range) to re-
duce pressure losses.
Consider SISO system, as shown in Figure A.11. The transfer function model is
y ( s ) = p ( s )u ( s ) + d ( s ) . (A.53)
The control objective is to maintain y close to its nominal steady-state value of 0 in presence
of disturbances. First we consider the steady-state case. The problem reduces to
y ( s ) = p(o)u (o) + d (o). (A.54)
526 Review Of Basic Concepts Appendix A
d ( s)
p ( s) y ( s)
u ( s)
Slope = p(0)
y max
Controllable
u min u u max Range of y
y min
Consider the column for separating a mixture of methanol/water. The nominal steady-state
operating conditions are shown. Given the following steady-state observations, determine the
steady-state regulation of the bottom product purity using the reboiler heat duty and the dis-
tillate flow rate when the feed can vary from 95 to 105 lb moles/hr.
Solution
xT k11 k12 d k f 1
x = k k q + k f ,
B 21 22 f 2
xT = ( X T − X T ) (0.02) , xB = ( X B − X B ) 0.2 , q = (Q − Q ) 4 ,
d = ( D − D ) 100 , f = F 200 .
yielding
3.0 − 2 0.4
K = , k f = .
8.05 − .14 6.74
528 Review Of Basic Concepts Appendix A
PROBLEMS
A.1 Control system design and block diagram construction. Consider the control of pH in a waste-
water treatment facility. Contaminated water from the plant flows into a large pH stabilization tank
where it is mixed with caustic solution (NaOH solution) to neutralize the water and bring its pH to 7.
The caustic flows into the stabilization tank from another storage tank. The treated water from the
stabilizer tank then flows into a lake about 1000 yards away.
Neutralizing Caustic
Solution
pH Stabilization
Waste Water Stream Tank
a. Design a feedback control system to control the pH of the water being discharged to the lake.
Show the process instrumentation diagram using standard symbols. Identify and label all
sensors, transmitters and control valves.
b. Draw a block diagram of the pH control system, identifying all variables used.
c. Discuss briefly the pros and cons of the following options from the point of view of the per-
formance of the control system. (i) Place the pH sensor in the pipe through which treated
water is leaving the stabilization tank. Sensor is located as close to the tank as possible.(ii)
Place the pH sensor at the point where the wastewater enters the lake. (iii) Place the pH sen-
sor inside the stabilization tank. (iv) Place the pH sensor in the pipe that brings the water
into the stabilization tank.
A.2 Feedback versus feedforward control. A process fluid is heated in a furnace, as shown in the
Figure. The available instrumentation on the system is also shown. It is desired to control the tempera-
ture of the process fluid at the exit at a given setpoint by manipulating the fuel input flow into the fur-
nace.
a. Show the instrumentation diagram of a feedback control strategy to achieve the ob-
jective stated. Label and clearly identify the instrumentation used.
b. Show the instrumentation diagram of a feedforward controller that will be used in
conjunction with the feedback controller of Part a. This feedforward controller
should compensate for variations in the input disturbance, namely, the process
fluid flow.
c. Draw a block diagram of the feedforward/feedback control system. Label and iden-
tify all blocks and arrows used in the block diagram.
d. It is important to maintain a constant ratio of air to fuel flow to the furnace. Show
the instrumentation necessary (transmitters, controllers, etc.) to maintain this ratio
constant.
Flue Gas
FT
TT
Air Fuel
A.3 Step response of a first-order system. You have been assigned the task of measuring the time
constant of a temperature sensor being used in a control loop. The following experiment was per-
530 Review Of Basic Concepts Appendix A
formed. The sensor was initially at room temperature of 25 °C. It was suddenly inserted into hot water
in a cup. You read the temperature as 40 °C after 1 minute. The reading reached a steady-state value of
45 C after a few minutes. Assume the sensor can be modeled as a first-order system.
a. Estimate the time constant of the temperature sensor.
b. Write the transfer function for the temperature sensor. Use your result from Part a. Show
that the gain is 1.
c. If the temperature can be read accurately to ± 1 °C on the sensor, what is the accuracy of
your time constant estimate in minutes?
A.4 Step response of a first-order system. A bare thermocouple, when suddenly moved from a steady
temperature of 80°F into a stream of hot air at a constant temperature of 100°F, reads 90°F after 1 sec.
Assume the thermocouple responds as a first-order system.
a. Sketch the behavior of the temperature surrounding the thermocouple as a function of time.
b. Sketch the expected behavior of the measured temperature as a function of time (qualita-
tively).
c. What will be the final (steady-state) temperature measured by the thermocouple?
d. Compute the measured temperature at 2 sec.
A.5 Pulse response of a first-order system. The time-dependent behavior of the outlet concentration
of a stirred tank to changes in inlet concentration is given by
dy
τ + y = x,
dt
where y = outlet tank concentration, x = inlet concentration t = time in minutes.
Suppose at t = 0, y = 0. The inlet concentration is suddenly increased from 0 to 1 and maintained at 1
for the duration of 1 minute, at which time it is brought back to zero. This is called a unit pulse input.
a. Sketch x(t) as a function of time.
b. Write an equation for x(t) in terms of unit step functions.
c. What is the Laplace transform of x(t)?
d. What is the Laplace transform of y(t)?
e. Solve for y(t) by inverting the result of Part d.
f. If τ = 1, sketch the response y(t).
A.6 Inverting Laplace transforms. Sketch the behavior of f(t) if its Laplace transform is given by
1 e − s − e −2 s e−3s
F (s) = + − .
s s2 s
A.7 Using dynamic response characteristics in design. A holding tank of uniform cross section has
been designed to smooth out the flow from a reactor to a purification column. The reactor has to be
shut down periodically for cleaning. Given the following data, estimate the maximum rate of change of
Problems 531
the flow into the column. Assume flow out of the holding tank is proportional to the height of liquid in
the holding tank.
Data: The tank has a cross-sectional area of 10 ft2 and a height of 10 ft. Normal (steady-state level) in
the tank is 5 ft. The normal flow from the reactor is 5 ft3/hr and the reactor is periodically shut down
for 2 hours.
Hint: First write the model of the holding tank. Then derive a transfer function relating changes in inlet
flow to the outlet flow. It is preferable to use deviation variables. This is a first-order transfer function.
Shutting down the reactor reduces the inlet flow to 0 suddenly. Hence it is a step change. The outlet
flow responds to this step change. Sketch its response. When is the slope of this curve a maximum?
A.8 Closed-loop stability. An open-loop unstable process can be represented by the following trans-
fer function:
K
G (s) = .
(4 s − 1)(τ v s + 1)(τ T s + 1)
where τ v and τ T are, respectively, the time constants of the control valve and the transmitter. Assum-
ing proportional control with gain K c is used, find the range of the loop gain KK c for which the
loop is stable if
A.9 Response of second-order systems. The temperature of a reactor responds to the feed variations
as a second-order system. Let
(15 F gpm)
T (s) F ( s) = .
4s 2 + 2s + 2
The reactor is operating normally at a feed rate of 70 gpm and at 300°F. Previous experience shows
that if the reactor temperature exceeds 350°F, the catalyst will be deactivated. The feed is subject to
sudden step changes. What is the maximum permissible step change in the feed rate, that will keep the
reactor temperature below its specified maximum of 350°F?
A.10 Control valve design. Control valves play an important role in the performance of the control
system. This problem considers a typical case where a control valve is installed in line with a process
unit, in this case a heat exchanger. The working design equation for the control valve is
532 Review Of Basic Concepts Appendix A
∆Pv
q = C v f (l )
gs
f (l ) = R l −1 , R = 20
l = valve position (0 − 1)
Cv = valve size
∆Pv = Pressure drop across valve, psi
gs = specific gravity
q = gal/min( gpm)
This is called an equal percentage valve, due to the nature of the dependence of f(l) (the area available
for flow) on the valve position.
a. Design a control valve (i.e., compute Cv) for a fluid with gs = 0.8. The normal design flow
is 250 gpm. The normal pressure drop across the valve is 10 psi and the normal pressure
drop across the line is 10 psi. The total pressure drop (pressure drop across valve + pres-
sure drop across line) is always constant at 20 psi. The pressure drop across the line is
variable and proportional to the square of the flow rate. The valve should be designed so
that it is half open (l = 0.5) at the design flow.
b. What will be the flow through the valve when the valve is fully open? What is the pressure
drop across the valve at this flow rate? (Hint: It is not 500 gpm.)
∆Pv ∆Pl
∆Pv
q = Cv f (l ) ,
gs
f (l ) = l ,
l = valve position (0-1).
The valve size C according to the specs, is 100. Valve is air-to-open (as the signal to the valve in-
creases, the valve opens more).
a. What is the maximum achievable flow through the cooling water valve? Assume that the
pressure drop across the heat exchanger is proportional to the square of the flow rate. The
total pressure drop (the sum of the pressure drops across the valve and the heat exchanger)
is a constant. (See figure in Problem A.10.)
b. Based on your calculations, could the valve be the cause of the problem with the control
loop? Explain why or why not. If the valve were the cause of the problem, what remedial
action would you recommend?
References
Cohen G. H., and G. A. Coon. 1953. “Theoretical Considerations of Retarded Control.” Trans. ASME,
75, 827.
Luyben, W. L. 1990. Process Modeling, Simulation and Control for Chemical Engineers. 2nd Ed.
McGraw-Hill, NY.
Marlin T. E. 1995. Process Control: Designing Processes and Control Systems for Dynamic Per-
formance. McGraw-Hill, NY.
Ogunnaike, B. A., and W. H. Ray. 1994. Process Dynamics, Modeling and Control. Oxford University
Press, NY.
Seborg, D. E., T. F. Edgar, and D.A. Mellichamp. 1989. Process Dynamics and Control. John Wiley
& Sons, NY.
Smith, C. A., and A. B. Corripio. 1985. Principles and Practice of Automatic Process Control. John
Wiley & Sons, NY.
Ziegler J. G., and N. B. Nichols. 1942. “Optimum Settings for Automatic Controllers” Trans. ASME,
64, 759.
534 Review Of Basic Concepts Appendix A