Nof S.Y. Springer Handbook of Automation

Подождите немного. Документ загружается.

Automating Errors and Conﬂicts Prognostics and Prevention References 525

30.91 X.W. Chen, S.Y. Nof: An agent-based conﬂict and

error detection model, submitted to Int. J. Prod.

Res. (2008)

30.92 C.L. Yang, S.Y. Nof: Analysis, detection policy, and

performance measures of detection task plan-

ning errors and conﬂicts, PRISM Res. Memorandum,

2004-P2 (School of Industrial Engineering, Purdue

University, West Lafayette 2004)

30.93 J. Avila-Soria: Interactive Error Recovery for Robotic

Assembly Using a Neural-Fuzzy Approach. Master

Thesis (School of Industrial Engineering, Purdue

University, West Lafayette 1999)

30.94 J.D. Velásquez, M.A. Lara, S.Y. Nof: Systematic res-

olution of conﬂict situation in collaborative facility

design, Int. J. Prod. Econ. 116(1), 139–153 (2008),

(2008)

30.95 S.Y. Nof, O.Z. Maimon, R.G. Wilhelm: Experiments

for Planning Error-Recovery Programs in Robotic

Work, Proc. Int. Comput. Eng. Conf. Exhib. 2,253–

264 (1987)

30.96 M. Imai, K. Hiraki, Y. Anzai: Human-robot interface

with attention, Syst. Comput. Jpn. 26(12), 83–95

(1995)

30.97 T.C. Lueth, U.M. Nassal, U. Rembold: Reliability and

integrated capabilities of locomotion and manipu-

lation for autonomous robot assembly, Robot.

Auton. Syst. 14, 185–198 (1995)

30.98 H.-J. Wu, S.B. Joshi: Error recovery in MPSG-based

controllers for shop ﬂoor control, Proc. IEEE Int.

Conf. Robot. Autom. ICRA 2, 1374–1379 (1994)

30.99 K. Sycara: Negotiation planning: An AI approach,

Eur. J. Oper. Res. 46(2), 216–234 (1990)

30.100 L. Fang, K.W. Hipel, D.M. Kilgour: Interactive Deci-

sion Making (Wiley, New York 1993)

30.101 J.-S.R. Jang: ANFIS: Adaptive-network-based fuzzy

inference systems, IEEE Trans. Syst. Man. Cybern.

23, 665–685 (1993)

30.102 A. Kusiak, J. Wang: Dependency analysis in con-

straint negotiation, IEEE Trans. Syst. Man. Cybern.

25(9), 1301–1313 (1995)

30.103 J.-S.R. Jang, N. Gulley: Fuzzy Systems Toolbox for

Use with MATLAB (The Math Works Inc., 1997)

30.104 C.Y. Huang, J.A. Ceroni, S.Y. Nof: Agility of net-

worked enterprises: parallelism, error recovery and

conﬂict resolution, Comput. Ind. 42, 73–78 (2000)

30.105 M. Klein, S.C.-Y. Lu: Conﬂict resolution in cooper-

ative design, Artif. Intell. Eng. 4(4), 168–180 (1989)

30.106 M. Klein: Supporting conﬂict resolution in cooper-

ative design systems, IEEE Trans. Syst. Man. Cybern.

21(6), 1379–1390 (1991)

30.107 M. Klein: Capturing design rationale in concur-

rent engineering teams, IEEE Computer 26(1), 39–47

(1993)

30.108 M. Klein: Conﬂict management as part of an inte-

grated exception handling approach, Artif. Intell.

Eng. Des. Anal. Manuf. 9, 259–267 (1995)

30.109 X. Li, X.H. Zhou, X.Y. Ruan: Study on conﬂict

management for collaborative design system, J.

Shanghai Jiaotong University (English ed.) 5(2), 88–

93 (2000)

30.110 X. Li, X.H. Zhou, X.Y. Ruan: Conﬂict management in

closely coupled collaborative design system, Int. J.

Comput. Integr. Manuf. 15(4), 345–352 (2000)

30.111 S.Y. Nof: Tools and models of e-Work, Proc. 5th Int.

Conf. Simul. AI (Mexico City 2000) pp. 249–258

30.112 S.Y. Nof: Collaborative e-Work and e-Manufacturing:

challenges for production and logistics managers,

J. Intell. Manuf. 17(6), 689–701 (2006)

30.113 X.F. Zha, H. Du: Knowledge-intensive collaborative

design modeling and support part I: review, dis-

tributed models and framework, Comput. Ind. 57,

39–55 (2006)

30.114 X.F. Zha, H. Du: Knowledge-intensive collabora-

tive design modeling and support part II: system

implementation and application, Comput. Ind. 57,

56–71 (2006)

30.115 R. Solomonoff, A. Rapoport: Connectivity of ran-

dom nets, Bull. Mater. Biophys. 13, 107–117

(1951)

30.116 P. Erdos, A. Renyi: On random graphs, Publ. Math.

Debr. 6, 290–291 (1959)

30.117 P. Erdos, A. Renyi: On the evolution of random

graphs, Magy. Tud. Akad. Mat. Kutato Int. Kozl. 5,

17–61 (1960)

30.118 P. Erdos, A. Renyi: On the strenth of connectedness

of a random graph, Acta Mater. Acad. Sci. Hung. 12,

261–267 (1961)

30.119 D.J. Watts, S.H. Strogatz: Collective dynamics of

‘small-world’ networks, Nature 393(6684), 440–

442 (1998)

30.120 R. Albert, H. Jeong, A.L. Barabasi: Internet: Diam-

eter of the World-Wide Web, Nature 401(6749),

130–131 (1999)

30.121 A.L. Barabasi, R. Albert: Emergence of scaling

in random networks, Science 286(5439), 509–512

(1999)

30.122 A.Broder,R.Kumar,F.Maghoul,P.Raghavan,

S. Rajagopalan, R. Stata, A. Tomkins, J. Wiener:

Graph structure in the Web, Comput. Netw. 33(1),

309–320 (2000)

30.123 D.J. de Solla Price: Networks of scientiﬁc papers,

Science 149, 510–515 (1965)

30.124 G. Bianconi, A.L. Barabasi: Bose-Einstein conden-

sation in complex networks, Phys. Rev. Lett. 86(24),

5632–5635 (2001)

30.125 S.Y. Nof: Collaborative control theory for e-Work,

e-Production, and e-Service, Annu. Rev. Control

31(2), 281–292 (2007)

30.126 C.L. Yang, X. Chen, S.Y. Nof: Design of a production

conﬂict and error detection model with active pro-

tocols and agents, Proc. 18th Int. Conf. Prod. Res.

(2005)

Part C 30

“This page left intentionally blank.”

527

Automati

Part D

Part D Automation Design:

Theory and Methods for Integration

31 Process Automation

Thomas F. Edgar, Austin, USA

Juergen Hahn, College Station, USA

32 Product Automation

Friedrich Pinnekamp, Zurich, Switzerland

33 Service Automation

Friedrich Pinnekamp, Zurich, Switzerland

34 Integrated Human and Automation Systems

Dieter Spath, Stuttgart, Germany

Martin Braun, Stuttgart, Germany

Wilhelm Bauer, Stuttgart, Germany

35 Machining Lines Automation

Xavier Delorme, Saint-Etienne, France

Alexandre Dolgui, Saint-Etienne, France

Mohamed Essaﬁ, Saint-Etienne, France

LaurentLinxe,Hagondang,France

Damien Poyard, Saint-Etienne, France

36 Large-Scale Complex Systems

Florin-Gheorghe Filip, Bucharest, Romania

Kauko Leiviskä, Oulun Yliopisto, Finland

37 Computer-Aided Design, Computer-Aided

Engineering, and Visualization

Gary R. Bertoline, West Lafayette, USA

Nathan Hartman, West Lafayette, USA

Nicoletta Adamo-Villani, West Lafayette, USA

38 Design Automation for Microelectronics

Deming Chen, Urbana, USA

39 Safety Warnings for Automation

Mark R. Lehto, West Lafayette, USA

Mary F. Lesch, Hopkinton, USA

William J. Horrey, Hopkinton, USA

528

Automation Design: Theory and Methods for Integration. Part D After focusing in the previous part on

the details, principles and practices of the methodologies for automation design, the chapters in this part cover the

basic design requirements for the automation and illustrate examples of how the challenging issues can be solved

for the deign and integration of automation with respect to its main purpose: Continuous and discrete processes

and industries, such as chemicals, reﬁneries, machinery, and instruments; process automation safety; automa-

tion products such as circuit breakers, motors, drives, robots, and other components for consumer products; and

services, such as maintenance, logistics, upgrade and repair, remote support operations, and tools for service per-

sonnel; design issues and criteria when integrating humans with the automation; design techniques, criteria and

algorithms for ﬂow lines, such as assembly lines, transfer lines, machining lines; and the design of complex,

large-scale, integrated automation. Another view of automation integration is with computer-aided design (CAD)

and computer-aided engineering (CAE), which are themselves ﬁne examples of integrated automation, and are

required for the design of any automation and non-automation components, products, microelectronics, and ser-

vices. Concluding this part is the design for safety of automation, and of automation for safety, as they have

become an obligatory and mandatory concern of automation integrators.

529

Process Auto

31. Process Automation

Thomas F. Edgar, Juergen Hahn

The ﬁeld of process automation is concerned

with the analysis of dynamic behavior of chem-

ical processes, design of automatic controllers,

and associated instrumentations. Process automa-

tion as practised in the process industries has

undergone signiﬁcant changes since it was ﬁrst

introduced in the 1940s. Perhaps the most signif-

icant inﬂuence on the changes in process control

technology has been the introduction of inex-

pensive digital computers and instruments with

greater capabilities than their analog predeces-

sors. During the past 20 years automatic control

has assumed increased importance in the pro-

cess industries, which has led to the application of

more sophisticated techniques.

31.1 Enterprise View of Process Automation ... 529

31.1.1 Measurement and Actuation

(Level 1) ...................................... 529

31.1.2 Safety and Environmental/

Equipment Protection (Level 2) ...... 529

31.1.3 Regulatory Control (Level 3a).......... 530

31.1.4 Multivariable and Constraint

Control (Level 3b) ......................... 530

31.1.5 Real-Time Optimization (Level 4).... 530

31.1.6 Planning and Scheduling (Level 5).. 531

31.2 Process Dynamics

and Mathematical Models ..................... 531

31.3 Regulatory Control................................ 533

31.4 Control System Design........................... 534

31.4.1 Multivariable Control .................... 535

31.5 Batch Process Automation ..................... 538

31.6 Automation and Process Safety.............. 541

31.7 Emerging Trends .................................. 543

31.8 Further Reading ................................... 543

References .................................................. 543

31.1 Enterprise View of Process Automation

Process automation is used in order to maximize pro-

duction while maintaining a desired level of product

quality and safety and making the process more eco-

nomical. Because these goals apply to a variety of

industries, process control systems are used in facilities

for the production of chemicals, pulp and paper, metals,

food, and pharmaceuticals. While the methods of pro-

duction vary from industry to industry, the principles

of automatic control are generic in nature and can be

universally applied, regardless of the size of the plant.

In Fig. 31.1 the process automation activities are

organized in the form of a hierarchy with required func-

tions at the lower levels and desirable functions at the

higher levels. The time scale for each activity is shown

on the left side of Fig. 31.1. Note that the frequency of

execution is much lower for the higher-level functions.

31.1.1 Measurement and Actuation (Level 1)

Measurement devices (sensors and transmitters) and

actuation equipment (for example, control valves) are

used to measure process variables and implement the

calculated control actions. These devices are interfaced

to the control system, usually digital control equipment

such as a digital computer. Clearly, the measurement

and actuation functions are an indispensable part of any

control system.

31.1.2 Safety and Environmental/

Equipment Protection (Level 2)

The level 2 functions play a critical role by ensuring

that the process is operating safely and satisﬁesenviron-

Part D 31

530 Part D Automation Design: Theory and Methods for Integration

5. Planning and

scheduling

Demand forecasting,

supply chain management,

raw materials and product

planning/scheduling

(Days–months)

4. Real-time

optimization

Plant-wide and individual unit

real-time optimization, parameter

estimation, supervisory

control, data reconcilation

(Hours–days)

3b. Multivariable

and constraint

control

Multivariable control, model

predictive control

(Minutes–hours)

3a. Regulatory

control

PID control, advanced control

techniques, control loop

performance monitoring

(Seconds–minutes)

2. Safety,

environmental/

equipment

protection

Alarm management, emergency

shutdown

(< 1 second)

1. Measurement

and actuation

Process

Sensor and actuator validation,

limit checking

(< 1 second)

Fig. 31.1 The ﬁve levelsof process controland optimization inman-

ufacturing. Time scales are shown for each level [31.1]

mental regulations. Process safety relies onthe principle

of multiple protection layers that involve groupings

of equipment and human actions. One layer includes

process control functions, such as alarm management

during abnormal situations, and safety instrumented

systems for emergency shutdowns. The safety equip-

ment (including sensors and control valves) operates

independently of the regular instrumentation used for

regulatory control in level 3a. Sensor validation tech-

niques can be employed to conﬁrm that the sensors are

functioning properly.

31.1.3 Regulatory Control (Level 3a)

Successful operation of a process requires that key

process variables such as ﬂow rates, temperatures, pres-

sures, and compositions be operated at, or close to,

their set points. This level 3a activity, regulatory con-

trol, is achieved by applying standard feedback and

feedforward control techniques. If the standard control

techniques are not satisfactory, a variety of advanced

control techniques are available. In recent years, there

has beenincreased interestin monitoringcontrol system

performance.

31.1.4 Multivariable

and Constraint Control (Level 3b)

Many difﬁcult process control problems have two dis-

tinguishing characteristics: (1) signiﬁcant interactions

occur among key process variables, and (2) inequality

constraints exist for manipulated and controlled vari-

ables. The inequality constraints include upper and

lower limits; for example, each manipulated ﬂow rate

has an upper limit determined by the pump and con-

trol valve characteristics. The lower limit may be zero

or a small positive value based on safety considera-

tions. Limits on controlled variables reﬂect equipment

constraints (for example, metallurgical limits) and the

operating objectives for the process; for example, a re-

actor temperature may have an upper limit to avoid

undesired side reactions or catalyst degradation, and

a lower limit to ensure that the reaction(s) proceed.

The ability to operate a process close to a limit-

ing constraint is an important objective for advanced

process control. For many industrial processes, the op-

timum operating condition occurs at a constraint limit,

for example, the maximum allowed impurity level in

a product stream. For these situations, the set point

should not be the constraintvalue because a process dis-

turbance could force the controlled variable beyond the

limit. Thus, the set point should be set conservatively,

based on the ability of the control system to reduce the

effects of disturbances.

The standard process control techniques of level 3a

may not be adequate for difﬁcult control problems

that have serious process interactions and inequality

constraints. For these situations, the advanced control

techniques of level 3b, multivariable control and con-

straint control, should be considered. In particular, the

model predictive control (MPC) strategy was developed

to deal with both process interactions and inequality

constraints.

31.1.5 Real-Time Optimization (Level 4)

The optimum operating conditions for a plant are de-

termined as part of the process design, but during

plant operations, the optimum conditions can change

frequently owing to changes in equipment availabil-

ity, process disturbances, and economic conditions (for

example, raw materials costs and product prices). Con-

sequently, it can be very proﬁtable to recalculate the

Part D 31.1

Process Automation 31.2 Process Dynamics and Mathematical Models 531

optimum operating conditions on a regular basis. The

new optimum conditions are then implemented as set

points for controlled variables.

Real-time optimization (RTO) calculations are

based on a steady-state model ofthe plant and economic

data such as costs and product values. A typical objec-

tive for the optimization is to minimize operating cost

or maximize the operating proﬁt. The RTO calculations

can be performed for a single process unit and/or on

a plant-wide basis.

The level 4 activities also include data analysis to

ensure that the process model used in the RTO cal-

culations is accurate for the current conditions. Thus,

data reconciliation techniques can be used to ensure

that steady-state mass and energy balances are satisﬁed.

Also, the process model can beupdated using parameter

estimation techniques and recent plant data.

31.1.6 Planning and Scheduling (Level 5)

The highest levelof theprocess controlhierarchy iscon-

cerned with planning and scheduling operations for the

entire plant. For continuous processes, the production

rates of all products and intermediates must be planned

and coordinated, based on equipment constraints, stor-

age capacity, sales projections, and the operation of

other plants, sometimes on a global basis. For the inter-

mittent operation of batch and semibatch processes, the

production control problem becomes a batch schedul-

ing problem based on similar considerations. Thus,

planning and scheduling activities pose large-scale op-

timization problems that are based on both engineering

considerations and business projections.

The activities of levels 1–3ain Fig.31.1 are required

for all manufacturing plants, while the activities in lev-

els 3b–5 are optional but can be very proﬁtable. The

decision to implement one or more of these higher-level

activities depends very much on the application and the

company. The decision hinges strongly on economic

considerations (for example, a cost–beneﬁt analysis),

and company priorities for their limited resources, both

human and ﬁnancial. The immediacy of the activity de-

creases fromlevel 1to level5 inthe hierarchy. However,

the amount of analysis and the computational require-

ments increase from the lowest to the highest level. The

process control activities at different levels should be

carefully coordinated and require information transfer

from one level to the next. The successful implementa-

tion of these process control activities is a critical factor

in making plant operation as proﬁtable as possible.

31.2 Process Dynamics and Mathematical Models

Development of dynamic models forms a key compo-

nent for process automation, as controller design and

tuning is often performed by using a mathematical rep-

resentation of the process. A model can be derived

either from ﬁrst-principles knowledge about the system

or from past plant data. Once a dynamic model has been

developed, it can be solved for a variety of conditions

that include changes in the input variables or variations

in the model parameters. The transient responses of the

output variables are calculated by numerical integration

after specifying boththe initialconditions andthe inputs

as functions of time.

A large number of numerical integration techniques

are available, ranging from simple techniques (e.g., the

Euler and Runge–Kutta methods) to more complicated

ones (e.g., the implicit Euler and Gear methods). All of

these techniques represent some compromise between

computational effort (computing time) and accuracy.

Although a dynamic model can always be solved in

principle, for some situations it may be difﬁcult to gen-

erate useful numerical solutions. Dynamic models that

exhibit a wide range of time scales (stiff equations)

are quite difﬁcult to solve accurately in a reasonable

amount of computation time. Software for integrating

ordinary and partial differential equations is readily

available. Popular software packages include MAT-

LAB, Mathematica, ACSL, IMSL, Mathcad, and GNU

Octave.

For dynamic models that contain large numbers

of algebraic and ordinary differential equations, gen-

eration of solutions using standard programs has been

developed to assist in this task. A graphical user inter-

face (GUI) allows the user to enter the algebraic and

ordinary differential equations and related information

such as the total integration period, error tolerances,

the variables to be plotted, and so on. The simulation

program then assumes responsibility for:

1. Checking to ensure that the set of equations is ex-

actly speciﬁed

2. Sorting the equations into an appropriate sequence

for iterative solution

3. Integrating the equations

4. Providing numerical and graphical output.

Part D 31.2

532 Part D Automation Design: Theory and Methods for Integration

Examples of equation-oriented simulators used in the

process industries include DASSL, ACSL, gPROMS,

and Aspen Custom Modeler.

One disadvantage of equation-oriented packages is

the amount of time and effort required to develop all

of the equations for a complex process. An alternative

approach is to use modular simulation where prewrit-

ten subroutines provide models of individual process

units such as distillation columns or chemical reac-

tors. Consequently, this type of simulator has a direct

correspondence to the process ﬂowsheet. The modular

approach has the signiﬁcant advantage that plant-scale

simulations only require the user to identify the appro-

priate modules and to supply the numerical values of

model parameters and initial conditions. This activity

requires much less effort than writing all of the equa-

tions. Furthermore, the software is responsible for all

aspects of the solution. Because each module is rather

general in form, the user can simulate alternative ﬂow-

sheets for a complex process, for example, different

conﬁgurations of distillation towers and heat exchang-

ers, or different types of chemical reactors. Similarly,

alternative process control strategies can be quickly

evaluated. Some software packages allow the user to

add custom modules for novel applications.

Modular dynamic simulators have been available

since the early 1970s. Several commercial products

are available from Aspen Technology and Honeywell.

Modelica is an example of a collaborative effort that

provides modeling capability for a number of applica-

tion areas. These packages also offer equation-oriented

capabilities. Modular dynamic simulators are achiev-

ing a high degree of acceptance in process engineering

and control studies because they allow plant dynamics,

real-time optimization, and alternative control conﬁgu-

rations to be evaluated for an existing or a new plant.

They also can be used for operator training. This fea-

ture allows dynamic simulators to be integrated with

software for other applications such as control system

design and optimization.

While most processes can be accurately represented

by a set of nonlinear differential equations, a process

is usually operated within a certain neighborhood of its

normal operating point (steady state), thus the process

model can be closely approximated by a linearized ver-

sion of themodel. Alinear model is beneﬁcialbecause it

permits the use of more convenient and compact meth-

ods for representing process dynamics, namely Laplace

transforms. The main advantage of Laplace transforms

is that they provide a compact representation of a dy-

namic system that is especially useful for the analysis

of feedback control systems. The Laplace transform of

a set of linear ordinary differential equations is a set

of algebraic equations in the new variable s, called the

Laplace variable. The Laplace transform is given by

F(s) = L[ f (t)]=

∞



f(t)e

−st

dt , (31.1)

where F(s) is the symbol for the Laplace transform,

f(t) is some function of time, and L is the Laplace

operator, deﬁned by the integral. Tables of Laplace

transforms are well documented for common func-

tions [31.1]. A linear differential equation with a single

input u and singleoutput y canbe converted intoa trans-

fer function using Laplace transforms as follows

Y(s) =G(s)U(s) ,

(31.2)

where U(s) is the Laplace transform of the input vari-

able u(t), Y(s) is the Laplace transform of the output

variable y(t), and G(s) is the transfer function, obtained

from transforming the differential equation.

The transfer function G(s) describes the dynamic

characteristic of the process. For linear systems it is in-

dependent of the input variable and so it can readily

be applied to any time-dependent input signal. As an

example, the ﬁrst-order differential equation

dy(t)

+y(t) = Ku(t)

(31.3)

can be Laplace-transformed to

Y(s) =

τs +1

U(s) .

(31.4)

Note that the parameters K and τ, known as the pro-

cess gain and time constant, respectively, map into the

transfer function as unspeciﬁed parameters. Numerical

values for parameters such as K and τ have to be deter-

mined for controller design or for simulation purposes.

Several different methods for the identiﬁcation of model

parameters in transfer functions are available. The most

common approach is to perform a step test on the pro-

cess and collect the data along the trajectory until it

reaches steady state. In order to identify the parameters,

the form of the transfer function model needs to be pos-

tulated and the parameters of the transfer function can

be estimated by using nonlinear regression. For more

details onthe development of various transfer functions,

see [31.1].

Part D 31.2

Process Automation 31.3 Regulatory Control 533

31.3 Regulatory Control

When the components of a control system are con-

nected, their overall dynamic behavior can be described

by combining the transfer functions for each compo-

nent. Each block describes how changes in the input

variables of the block will affect the output variables

of the block. One example is the feedback/feedforward

control block diagram shown in Fig.31.2, which con-

tains the important components of a typical control

system, namely process, controller, sensor, and ﬁnal

control element. Regulatory control deals with treat-

ment of disturbances that enter the system, as shown

in Fig. 31.2. These components are discussed in more

detail below.

Most modern control equipment require a digital

signal for displays and control algorithms, thus the

analog-to-digital converter (ADC) transforms the trans-

mitter analog signal to a digital format. Because ADCs

may be relatively expensive if adequate digital resolu-

tion is required, incoming digital signals are usually

multiplexed. Prior to sending the desired control ac-

tion, which is often in a digital format, to the ﬁnal

control element in the ﬁeld, the desired control action

is usually transformed by a digital-to-analog (DAC)

converter to an analog signal for transmission. DACs

are relatively inexpensive and are not normally multi-

plexed. Widespread use of digital control technologies

has made ADCsandDACs standard parts of the control

system.

Sensors

The hardware components of a typical modern digi-

tal control loop shown in Fig.31.2 are discussed next.

The function of the process measurement device is to

Feedback

controller

Final control

element

Set point,

Manipulated

variable, U

Error,

Feedforward

controller

Measurement

device

Process

Disturbance, D

Controlled

variable, Y

Supervisory

control

Fig. 31.2 Block diagram of a process

sense the values, or changes in values, of process vari-

ables. The actual sensing device may generate, e.g.,

a physical movement, a pressure signal, or a millivolt

signal. A transducer transforms the measurement signal

from one physical or chemical quantity to another, e.g.,

pressure to milliamps. The transduced signal is then

transmitted to a control room through the transmission

line. The transmitter is therefore a signal generator and

a linedriver. Often thetransducer andthe transmitterare

contained in the same device.

The most commonly measured process variables

are temperature, ﬂow, pressure, level, and composition.

When appropriate, other physical properties are also

measured. The selection of the proper instrumentation

for a particular application is dependent on factors such

as: the type and nature of the ﬂuid or solid involved;

relevant process conditions; range, accuracy, and re-

peatability required; response time; installed cost; and

maintainability and reliability. Various handbooks are

available that can assist in selecting sensors for partic-

ular applications (e.g., [31.2]). Sensors are discussed in

detail on Chap.20.

Control Valves

Material and energy ﬂow rates are the most commonly

selected manipulated variables for control schemes.

Thus, good control valve performance is an essen-

tial ingredient for achieving good control performance.

A control valve consists of two principal assem-

blies: a valve body and an actuator. Good control

valve performance requires consideration of the process

characteristics and requirements such as ﬂuid charac-

teristics, range, shut-off, and safety, as well as control

Part D 31.3

534 Part D Automation Design: Theory and Methods for Integration

requirements, e.g., installed control valve characteris-

tics and response time. The proper selection and sizing

of control valves and actuators is an extensive topic in

its own right [31.2].

Controllers

The most commonly employed feedback controller in

the process industry is the proportional–integral (PI)

controller, which can be described by the following

equation

u(t) =

u +K

⎛

⎝

e(t)+



e(t



)dt



⎞

⎠

(31.5)

Note that the controller includes proportional as well as

integrating action. The controller has two tuning param-

eters: the proportional constant K

and the integral time

constant τ

. The integral action will eliminate offset for

constant load disturbances but it can potentially lead to

a phenomenon known as reset windup. When there is

a sustained error, the large integral term in (31.5) causes

the controller output to saturate. This can occur dur-

ing start-up of batch processes, or after large set point

changes or large sustained disturbances.

PI controllers make up the vast majority of con-

trollers that are currently used in the chemical process

industries. If it is important to achieve a faster response

that is offset-free, a PID (D =derivative) controller can

be utilized, described by the following expression

u(t) =

u +K

⎛

⎝

e(t)+



e(t



)dt



+τ

de(t)

⎞

⎠

(31.6)

The PID controller of (31.6) contains three tuning

parameters because the derivative mode adds a third

adjustable parameter τ

. However, if the process mea-

surement is noisy, the value of the derivative of the error

may change rapidly and derivative action will amplify

the noise, as a ﬁlter on the error signal can be employed.

In the 21st century, digital control systems are ubi-

quitous in process plants, mostly employing a discrete

(ﬁnite-difference) form of the PID controller equation

given by

u +K



Δt



i=0

+τ

−e

k−1

Δt



(31.7)

where Δt is the sampling period for the control calcu-

lations and k represents the current sampling time. If

the process and the measurements permit to chose the

sampling period Δt to be small then the behavior of the

digital PID controller will essentially be the same as for

an analog PID controller.

31.4 Control System Design

Traditionally, process design and control system de-

sign have been separate engineering activities. Thus, in

the traditional approach, control system design is not

initiated until after plant design is well underway and

major pieces of equipment may even have been or-

dered. This approach has serious limitations because

the plant design determines the process dynamics as

well as the operability of the plant. In extreme situa-

tions, the process may be uncontrollable, even though

the design appears satisfactory from a steady-state point

of view. A more desirable approach is to consider pro-

cess dynamics and control issues early in the process

design.

The two general approaches to control system de-

sign are:

1. Traditional approach. The control strategy and

control system hardware are selected based on

knowledge of the process, experience, and insight.

After the control system is installed in the plant,

the controller settings (such as in a PID controller)

are adjusted. This activity is referred to as controller

tuning.

2. Model-based approach. A dynamic model of the

process is ﬁrst developed that can be helpful in at

least three ways: (a) it can be used as the basis

for model-based controller design methods, (b) the

dynamic model can be incorporated directly in the

control law (for example, model predictive control),

and (c) the model can be used in a computer simu-

lation to evaluate alternative control strategies and

to determine preliminary values of the controller

settings.

For many simple process control problems con-

troller speciﬁcation is relatively straightforward and

Part D 31.4