Handbook Section 8 1132454
Handbook Section 8 1132454
https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374
https://ebooknice.com/product/matematik-5000-kurs-2c-larobok-23848312
https://ebooknice.com/product/sat-ii-success-
math-1c-and-2c-2002-peterson-s-sat-ii-success-1722018
(Ebook) Master SAT II Math 1c and 2c 4th ed (Arco Master the SAT
Subject Test: Math Levels 1 & 2) by Arco ISBN 9780768923049,
0768923042
https://ebooknice.com/product/master-sat-ii-math-1c-and-2c-4th-ed-
arco-master-the-sat-subject-test-math-levels-1-2-2326094
(Ebook) Cambridge IGCSE and O Level History Workbook 2C - Depth Study:
the United States, 1919-41 2nd Edition by Benjamin Harrison ISBN
9781398375147, 9781398375048, 1398375144, 1398375047
https://ebooknice.com/product/cambridge-igcse-and-o-level-history-
workbook-2c-depth-study-the-united-states-1919-41-2nd-edition-53538044
https://ebooknice.com/product/process-dynamics-and-control-2nd-
edition-2031102
https://ebooknice.com/product/process-dynamics-and-control-6821516
https://ebooknice.com/product/perry-s-chemical-engineers-handbook-
section-13-1132436
Copyright © 2008, 1997, 1984, 1973, 1963, 1950, 1941, 1934 by The McGraw-Hill Companies, Inc. All rights reserved. Manufactured in the United
States of America. Except as permitted under the United States Copyright Act of 1976, no part of this publication may be reproduced or distributed
in any form or by any means, or stored in a database or retrieval system, without the prior written permission of the publisher.
0-07-154215-9
The material in this eBook also appears in the print version of this title: 0-07-151131-8.
All trademarks are trademarks of their respective owners. Rather than put a trademark symbol after every occurrence of a trademarked name, we use
names in an editorial fashion only, and to the benefit of the trademark owner, with no intention of infringement of the trademark. Where such
designations appear in this book, they have been printed with initial caps.
McGraw-Hill eBooks are available at special quantity discounts to use as premiums and sales promotions, or for use in corporate training programs.
For more information, please contact George Hoare, Special Sales, at [email protected] or (212) 904-4069.
TERMS OF USE
This is a copyrighted work and The McGraw-Hill Companies, Inc. (“McGraw-Hill”) and its licensors reserve all rights in and to the work. Use of this
work is subject to these terms. Except as permitted under the Copyright Act of 1976 and the right to store and retrieve one copy of the work, you may
not decompile, disassemble, reverse engineer, reproduce, modify, create derivative works based upon, transmit, distribute, disseminate, sell, publish
or sublicense the work or any part of it without McGraw-Hill’s prior consent. You may use the work for your own noncommercial and personal use;
any other use of the work is strictly prohibited. Your right to use the work may be terminated if you fail to comply with these terms.
THE WORK IS PROVIDED “AS IS.” McGRAW-HILL AND ITS LICENSORS MAKE NO GUARANTEES OR WARRANTIES AS TO THE
ACCURACY, ADEQUACY OR COMPLETENESS OF OR RESULTS TO BE OBTAINED FROM USING THE WORK, INCLUDING ANY
INFORMATION THAT CAN BE ACCESSED THROUGH THE WORK VIA HYPERLINK OR OTHERWISE, AND EXPRESSLY DISCLAIM
ANY WARRANTY, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF MERCHANTABILITY OR
FITNESS FOR A PARTICULAR PURPOSE. McGraw-Hill and its licensors do not warrant or guarantee that the functions contained in the work will
meet your requirements or that its operation will be uninterrupted or error free. Neither McGraw-Hill nor its licensors shall be liable to you or
anyone else for any inaccuracy, error or omission, regardless of cause, in the work or for any damages resulting therefrom. McGraw-Hill has no
responsibility for the content of any information accessed through the work. Under no circumstances shall McGraw-Hill and/or its licensors be liable
for any indirect, incidental, special, punitive, consequential or similar damages that result from the use of or inability to use the work, even if any of
them has been advised of the possibility of such damages. This limitation of liability shall apply to any claim or cause whatsoever whether such claim
or cause arises in contract, tort or otherwise.
DOI: 10.1036/0071511318
This page intentionally left blank
Section 8
Process Control
Cecil L. Smith, Ph.D. Principal, Cecil L. Smith Inc. (Batch Process Control, Telemetering
and Transmission, Digital Technology for Process Control, Process Control and Plant Safety)
8-1
Copyright © 2008, 1997, 1984, 1973, 1963, 1950, 1941, 1934 by The McGraw-Hill Companies, Inc. Click here for terms of use.
8-2 PROCESS CONTROL
Nomenclature
Symbol Definition Symbol Definition
A Area s Laplace transform variable
Aa Actuator area s Search direction
Ac Output amplitude limits Si Step response coefficient
Av Amplitude of controlled variable t Time
Al Cross-sectional area of tank T Temperature, target
b Controller output bias T(s) Decoupler transfer function
B Bottoms flow rate Tb Base temperature
B*i Limit on control Tf Exhaust temperature
cA Concentration of A TR Reset time
C Cumulative sum U Heat-transfer coefficient
Cd Discharge coefficient u, U Manipulated variable, controller output
Ci Inlet concentration V Volume
C*i Limit on control move Vs Product value
CL Specific heat of liquid w Mass flow rate
C0 Integration constant wi Weighting factor
Cp Process capability W Steam flow rate
Cr Heat capacity of reactants x Mass fraction
CV Valve flow coefficient ⎯x Sample mean
D Distillate flow rate, disturbance xi Optimization variable
D*i Limit on output xT Pressure drop ratio factor
e Error X Transform of deviation variable
E Economy of evaporator y, Y Process output, controlled variable, valve travel
f Function of time Ysp Set point
F, f Feed flow rate z Controller tuning law, expansion factor
FL Pressure recovery factor zi Feed mole fraction (distillation)
gc Unit conversion constant Z Compressibility factor
gi Algebraic inequality constraint
G Transfer function Greek Symbols
Gc Controller transfer function α Digital filter coefficient
Gd Disturbance transfer function aT Temperature coefficient of resistance
Gf Feedforward controller transfer function β Resistance thermometer parameter
Gm Sensor transfer function γ Ratio of specific heats
Gp Process transfer function δ Move suppression factor, shift in target value
Gt Transmitter transfer function ∆q Load step change
Gv Valve transfer function ∆t Time step
hi Algebraic equality constraints ∆T Temperature change
h1 Liquid head in tank ∆u Control move
H Latent heat of vaporization, control limit or threshold ε Spectral emissivity, step size
i Summation index ζ Damping factor (second-order system)
Ii Impulse response coefficient θ Time delay
j Time index λ Relative gain array parameter, wavelength
J Objective function or performance index Λ Relative gain array
k Time index Deviation variable
kf Flow coefficient ρ Density
kr Kinetic rate constant σ Stefan-Boltzmann constant, standard deviation
K Gain, slack parameter Σt Total response time
Kc Controller gain τ Time constant
Kd Disturbance transfer function gain τd Natural period of closed loop, disturbance time constant
Km Measurement gain D Derivative time (PID controller)
Kp Process gain τF Filter time constant
Ku Ultimate controller gain (stability) τI Integral time (PID controller)
L Load variable τP Process time constant
Lp Sound pressure level τo Period of oscillation
M Manipulated variable φPI Phase lag
mc Number of constraints
Mv Mass flow Subscripts
Mr Mass of reactants A Species A
Mw Molecular weight b Best
n Number of data points, number of stages or effects c Controller
N Number of inputs/outputs, model horizon d Disturbance
p Proportional band (%) eff Effective
pc Vapor pressure f Feedforward
pd Actuator pressure i Initial, inlet
pi Pressure L Load, disturbance
pu Proportional band (ultimate) m Measurement or sensor
q Radiated energy flux p Process
qb Energy flux to a black body s Steady state
Q Flow rate sp Set-point value
rc Number of constraints t Transmitter
R Equal-percentage valve characteristic u Ultimate
RT Resistance in temperature sensor v Valve
R1 Valve resistance
FUNDAMENTALS OF PROCESS DYNAMICS AND CONTROL
THE GENERAL CONTROL SYSTEM by the proper selection of control modes to satisfy the requirements of
the process and, second, by the appropriate tuning of those modes.
A process is shown in Fig. 8-1 with a manipulated input U, a load input Feedforward Control A feedforward system uses measure-
D, and a controlled output Y, which could be flow, pressure, liquid ments of disturbance variables to position the manipulated variable in
level, temperature, composition, or any other inventory, environmen- such a way as to minimize any resulting deviation. The disturbance
tal, or quality variable that is to be held at a desired value identified as variables could be either measured loads or the set point, the former
the set point Ysp. The load may be a single variable or an aggregate of being more common. The feedforward gain must be set precisely to
variables either acting independently or manipulated for other pur- reduce the deviation of the controlled variable from the set point.
poses, affecting the controlled variable much as the manipulated vari- Feedforward control is usually combined with feedback control to
able does. Changes in load may occur randomly as caused by changes eliminate any offset resulting from inaccurate measurements and calcula-
in weather, diurnally with ambient temperature, manually when oper- tions and unmeasured load components. The feedback controller can be
ators change production rate, stepwise when equipment is switched used as a bias on the feedforward controller or in a multiplicative form.
into or out of service, or cyclically as the result of oscillations in other Computer Control Computers have been used to replace analog
control loops. Variations in load will drive the controlled variable away PID controllers, either by setting set points of lower-level controllers in
from the set point, requiring a corresponding change in the manipu- supervisory control or by driving valves directly in direct digital control.
lated variable to bring it back. The manipulated variable must also Single-station digital controllers perform PID control in one or two
change to move the controlled variable from one set point to another. loops, including computing functions such as mathematical operations,
An open-loop system positions the manipulated variable either characterization, lags, and dead time, with digital logic and alarms. Dis-
manually or on a programmed basis, without using any process mea- tributed control systems provide all these functions, with the digital
surements. This operation is acceptable for well-defined processes processor shared among many control loops; separate processors may
without disturbances. An automated transfer switch is provided to be used for displays, communications, file servers, and the like. A host
allow manual adjustment of the manipulated variable in case the computer may be added to perform high-level operations such as
process or the control system is not performing satisfactorily. scheduling, optimization, and multivariable control. More details on
A closed-loop system uses the measurement of one or more process computer control are provided later in this section.
variables to move the manipulated variable to achieve control. Closed-
loop systems may include feedforward, feedback, or both. PROCESS DYNAMICS AND MATHEMATICAL MODELS
Feedback Control In a feedback control loop, the controlled
variable is compared to the set point Ysp, with the error E acted upon GENERAL REFERENCES: Seborg, Edgar, and Mellichamp, Process Dynamics
by the controller to move U in such a way as to minimize the error. and Control, Wiley, New York, 2004; Marlin, Process Control, McGraw-Hill,
This action is specifically negative feedback, in that an increase in New York, 2000; Ogunnaike and Ray, Process Dynamics Modeling and Control,
error moves U so as to decrease the error. (Positive feedback would Oxford University Press, New York, 1994; Smith and Corripio, Principles and
cause the error to expand rather than diminish and therefore does not Practices of Automatic Process Control, Wiley, New York, 1997.
regulate.) The action of the controller is selectable to allow use on Open-Loop versus Closed-Loop Dynamics It is common in
process gains of both signs. industry to manipulate coolant in a jacketed reactor in order to control
The controller has tuning parameters related to proportional, integral, conditions in the reactor itself. A simplified schematic diagram of such
derivative, lag, dead time, and sampling functions. A negative feedback a reactor control system is shown in Fig. 8-2. Assume that the reactor
loop will oscillate if the controller gain is too high; but if it is too low, con- temperature is adjusted by a controller that increases the coolant flow in
trol will be ineffective. The controller parameters must be properly proportion to the difference between the desired reactor temperature
related to the process parameters to ensure closed-loop stability while and the temperature that is measured. The proportionality constant is
still providing effective control. This relationship is accomplished, first, Kc. If a small change in the temperature of the inlet stream occurs, then
Disturbance, D
Feedforward
Controller
Controlled
Error Feedback Variable, Y
Set Process
Controller Manipulated Variable, U
Point, Ysp
Feedback Loop
8-5
8-6 PROCESS CONTROL
shows that there is an optimal choice for Kc, somewhere between 0 (no
control) and Ku (stability limit). If one has a dynamic model of a process,
then this model can be used to calculate controller settings. In Fig. 8-3,
no time scale is given, but rather the figure shows relative responses. A
well-designed controller might be able to speed up the response of a
process by a factor of roughly 2 to 4. Exactly how fast the control system
responds is determined by the dynamics of the process itself.
Physical Models versus Empirical Models In developing a
dynamic process model, there are two distinct approaches that can be
taken. The first involves models based on first principles, called phys-
ical or first principles models, and the second involves empirical mod-
els. The conservation laws of mass, energy, and momentum form the
basis for developing physical models. The resulting models typically
FIG. 8-2 Reactor control system. involve sets of differential and algebraic equations that must be solved
simultaneously. Empirical models, by contrast, involve postulating the
form of a dynamic model, usually as a transfer function, which is dis-
depending on the value of Kc, one might observe the reactor tempera- cussed below. This transfer function contains a number of parameters
ture responses shown in Fig. 8-3. The top plot shows the case for no that need to be estimated from data. For the development of both
control (Kc = 0), which is called the open loop, or the normal dynamic physical and empirical models, the most expensive step normally
response of the process by itself. As Kc increases, several effects can be involves verification of their accuracy in predicting plant behavior.
noted. First, the reactor temperature responds faster and faster. Sec- To illustrate the development of a physical model, a simplified
ond, for the initial increases in K, the maximum deviation in the reactor treatment of the reactor, shown in Fig. 8-2, is used. It is assumed that
temperature becomes smaller. Both of these effects are desirable so that the reactor is operating isothermally and that the inlet and exit volu-
disturbances from normal operation have as small an effect as possible metric flows and densities are the same. There are two components, A
on the process under study. As the gain is increased further, eventually and B, in the reactor, and a single first-order reaction of A→B takes
a point is reached where the reactor temperature oscillates indefinitely, place. The inlet concentration of A, which we call ci, varies with time.
which is undesirable. This point is called the stability limit, where Kc = A dynamic mass balance for the concentration of A, denoted cA, can be
Ku, the ultimate controller gain. Increasing Kc further causes the magni- written as follows:
tude of the oscillations to increase, with the result that the control valve
will cycle between full open and closed.
The responses shown in Fig. 8-3 are typical of the vast majority of dcA
V = Fci − FcA − krVcA (8-1)
regulatory loops encountered in the process industries. Figure 8-3 dt
In Eq. (8-1), the flow in of component A is Fci, the flow out is FcA, and Define
the loss via reaction is krVcA, where V = reactor volume and kr =
kinetic rate constant. In this example, ci is the input, or forcing, variable Fci(t) − FcA − krVc2A
f(cAt) = (8-5)
and cA is the output variable. If V, F, and kr are constant, Eq. (8-1) can V
be rearranged by dividing by F + krV so that it contains only two
groups of parameters. The result is then
values for τ and K so that the postulated model described the reactor’s
dynamic response. The advantage of the physical model over the m2 ∆t ∆t
empirical model is that the physical model gives insight into how reac-
tor parameters affect the values of τ and K, which in turn affects the
m3 = f cA(t) + 2 , t +
2 2 (8-9)
Denoting £(CA) as CA(s) and using the relationships in Table 8-1 give
Kc⎯i
τ sCA(s) = s − CA(s) (8-12)
Kc⎯i s
CA(s) = (8-13)
τs + 1 FIG. 8-4 First-order transfer function.
By using the entries in Table 8-1, Eq. (8-13) can be inverted to give
time. Most real processes have variables that are continuous, such as
the transient response of cA as
temperature, pressure, and flow. However, some processes involve
cA(t) = (Kc⎯i)(1 − e−1τ) (8-14) discrete events, such as the starting or stopping of a pump. In addi-
tion, modern plants are controlled by digital computers, which are dis-
Equation (8-14) shows that cA starts from 0 and builds up exponen- crete. In controlling a process, a digital system samples variables at a
tially to a final concentration of Kci. Note that to get Eq. (8-14), it was fixed rate, and the resulting system is a sampled data system. From
only necessary to solve the algebraic Eq. (8-12) and then find the one sampling instant until the next, variables are assumed to remain
inverse of CA(s) in Table 8-1. The original differential equation was fixed at their sampled values. Similarly, in controlling a process, a dig-
not solved directly. In general, techniques such as partial fraction ital computer sends out signals to control elements, usually valves, at
expansion must be used to solve higher-order differential equations discrete instants of time. These signals remain fixed until the next
with Laplace transforms. sampling instant.
Transfer Functions and Block Diagrams A very convenient Figure 8-5 illustrates the concept of sampling a continuous func-
and compact method of representing the process dynamics of linear tion. At integer values of the sampling rate ∆t, the value of the variable
systems involves the use of transfer functions and block diagrams. A trans- to be sampled is measured and held until the next sampling instant. To
fer function can be obtained by starting with a physical model, as dis- deal with sampled data systems, the z transform has been developed.
cussed previously. If the physical model is nonlinear, first it needs to The z transform of the function given in Fig. 8-5 is defined as
be linearized around an operating point. The resulting linearized
∞
model is then approximately valid in a region around this operating
point. To illustrate how transfer functions are developed, Eq. (8-2) Z(f) = f(n ∆t)z−n (8-19)
n=0
will again be used. First, one defines deviation variables, which are the
process variables minus their steady-state values at the operating
In an analogous manner to Laplace transforms, one can develop trans-
point. For Eq. (8-2), there would be deviation variables for both cA
fer functions in the z domain as well as block diagrams. Tables of z
and ci, and these are defined as
transform pairs have been published (Seborg, Edgar, and Mel-
ξ = cA − c⎯A (8-15) lichamp, Process Dynamics and Control, Wiley, New York, 2004) so
that the discrete transfer functions can be inverted back to the time
ξi = ci − c⎯i (8-16) domain. The inverse gives the value of the function at the discrete
sampling instants. Sampling a continuous variable results in a loss of
where the overbar stands for steady state. Substitution of Eqs. (8-15) information. However, in practical applications, sampling is fast
and (8-16) into Eq. (8-2) gives enough that the loss is typically insignificant and the difference
between continuous and discrete modeling is small in terms of its
dξ effect on control. Increasingly, model predictive controllers that make
τ = Kξi − ξ + (K ⎯ci − ⎯cA) (8-17) use of discrete dynamic models are being used in the process indus-
dt tries. The purpose of these controllers is to guide a process to opti-
mum operating points. These model predictive control algorithms are
The term in parentheses in Eq. (8-17) is zero at steady state, and thus typically run at much slower sampling rates than are used for basic
it can be dropped. Next the Laplace transform is taken, and the result- control loops such as flow control or pressure control. The discrete
ing algebraic equation is solved. dynamic models used are normally developed from data generated
By denoting X(s) as the Laplace transform of ξ and Xi(s) as the
transform of ξi, the final transfer function can be written as
X K
= (8-18)
Xi τs + 1
Distance-Velocity Lag (Dead-Time Element) The dead-time arrives at the temperature measurement point. If some element of a
or time-delay element, commonly called a distance-velocity lag, is system produces a dead time of θ time units, then an input to that unit
often encountered in process systems. For example, if a temperature- f(t) will be reproduced at the output as f(t − θ). The transfer function
measuring element is located downstream from a heat exchanger, a for a pure dead-time element is shown in Fig. 8-17, and the transient
time delay occurs before the heated fluid leaving the exchanger response of the element is shown in Fig. 8-18.
FIG. 8-11 Response of pure capacity system. FIG. 8-12 Two tanks in series.
FUNDAMENTALS OF PROCESS DYNAMICS AND CONTROL 8-11
FIG. 8-13 Second-order transfer function. FIG. 8-17 Dead-time transfer function.
K
G(s) = n (8-25)
(τs + 1)
The step response of this transfer function is shown in Fig. 8-19. Note
that all curves reach about 60 percent of their final value at t = nτ.
FIG. 8-16 Response of general second-order system. FIG. 8-19 Response of nth-order lags.
8-12 PROCESS CONTROL
U1 Y1
G11
G21
G12
U2 Y2
G22
Higher-order systems can be approximated by a first- or second- model parameters K, τ, and θ fitted to the data. These parameters are
order plus dead-time system for control system design. calculated by using optimization to minimize the squared difference
Multi-input, Multioutput Systems The dynamic systems con- between the model predictions and the data, i.e., a least squares
sidered up to this point have been examples of single-input, single- approach. Let each measured data point be represented by yj (mea-
output (SISO) systems. In chemical processes, one often encounters sured response), tj (time of measured response), j = 1 to n. Then the
systems where one input can affect more than one output. For exam- least squares problem can be formulated as
ple, assume that one is studying a distillation tower in which both
reflux and boil-up are manipulated for control purposes. If the output n
variables are the top and bottom product compositions, then each minτ,θ,κ [yj − ^y(tj)]2 (8-27)
input affects both outputs. For this distillation example, the process is j=1
referred to as a 2 × 2 system to indicate the number of inputs and out-
^
puts. In general, multi-input, multioutput (MIMO) systems can have where y(t j) is the predicted value of y at time tj and n is the number of
n inputs and m outputs with n ≠ m, and they can be nonlinear. Such a data points. This optimization problem can be solved to calculate the
system would be called an n × m system. An example of a transfer optimal values of K, τ, and θ. A number of software packages such as
function for a 2 × 2 linear system is given in Fig. 8-20. Note that since Excel Solver are available for minimizing Eq. (8-27).
linear systems are involved, the effects of the two inputs on each out- One operational problem caused by step forcing is the fact that
put are additive. In many process control systems, one input is the process under study is moved away from its steady-state operat-
selected to control one output in a MIMO system. For m outputs ing point. Plant managers may be reluctant to allow large steady-
there would be m such selections. For this type of control strategy, one state changes, since normal production will be disturbed by the
needs to consider which inputs and outputs to couple, and this prob- changes. As a result, alternative methods of forcing actual processes
lem is referred to as loop pairing. Another important issue that arises have been developed, and these included pulse testing and pseudo-
involves interaction between control loops. When one loop makes a random binary signal (PRBS) forcing, both of which are illustrated
change in its manipulated variable, the change affects the other loops in Fig. 8-22. With pulse forcing, one introduces a step, and then
in the system. These changes are the direct result of the multivariable after a period of time the input is returned to its original value. The
nature of the process. In some cases, the interaction can be so severe result is that the process dynamics are excited, but after the forcing
that overall control system performance is drastically reduced. Finally, the process returns to its original steady state. PRBS forcing involves
some of the modern approaches to process control tackle the MIMO a series of pulses of fixed height and random duration, as shown in
problem directly, and they simultaneously use all manipulated vari- Fig. 8-22. The advantage of PRBS is that forcing can be concen-
ables to control all output variables rather than pair one input to one trated on particular frequency ranges that are important for control
output (see later section on multivariable control). system design.
Fitting Dynamic Models to Experimental Data In developing Transfer function models are linear, but chemical processes are
empirical transfer functions, it is necessary to identify model parameters known to exhibit nonlinear behavior. One could use the same type of
from experimental data. There are a number of approaches to process optimization objective as given in Eq. (8-27) to determine parame-
identification that have been published. The simplest approach involves ters in nonlinear first-principles models, such as Eq. (8-3) presented
introducing a step test into the process and recording the response of the earlier. Also, nonlinear empirical models, such as neural network
process, as illustrated in Fig. 8-21. The x’s in the figure represent the models, have recently been proposed for process applications. The
recorded data. For purposes of illustration, the process under study will be key to the use of these nonlinear empirical models is to have high-
assumed to be first-order with dead time and have the transfer function quality process data, which allows the important nonlinearities to be
identified.
Y(s)
G(s) = = K exp (−θs)τs + 1 (8-26)
U(s) FEEDBACK CONTROL SYSTEM CHARACTERISTICS
GENERAL REFERENCES: Shinskey, Process Control Systems, 4th ed., McGraw-
The response y(t), produced by Eq. (8-26) can be found by inverting Hill, New York, 1996; Seborg, Edgar, and Mellichamp, Process Dynamics and
the transfer function, and it is also shown in Fig. 8-21 for a set of Control, Wiley, New York, 1989.
FUNDAMENTALS OF PROCESS DYNAMICS AND CONTROL 8-13
There are two objectives in applying feedback control: (1) regulate the troller with certain parameter settings regulates a variable, relative to
controlled variable at the set point following changes in load and (2) the best response that can be achieved for that particular process.
respond to set-point changes, the latter called servo operation. In fluid Robustness is a measure of how small a change in a process parameter
processes, almost all control loops must contend with variations in load, is required to bring the loop from its current state to the limit of sta-
and therefore regulation is of primary importance. While most loops will bility (ΠG = 1.0). The well-damped loop in Fig. 8-24 has a robustness
operate continuously at fixed set points, frequent changes in set points of 79 percent, in that increasing the gain of any element in the loop by
can occur in flow loops and in batch production. The most common a factor of 1/0.56, or 1.79, would bring the loop to the limit of stability.
mechanism for achieving both objectives is feedback control, because it Increasing controller performance by raising its gain can therefore be
is the simplest and most universally applicable approach to the problem. expected to decrease robustness. Both performance and robustness
Closing the Loop The simplest representation of the closed feed- are functions of the dynamics of the process being controlled, the
back loop is shown in Fig. 8-23. The load is shown entering the process selection of the controller, and the tuning of the controller parameters.
at the same point as the manipulated variable because that is the most On/Off Control An on/off controller is used for manipulated
common point of entry, and because, lacking better information, the ele- variables having only two states. They commonly control tempera-
ments in the path of the manipulated variable are the best estimates of tures in homes, electric water heaters and refrigerators, and pressure
those in the load path. The load rarely impacts directly on the controlled and liquid level in pumped storage systems. On/off control is satisfac-
variable without passing through the dominant lag in the process. Where tory where slow cycling is acceptable, because it always leads to
the load is unmeasured, its current value can be observed as the con- cycling when the load lies between the two states of the manipulated
troller output required to keep the controlled variable Y at set point Ysp. variable. The cycle will be positioned symmetrically about the set
If the loop is opened—either by placing the controller in manual point only if the load happens to be equidistant between the two states
operation or by setting its gains to zero—the load will have complete of the manipulated variable. The period of the symmetric cycle will be
influence over the controlled variable, and the set point will have none. approximately 4θ, where θ is the dead time in the loop. If the load is
Only by closing the loop with controller gain as high as possible will the not centered between the states of the manipulated variable, the period
influence of the load be minimized and that of the set point be maxi- will tend to increase and the cycle will follow a sawtooth pattern.
mized. There is a practical limit to the controller gain, however, at the Every on/off controller has some degree of dead band, also known as
point where the controlled variable develops a uniform oscillation. This lockup, or differential gap. Its function is to prevent erratic switching
is defined as the limit of stability, and it is reached when the product of between states, thereby extending the life of contacts and motors.
gains in the loop ΠG = GcGvGp is equal to 1.0 at the period of the oscil- Instead of changing states precisely when the controlled variable crosses
lation. If the gain of any element in the loop increases from this condi- the set point, the controller will change states at two different points for
tion, oscillations will expand, creating a dangerous situation where safe increasing and decreasing signals. The difference between these two
limits of operation could be exceeded in a few cycles. Consequently, switching points is the dead band (see Fig. 8-25); it increases the ampli-
control loops should be left in a condition where the loop gain is less tude and period of the cycle, similar to the effects of dead time.
than 1.0 by a safe margin that allows for possible variations in process A three-state controller is used to drive either a pair of independent
parameters. Figure 8-24 describes a load response under PID (propor- two-state actuators, such as heating and cooling valves, or a bidirec-
tional-integral-derivative) control where the loop is well damped at a tional motorized actuator. The controller is comprised of two on/off
loop gain of 0.56; loop gain is then increased to 0.93 and to 1.05, creat- controllers, each with dead band, separated by a dead zone. While the
ing a lightly damped and then an expanding cycle, respectively. controlled variable lies within the dead zone, neither output is ener-
In controller tuning, a choice must be made between performance gized. This controller can drive a motorized valve to the point where
and robustness. Performance is a measure of how well a given con- the manipulated variable matches the load, thereby avoiding cycling.
8-14 PROCESS CONTROL
Proportional Control A proportional controller moves its out- should be set at the best estimate of the load, and the proportional band
put proportional to the deviation e between the controlled variable y set as low as possible. Processes requiring a proportional band of more
and its set point ysp: than a few percent may control with unacceptably large values of offset.
Proportional control is most often used to regulate liquid level,
100 where variations in the controlled variable carry no economic penalty
u = Kc e + b = e + b (8-28)
P and where other control modes can easily destabilize the loop. It is
actually recommended for controlling the level in a surge tank when
where e = ±(y − ysp), the sign selected to produce negative feedback. In manipulating the flow of feed to a critical downstream process. By set-
some controllers, proportional gain Kc is introduced as a pure number; in ting the proportional band just under 100 percent, the level is allowed
others, it is set as 100/P, where P is the proportional band in percent. The to vary over the full range of the tank capacity as inflow fluctuates,
output bias b of the controller is also known as manual reset. The propor- thereby minimizing the resulting rate of change of manipulated out-
tional controller is not a good regulator, because any change in output flow. This technique is called averaging level control.
required to respond to a change in load results in a corresponding change Proportional-plus-Integral (PI) Control Integral action elimi-
in the controlled variable. To minimize the resulting offset, the bias nates the offset described above by moving the controller output at a
FUNDAMENTALS OF PROCESS DYNAMICS AND CONTROL 8-15
FIG. 8-23 Both load regulation and set-point response require high gains for the feedback controller.
rate proportional to the deviation from set point—the output will then
not stop moving until the deviation is zero. Although available alone in
100 1
dy
u = e + e dt ± τD + C0
P τI dt
(8-31)
an integral controller, it is most often combined with proportional
action in a PI controller: where τD is the derivative time constant. Note that derivative action
100
1
u = e + edt + C0
P τI
(8-29)
is applied to the controlled variable rather than to the deviation, as it
should not be applied to the set point; the selection of the sign for the
derivative term must be consistent with the action of the controller.
In some PID controllers, the integral and derivative terms are com-
where τI is the integral time constant in minutes; in some controllers bined serially rather than in parallel, as done in the last equation. This
it is introduced as integral gain or reset rate 1τI in repeats per minute. results in interaction between these modes, such that the effective val-
The last term in the equation is the constant of integration, the value ues of the controller parameters differ from their set values as follows:
of the controller output when integration begins. The PI controller is
by far the most commonly used controller in the process industries.
Because the integral term lags the proportional term by 90° in phase, τI,eff = τI + τD
the PI controller then always produces a phase lag between 0° and 90°:
1
τD,eff = (8-32)
τo 1τD + 1τI
φPI = −tan
−1
(8-30)
2πτI
τD
100
where τo is the period of oscillation of the loop. The phase angle Kc = 1 +
P τI
should be kept between 15° for lag-dominant processes and 45° for
dead-time-dominant processes for optimum results.
Proportional-plus-Integral-plus-Derivative (PID) Control The The performance of the interacting controller is almost as high as that
derivative mode moves the controller output as a function of the rate of of the noninteracting controller on most processes, but the tuning
change of the controlled variable, which adds phase lead to the con- rules differ because of the above relationships. Both controllers are in
troller, increasing its speed of response. It is normally combined with common use in digital systems.
proportional and integral modes. The noninteracting or ideal form of There is always a gain limit placed upon the derivative vector—a
the PID controller appears functionally as value of 10 is typical. However, interaction decreases the derivative
∏G = 0.56
1.05
Controlled
variable Set
0.93
Time
FIG. 8-24 Transition from well-damped load response to instability develops as loop gain
increases.
8-16 PROCESS CONTROL
gain below this value by the factor 1 + τDτI, which is the reason for the gain—in Fig. 8-26, that fraction is 0.13. Increasing Kc can reduce the
decreased performance of the interacting PID controller. Sampling in a offset, but with an accompanying loss in damping.
digital controller has a similar effect, limiting derivative gain to the ratio The PI and PID controller were tuned to produce a minimum inte-
of derivative time to the sample interval of the controller. Noise on the grated absolute error (IAE). Their response curves are similar in
controlled variable is amplified by the derivative gain, preventing its use appearance to a gaussian distribution curve, but with a damped cycle
in controlling flow and liquid level. Derivative action is recommended in the trailing edge. The peak deviation of the PID response curve is
for control of temperature and composition in multiple-capacity only 0.12 times the uncontrolled offset, occurring at 0.36Στ; the peak
processes with little measurement noise. deviation of the PI response curve is 0.21 times the uncontrolled off-
Controller Comparison Figure 8-26 compares the step load set, occurring at 0.48Στ. These values can be used to predict the load
response of a distributed lag without control, and with P, PI, and inter- response of any distributed lag whose parameters K and Στ are known
acting PID control. A distributed lag is a process whose resistance and or can be estimated as described below.
capacity are distributed throughout its length—a heat exchanger is
characteristic of this class, its heat-transfer surface and heat capacity CONTROLLER TUNING
being uniformly distributed. Other examples include imperfectly
stirred tanks and distillation columns—both trayed and packed. The sig- The performance of a controller depends as much on its tuning as on
nature of a distributed lag is its open-loop (uncontrolled) step response, its design. Tuning must be applied by the end user to fit the controller
featuring a relatively short dead time followed by a dominant lag called to the controlled process. There are many different approaches to con-
Στ, which is the time required to reach 63.2 percent complete response. troller tuning, based on the particular performance criteria selected,
The proportional controller is unable to return the controlled vari- whether load or set-point changes are more important, whether the
able to the set point following the step load change, as a deviation is process is lag- or dead-time-dominant, and the availability of informa-
required to sustain its output at a value different from its fixed bias b. tion about the process dynamics. The earliest definitive work in this
The amount of proportional offset produced as a fraction of the field was done at the Taylor Instrument Company by Ziegler and
uncontrolled offset is 1/(1 + KKc), where K is the steady-state process Nichols (Trans. ASME, p. 759, 1942), tuning PI and interacting PID
FIG. 8-26 Minimum-IAE tuning gives very satisfactory load response for a distributed lag.
FUNDAMENTALS OF PROCESS DYNAMICS AND CONTROL 8-17
controllers for optimum response to step load changes applied to lag- an effective test of its tuning, however, as its steady-state output is pro-
dominant processes. While these tuning rules are still in use, they do portional to its set point. Other loops should be load-tested: simulate a
not apply to set-point changes, dead-time-dominant processes, or load change from a steady state at zero deviation by transferring the
noninteracting PID controllers (Seborg, Edgar, and Mellichamp, controller to manual and stepping its output, and then immediately
Process Dynamics and Control, Wiley, New York, 2004). transferring back to automatic before a deviation develops.
Controller Performance Criteria The most useful measures of Figure 8-27a and b shows variations in the response of a distrib-
controller performance in an industrial setting are the maximum devi- uted lag to a step change in load for different combinations of pro-
ation in the controlled variable resulting from a disturbance, and its portional and integral settings of a PI controller. The maximum
integral. The disturbance could be to the set point or to the load, deviation is the most important criterion for variables that could
depending on the variable being controlled and its context in the exceed safe operating levels, such as steam pressure, drum level, and
process. The size of the deviation and its integral are proportional to steam temperature in a boiler. The same rule can apply to product
the size of the disturbance (if the loop is linear at the operating point). quality if violating specifications causes it to be rejected. However, if
While actual disturbances arising in a plant may appear to be random, the product can be accumulated in a downstream storage tank, its
the controller needs a reproducible test to determine how well it is average quality is more important, and this is a function of the devia-
tuned. The disturbance of choice for test purposes is the step, because tion integrated over the residence time of the tank. Deviation in the
it can be applied manually, and by containing all frequencies including other direction, where the product is better than specification, is safe
zero it exercises all modes of the controller. (The step actually has the but increases production costs in proportion to the integrated deviation
same frequency distribution as integrated white noise, a “random because quality is given away.
walk.”) When tuned optimally for step disturbances, the controller For a PI or PID controller, the integrated deviation—better known
will be optimally tuned for most other disturbances as well. as integrated error IE—is related to the controller settings
A step change in set point, however, may be a poor indicator of a
loop’s load response. For example, a liquid-level controller does not PτI
have to integrate to follow a set-point change, as its steady-state output IE = ∆u
100 (8-33)
is independent of the set point. Stepping a flow controller’s set point is
Poptx3
Poptx2
Controlled Popt
variable
Set
Popt/1.5
0 1 2 3 4
Time, t/ ∑τ
(a)
τI optx2
Controlled
variable τI optx3
τI opt
Set
I opt/1.5
0 1 2 3 4
Time, t/ ∑τ
(b)
FIG. 8-27 The optimum settings produce minimum-IAE load response. (a) The pro-
portional band primarily affects damping and peak deviation. (b) Integral time deter-
mines overshoot.
Exploring the Variety of Random
Documents with Different Content
Veri valui ryssän edetessä,
Säästäneet ei lasta kätkyessä;
Taas ruotsalaiset joutuivat uuteen huoleen,
Kun ryssä alkoi siirtyä Kruunulinnan puoleen,
On Hämeenlinna myös nimenä sillä, —
Miks näin, lie silläkin syynsä kyllä —;
Siellä riehuvat ryssät tuhoisasti
Ja tahtovat mennä aina Turkuun asti.
Laajalti ja säälimättä tuhoaapi ryssä,
Kirkot, kylät häviävät tulen syleilyssä.
Heit' ei ollut kahtasataakaan, vannon sen
Kautta Jumalan ja pyhän Neitsyen!
Ja Ruotsin sotaväki vaan
Ei tohtinut tulla auttamaan!
Yhä kauhunäytelmät uudistuupi,
Hattulan kirkkoon ryssä tunkeuupi,
Sieltä raastavat pyhäinkuvat ja ristit,
Turkuun saapuu ne kauhun viestit.
On luoksensa kahden päivän matka vaan,
Mutta Herra Steen vain ei tohdi auttamaan.
Ritarin nimeä kannatte. Herra Steen,
Älkää pistäkö päätänne pensaaseen.
Näin Turussa hällä surun sydämessä,
Vaan tointa ei kylliks kohtalon kääntyessä.
Hän Ahvenanmaalla ja Uudellamaalla antoi käskyn julki,
Ja tietokapula myös yli koko Suomen kulki.
Mies aseineen joka talosta,
Niin ryssän hän voisi torjua.
Joka mies oli varmasti silloin matkassa,
Kun Herra Steenin tykistö tuotiin Turusta,
Siin' aika meluin kulkea mahdettiin,
Kun kolmesataa tykkiä laahattiin!
Pyhän Mathiaksen päivänä — määrä on se —
Lähtee Ruotsin miehet liikkeelle,
Ruotsalaisia olivat kaikkein useimmat,
Ja Ruotsin olitkin miehistä parahat;
Toiset lähtivät reellä ja toiset hevosin,
Kuinka kukin vaan pääsi parhaimmin.
Yhdeksänsataa siin' oli huoviakin,
Jotka lähtivät sodan vaaroihin.
Ahvenanmaalta tuli kaksituhatta,
Talonpoikaa kelpoa ja kaunista,
Ja yhteenkoottuna nyt olit kaikin,
Hän heistä mahtavan joukon saikin,
Neljäkymmentä tuhatta miest' oli heitä,
Kaikk' aseilla hyvillä varustuneita.
Kun ryssät tiedon saivat tästä,
He eivät viivytelleet lähtemästä,
Ei heistä enää näkynyt jälkeäkään,
Ei mistään löytty ryssää yhtäkään.
Ei kiinni heitä saa, se tieto on vakaa,
Neljäkolmatta peninkulmaa vaikk' ajettiin takaa.
Lepoon väkemme asettui allapäin
Ja moni kai ajatteli mielessään näin:
Mitä nyt auttaa enää kaikki puuha ja huoli!
Autiona on Karjala, Savo ja Hämettä puoli,
Niissä ei kuulu yhtään ihmissanaa,
Eik' ole lehmää, koiraa, kukkoa tai kanaa!
Maa autio on: tääll' olla ei halaa,
Tyhjä on kahdeksankymmenen peninkulman ala.
Varustusten määrä kyll' oli suuri
Tätä mahtavaa retkeä varten juuri.
Yli viissataa luotipyssyä ja monta ampujaa,
Mutta Ruotsin hyödyksi mitään he ei voineet toimittaa.
Palkkasoturit Tukholman
Saivat silloin sen kunnian,
Että heidän tuli Suomehen jäädä
Ja Olavinlinnaa vartioimaan käydä. —
Eipä muuta nyt tullut tehtäväksi,
Muut kaikki jälleen kotiin läksi.
Ja huovimiehet, ollen niin,
Ne tulivat Turun kaupunkiin.
Herra Steen heidät sinne tänne majoittaapi,
Moni pappila heitä vieraiksensa saapi.
Herra Steen itse tahtoi Ruotsihin takaisin,
Siell' aikoi uuden joukon hän koota vieläkin.
Näin Suomessa hän asiat suunnittaa:
Svante Niilonpoika täällä nyt päällikkyyden saa.
Ja tälle päällikölle vastavalitulle
Jää avuks' Steen Tuurenpoika ynnä Erik Trulle;
Häntä totelkoot Suomen kaikk' asujat,
Niin läheiset kuin myös kaukaisat,
Päästä kuukauden Suomen parhaaksi
Herra Steen taas palata lupasi.
Savon synty.
"tora totkusista
Artti ahven-maimasista".
Savonlinnan ihmeet.
*****
Vaan palatkaamme vielä hetkisen ajaksi takaisin Savonlinnan
kummitus-taruihin. Pari vuosisataa sen jälkeen, kun Olaus Magnus
uutena Pliniona kirjoitti merkilliset juttunsa Arx novam ihmeistä, on
Säämingin pappi kaikessa hiljaisuudessa piirtänyt kirkonkirjaansa
seuraavan muistelman:
Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.
ebooknice.com