Nof S.Y. Springer Handbook of Automation

Подождите немного. Документ загружается.

Process Automation 31.4 Control System Design 535

a detailed analysis or an explicit model is not required.

However, for complex processes, a process model is

invaluable both for control system design and for an

improved understanding of the process.

The major steps involved in designing and installing

a control system using the model-based approach are

shown in the ﬂowchart of Fig. 31.3. The ﬁrst step, for-

mulation of the control objectives, is a critical decision.

The formulation is based on the operating objectives for

the plants and the process constraints; for example, in

the distillation column control problem, the objective

might be to regulate a key component in the distil-

late stream, the bottoms stream, or key components in

both streams. An alternative would be to minimize en-

ergy consumption (e.g., heat input to the reboiler) while

meeting product quality speciﬁcations on one or both

product streams. The inequality constraints should in-

clude upper and lower limits on manipulated variables,

conditions that lead to ﬂooding or weeping in the col-

umn, and product impurity levels.

After the control objectives have been formulated,

a dynamic model of the process is developed. The dy-

namic model can have a theoretical basis, for example,

physical and chemical principles such as conserva-

tion laws and rates of reactions, or the model can be

developed empirically from experimental data. If exper-

imental data are available, the dynamic model should be

validated, with the data and the model accuracy char-

acterized. This latter information is useful for control

system design and tuning.

The next step in the control system design is to de-

vise an appropriate control strategy that will meet the

control objectives while satisfying process constraints.

As indicated in Fig.31.3, this design activity is based on

models and plant data. Finally the control system can

be installed, with ﬁnal adjustments performed once the

plant is operating.

31.4.1 Multivariable Control

In most industrial processes, there are a number of

variables that must be controlled, and a number of

variables can be manipulated. These problems are re-

ferred to as multiple-input multiple-output (MIMO)

control problems. For almost all important processes,

at least two variables must be controlled: product qual-

ity and throughput. Several examples of processes with

two controlled variables and two manipulated vari-

ables are shown in Fig.31.4. These examples illustrate

a characteristic feature of MIMO control problems,

namely, the presence of process interactions; that is,

Information from

existing plants

(if available)

Formulate

control

objectives

Management

objectives

Physical

and chemical

principles

Develop

process

model

Computer

simulation

Plant data

(if available)

Process control

theory

Experience with

existing plants

(if available)

Device control

strategy

Select control

hardware

and software

Install control

system

Adjust controller

settings

Final control

system

Computer

simulation

Vendor

information

Information baseEngineering activity

Fig. 31.3 Major steps in control system development [31.1]

each manipulated variable can affect both controlled

variables. Consider the inline blending system shown

in Fig. 31.4a. Two streams containing species A and B,

respectively, are to be blended to produce a produce

stream with mass ﬂow rate w and composition x,the

mass fraction of A. Adjusting either manipulated ﬂow

rate, w

or w

, affects both w and x.

Similarly, for the distillation column in Fig.31.4b,

adjusting either reﬂux ﬂow rate R or steam ﬂow S will

affect both distillate composition x

and bottoms com-

position x

. For the gas–liquid separator in Fig.31.4c,

adjusting the gas ﬂow rate G will have a direct effect on

pressure P and a slower, indirect effect on liquid level h

because changing the pressure in the vessel will tend to

change the liquid ﬂow rate L and thus affect h. In con-

trast, adjusting theother manipulatedvariable L directly

affects h but has only a relatively small and indirect

effect on P.

Part D 31.4

536 Part D Automation Design: Theory and Methods for Integration

a) Inline blending system

b) Destillation column

c) Gas–liquid separator

Gas

Liquid

Coolant

Steam

Feed

Fig. 31.4a–c Physical examples of multivariable control prob-

lems [31.1]

Pairing of a single controlled variable and a sin-

gle manipulated variable via a PID feedback controller

is possible, if the number of manipulated variables is

equal to the number of controlled variables. On the

other hand, more general multivariable control strate-

gies do not make such restrictions. MIMO control

problems are inherently more complexthan single-input

single-output (SISO) control problems because process

interactions occur between controlled and manipulated

variables. In general, a change in a manipulated vari-

able, say u

, will affect all of the controlled variables

, y

, ..., y

. Because of process interactions, selec-

tion of the best pairing of controlled and manipulated

variables for a multiloop control scheme can be a dif-

ﬁcult task. In particular, for a control problem with n

controlled variables and n manipulated variables, there

are n! possible multiloop control conﬁgurations. Hence

there is a growing trend to use multivariable control,

in particular an approach called model predictive con-

trol (MPC).

Model predictive control offers several important

advantages: (1) the process model captures the dynamic

and static interactions between input, output, and dis-

turbance variables, (2) constraints on inputs and outputs

are considered in a systematic manner, (3) the con-

trol calculations can be coordinated with the calculation

of optimum set points, and (4) accurate model predic-

tions can provide early warnings of potential problems.

Clearly, the success of MPC (or any other model-

based approach) depends on the accuracy of the process

model. Inaccurate predictions can make matters worse,

instead of better.

First-generation MPC systems were developed in-

dependently in the 1970s by two pioneering industrial

research groups. Dynamic matrix control (DMC)was

devised by Shell Oil [31.3], and a related approach was

developed by ADERSA [31.4]. Model predictive con-

trol has had a major impact on industrial practice; for

example, an MPC survey by Qin and Badgwell [31.5]

reported that there were over 4500 applications world-

wide by the end of 1999, primarily in oil reﬁneries and

petrochemical plants. In these industries, MPC has be-

come the method of choice for difﬁcult multivariable

control problems that include inequality constraints.

The overall objectives of an MPC controller are as

follows:

1. Prevent violations of input and output constraints

2. Drive some output variables to their optimal set

points, while maintaining other outputs within spec-

iﬁed ranges

3. Prevent excessive movement of the manipulated

variables

4. Control as many process variables as possible when

a sensor or actuator is not available.

A block diagram of a model predictive control sys-

tem is shown in Fig.31.5. A process model is used

to predict the current values of the output variables.

The residuals (the differences between the actual and

predicted outputs) serve as the feedback signal to a pre-

diction block. The predictions are used in two types of

MPC calculations that are performed at each sampling

instant: set-point calculations and control calculations.

Inequality constraints on the input and output variables,

such as upper and lower limits, can be included in either

type of calculation.

The model acts in parallel with the process and

the residual serves as a feedback signal, however, it

should be noted that the coordination of the control

and set-point calculation is a unique feature of MPC.

Part D 31.4

Process Automation 31.4 Control System Design 537

Furthermore, MPC has had a signiﬁcant impact on

industrial practice because it is more suitable for con-

strained MIMO control problems.

The set points for the control calculations, also

called targets, are calculated from an economic opti-

mization based on a steady-state model of the pro-

cess, traditionally, a linear steady-state model. Typical

optimization objectives include maximizing a proﬁt

function, minimizing a cost function, or maximizing

a production rate. The optimum values of set points

are changed frequently owing to varying process con-

ditions, especially changes in the inequality constraints.

The constraint changes are due to variations in process

conditions, equipment, and instrumentation, as well as

economic data such as prices and costs. In MPC the

set points are typically calculated each time the control

calculations are performed.

The control calculations are based on current mea-

surements and predictions of the future values of the

outputs. The predictions are made using a dynamic

model, typically a linear empirical model such as a mul-

tivariable version of the step response models that were

discussed in Sect.31.2. Alternatively, transfer func-

tion or state-space models can be employed. For very

nonlinear processes, it can be advantageous to pre-

dict future output values using a nonlinear dynamic

model. Both physical models and empirical models,

such as neural networks, have been used in nonlinear

MPC [31.5].

The objective of the MPC control calculations is

to determine a sequence of control moves (that is, ma-

nipulated input changes) so that the predicted response

moves to the set point in an optimal manner. The actual

output y, predicted output

y, and manipulated input u

are shown in Fig.31.6. At the current sampling instant,

denoted by k,theMPC strategy calculates a set of M

values of the input {u{k+i −1), i =1, 2,...,M}. The

set consists of the current input u(k)andM −1 future

inputs. The input is held constant after the M control

moves. The inputs are calculated so that a set of P pre-

dicted outputs {

y(k+i), i =1, 2,...,P} reaches the set

point in an optimal manner. The control calculations are

based on optimizing an objective function. The number

of predictions P is referred to as the prediction hori-

zon while the number of control moves M is called the

control horizon.

A distinguishing featureof MPCis itsreceding hori-

zon approach. Although a sequence of

M control moves

is calculated at each sampling instant, only the ﬁrst

move is actually implemented. Then a new sequence

is calculated at the next sampling instant, after new

Set-point

calculations

Control

calculations

Prediction

Predicted

outputs

Inputs

Residuals

Set points

(targets)

Inputs

Process

outputs

Model

outputs

Process

Model

Fig. 31.5 Block diagram for model predictive control [31.1]

measurements becomeavailable; again, only the ﬁrst in-

put move is implemented. This procedure is repeated at

each sampling instant.

In MPC applications, the calculated input moves

are usually implemented as set points for regulatory

control loops at the distributed control system (DCS)

level, such as ﬂow control loops. If a DCS control loop

has been disabled or placed in manual mode, the input

variable is no longer available for control. In this situa-

tion, the control degrees of freedom are reduced by one.

Even though an input variable is unavailable for con-

trol, it can serve as a disturbance variable if it is still

measured.

Before each control execution, it is necessary to

determine which outputs (controlled variables (CV)),

inputs (manipulated variables (MV)), and disturbance

variables (DVs) are currently available for the MPC

Past output

Predicted future output

Past control action

Future control action

Past Future

Set point (target)

Control horizon, M

Prediction horizon, P

k–1 kk+1 k+2 k+M–1

k+P

Sampling instant

Fig. 31.6 Basic concept for model predictive control

Part D 31.4

538 Part D Automation Design: Theory and Methods for Integration

calculations. The variables available for the control cal-

culations can change from one control execution time

to the next for a variety of reasons; for example, a sen-

sor may not be available owing to routine maintenance

or recalibration.

Inequality constraints on input and output variables

are important characteristics for MPC applications.

In fact, inequality constraints were a primary mo-

tivation for the early development of MPC. Input

constraints occur as a result of physical limitations on

plant equipment such as pumps, control valves, and

heat exchangers; for example, a manipulated ﬂow rate

might have a lower limit of zero and an upper limit

determined by the pump, control valve, and piping

characteristics. The dynamicsassociated withlarge con-

trol valves impose rate-of-change limits on manipulated

ﬂow rates.

Constraints onoutput variables are a key component

of the plant operating strategy; for example, a common

distillation column control objective is to maximize the

production rate while satisfying constraints on prod-

uct quality and avoiding undesirable operating regimes

such as ﬂooding or weeping. It is convenient to make

a distinction between hard and soft constraints. As the

name implies, a hard constraint cannot be violated at

any time. By contrast, a soft constraint can be violated,

but the amount of violation is penalized by a modiﬁ-

cation of the cost function. This approach allows small

constraint violations to be tolerated for short periods of

time [31.1].

31.5 Batch Process Automation

Batch processing is an alternative to continuous pro-

cessing. In batch processing, a sequence of one or more

steps, either in a single vessel or in multiple vessels,

is performed in a deﬁned order, yielding a speciﬁc

quantity of a ﬁnished product. Because the volume of

product is normally small, large production runs are

achieved by repeating the process steps on a prede-

termined schedule. In batch processing, the production

amounts are usually smaller than for continuous pro-

cessing; hence, it is usually not economically feasible

to dedicate processing equipment to the manufacture

of a single product. Instead, batch processing units are

organized so that a rangeof products(from a few to pos-

sibly hundreds) can be manufactured with a given set

of process equipment. Batch processing can be compli-

cated byhavingmultiple stages, multiple products made

from the same equipment, or parallel processing lines.

The key challenge for batch plants is to consistently

manufacture each product in accordance with its speci-

Run-to-run

control

Sequential

control

Logic

control

Control

during the

batch

Production

management

Equipment

control

Safety

interlocks

Fig. 31.7 Overview of a batch control system

ﬁcations while maximizing the utilization of available

equipment. Beneﬁts include reduced inventories and

shortened response times to make a specialty product

compared with continuous processing plants. Typically,

it is not possible to use blending of multiple batches in

order to obtain the desired product quality, so product

quality speciﬁcations must be satisﬁed by each batch.

Batch processingis widelyused to manufacture spe-

cialty chemicals, metals, electronic materials, ceramics,

polymers, food and agricultural materials, biochemicals

and pharmaceuticals, multiphasematerials/blends, coat-

ings, and composites – an extremely broad range of

processes and products. The unit operations in batch

processing are also quite diverse, and some are analo-

gous to operations for continuous processing.

In analogy with the different levels of plant control

depicted in Fig.31.1, batch control systems operate at

various levels:

•

Batch sequencing and logic controls (levels 1 and 2)

•

Control during the batch (level 3)

•

Run-to-run control (levels 4 and 5)

•

Batch production management (level 5).

Figure 31.7 shows the interconnections of the differ-

ent types of control used in a typical batch process.

Run-to-run control is a type of supervisory control

that resides principally in the production management

block. In contrast to continuous processing, the focus of

control shifts from regulation to set-point changes, and

sequencing of batches and equipment takes on a much

greater role.

Part D 31.5

Process Automation 31.5 Batch Process Automation 539

Batch control systems must be very versatile to be

able to handle pulse inputs and discrete input/output

(I/O) as well as analog signals for sensors and actuators.

Functional control activities are summarized as follows:

1. Batch sequencing and logic control: Sequencing of

control steps that follow a recipe involves, for ex-

ample, mixing of ingredients, heating, waiting for

a reaction to complete, cooling, and discharging the

resulting products.Transferof materials to and from

batch tanks or reactors includes metering of mater-

ials as they are charged (as speciﬁed by the recipe),

as well as transfer of materials at the completion of

the process operation. In addition to discrete logic

for the control steps, logic is needed for safety in-

terlocks to protect personnel, equipment, and the

environment from unsafe conditions. Process inter-

locks ensure that process operations can only occur

in the correct time sequence.

2. Control during the batch: Feedback control of

ﬂow rate, temperature, pressure, composition, and

level, including advanced control strategies, falls

in this category, which is also called within-the-

batch control [31.6]. In sophisticated applications,

this requires speciﬁcation of an operating trajectory

for the batch (that is, temperature or ﬂow rate as

a function of time). In simpler cases, it involves

tracking of set points of the controlled variables,

which includes ramping the controlled variables up

and down and/or holding them constant for a pre-

scribed period of time. Detection of when the batch

operations should be terminated (end point) may be

performed by inferential measurements of product

quality, if direct measurement is not feasible.

3. Run-to-run control: Also called batch-to-batch con-

trol, this supervisory function is based on ofﬂine

product quality measurements at the end of a run.

Operating conditions and proﬁles for the batch are

adjusted between runs to improve the product qual-

ity using tools such as optimization.

4. Batch production management: This activity entails

advising the plant operator of process status and

how to interact with the recipes and the sequential,

regulatory, and discretecontrols. Complete informa-

tion (recipes) is maintained for manufacturing each

product grade, including the names and amounts of

ingredients, process variable set points, ramp rates,

processing times, and sampling procedures. Other

database information includes batches produced on

a shift, daily, or weekly basis, as well as material

and energy balances. Scheduling of process units

is based on availability of raw materials and equip-

ment and customer demand.

Recipe modiﬁcations from one run to the next are com-

mon in many batch processes. Typical examples are

modifying the reaction time, feed stoichiometry, or re-

actor temperature. When such modiﬁcations are done

at the beginning of a run (rather than during a run),

the control strategy is called run-to-run control. Run-to-

run control is frequently motivated by the lack of online

measurements of the product quality during a batch run.

In batch chemical production, online measurements are

often not available during the run, but the product can

be analyzed by laboratory samples at the end of the run.

The processengineer mustspecify a recipe that contains

the values of the inputs (which may be time-varying)

that will meet the product requirements. The task of the

run-to-run controller is to adjust the recipe after each

run to reduce variability in the output product from the

stated speciﬁcations.

Batch run-to-run control is particularly useful to

compensate for processes where the controlled vari-

able drifts over time; for example, in a chemical

vapor deposition process the reactor walls may become

fouled owing to byproduct deposition. This slow drift

in the reactor chamber condition requires occasional

changes to the batch recipe in order to ensure that the

controlled variables remain on target. Eventually, the

reactor chamber must be cleaned to remove the wall

deposits, effectively causing a step disturbance to the

process outputs when the inputs are held constant. Just

as the run-to-run controller compensates for the drift-

ing process, it can also return the process to target after

a step disturbance change [31.7,8].

The Instrument Society of America (ISP) SP-88

standard deals with the terminology involved in batch

control [31.9]. There are a hierarchy of activities that

take place in a batch processing system [31.10]. At the

highest level, procedures identify how the products are

made, that is, the actions to be performed (and their or-

der) as well as the associated control requirements for

these actions. Operations are equivalent to unit opera-

tions in continuous processing and include such steps as

charging, reacting, separating, and discharging. Within

each operation are logical points called phases, where

processing can be interrupted by operator or computer

interaction. Examples of different phases include the

sequential addition of ingredients, heating a batch to

a prescribed temperature, mixing, and so on. Control

steps involvedirect commands toﬁnal control elements,

speciﬁed by individual control instructions in software.

Part D 31.5

540 Part D Automation Design: Theory and Methods for Integration

As an example, for {operation = charge reactant} and

{phase =add ingredient B}, the control steps would be:

(1) open the B supply valve, (2) total the ﬂow of B over

a period of time until the prescribed amount has been

added, and (3) close the B supply valve.

The term recipe has a range of deﬁnitions in batch

processing, but in general a recipe is a procedure with

the set of data, operations, and control steps required to

manufacture a particular grade of product. A formula

is the list of recipe parameters, which includes the raw

materials, processing parameters, and product outputs.

A recipe procedure has operations for both normal and

abnormal conditions. Each operation contains resource

requests forcertain ingredients (and their amounts). The

operations in the recipe can adjust set points and turn

equipment on and off. The complete production run for

a speciﬁc recipeis calleda campaign(multiple batches).

In multigrade batch processing, the instructions re-

main the same from batch to batch, but the formula can

be changedto yieldmodest variations in the product;for

example, in emulsion polymerization, different grades

of polymers are manufactured by changing the formula.

In ﬂexible batch processing, both the formula (recipe

parameters) and the processing instructions can change

from batch to batch. The recipe for each product must

specify both the raw materials required and how condi-

tions within the reactor are to be sequenced in order to

make the desired product.

Many batch plants, especially those used to man-

ufacture pharmaceuticals, are certiﬁed by the Interna-

tional Standards Organization (ISO). ISO 9000 (and

the related ISO standards 9001–9004) state that every

manufactured product should havean established,docu-

mented procedure, and the manufacturer should be able

to document that the procedure was followed. Compa-

nies must pass periodic audits to main ISO 9000 status.

Both ISO 9000 and the US Food and Drug Admin-

istration (FDA) require that only a certiﬁed recipe be

used. Thus, if the operation of a batch becomes ab-

normal, performing any unusual corrective action to

bring it back within the normal limits is not an option.

In addition, if a slight change in the recipe apparently

produces superior batches, the improvement cannot be

implemented unless the entire recipe is recertiﬁed. The

FDA typically requires product and raw materials track-

ing, so that product abnormalities can be traced back to

their sources.

Recently, in an effort to increase the safety, efﬁ-

ciency, and affordability of medicines, the FDA has

proposed a new framework for the regulation of phar-

maceutical development, manufacturing, and quality

assurance. The primary focus of the initiative is to

reduce variability through a betterunderstanding ofpro-

cesses than can be obtained by the traditional approach.

Process analytical technology (PAT) has become an

acronym in the pharmaceutical industry for design-

ing, analyzing, and controlling manufacturing through

timely measurements (i. e., during processing) of crit-

ical quality and performance attributes of raw and

in-process materials and processes, with the goal of

ensuring ﬁnal product quality. Process variations that

could possibly contribute to patient risk are determined

through modeling and timely measurements of critical

quality attributes, which are then addressed by pro-

cess control. In this manner processes can be developed

and controlled in such a way that quality of product is

guaranteed.

Semiconductor manufacturing is an example of

a large-volume batch process [31.7]. In semiconductor

manufacturing an integrated circuit consists of several

layers of carefully patterned thin ﬁlms, each chemically

altered to achieve desired electrical characteristics.

These devices are manufactured through a series of

physical and/or chemical batch unit operations simi-

lar to the way in which speciality chemicals are made.

From 30 to 300 process steps are typically required to

construct a set of circuits on a single-crystalline sub-

strate called a wafer. The wafers are 4 to 4–12inch

(100–300 mm) in diameter, 400–700μm thick, and

serve as the substrate upon which microelectronic cir-

cuits (devices) are built. Circuits are constructed by

depositing the thin ﬁlms (0.01–10μm) of material of

carefully controlled composition in speciﬁc patterns

and then etching these ﬁlms to exacting geometries

(0.35–10μm).

The main unit operations in semiconductor man-

ufacturing are crystal growth, oxidation, deposition

(dielectrics, silicon, metals), physical vapor deposition,

dopant diffusion, dopant-ion implantation, photolithog-

raphy, etch, and chemical–mechanical polishing. Most

processes in semiconductor manufacturing are semi-

batch; for example, in a single-waferprocessing tool the

following steps are carried out:

1. A robotic arm loads the boat of wafers

2. The machine transfers a single wafer into the pro-

cessing chamber

3. Gases ﬂow continuously and reaction occurs

4. The machine removes the wafer

5. The next wafer is processed.

When all wafers are ﬁnished processing, the operator

takes the boat of wafers to the next machine. All of

Part D 31.5

Process Automation 31.6 Automation and Process Safety 541

these steps are carried out in a clean room designed to

minimize device damage by particulate matter.

For a given tool or unit operationa speciﬁed number

of wafers are processed together in a lot, which is car-

ried in a boat. There is usually an extra slot in the boat

for a pilot wafer, which is used for metrology reasons.

A cluster tool refers to equipment which has several

single-wafer processing chambers. The chambers may

carry out the same process or different processes; some

vendors base their chamber designs on series operation,

while others utilize parallel processing schemes.

The recipe for the batch consists of the regulatory

set points and parameters for the real-time controllers

on the equipment. The equipment controllers are nor-

mally not capable of receiving a continuous set-point

trajectory. Only furnaces and rapid thermal processing

tools are able to ramp up, hold, and ramp down their

temperature or power supply. A recipe can consist of

several steps; each step processes a different ﬁlm based

on speciﬁc chemistry. The same recipe on the same type

of chamber may produce different results, due to differ-

ent processes used in the chamber previously. This lack

of repeatability across chambers is a big problem with

cluster tools or when a fabrication plant (fab) has multi-

ple machines of the same type, because it requires that

a fab keep track of different recipes for each chamber.

The controller translates the desired specs into a ma-

chine recipe. Thus, the fab supervisory controller only

keeps track of the product speciﬁcations.

Factory automation in semiconductor manufactur-

ing integrates the individual equipment into higher

levels of automation, in order to reduce the total cy-

cle time, increase fab productivity, and increase product

yield [31.8]. The major functions provided by the au-

tomation system include:

1. Planning of factory operation from order entry

through wafer production

2. Scheduling of factory resources to meet the produc-

tion plan

3. Modeling and simulation of factory operation

4. Generation and maintenance of process and product

speciﬁcation and recipes

5. Tracking of work-in-progress (WIP)

6. Monitoring of factory performance

7. Machine monitoring, control, and diagnosis

8. Process monitoring, control, and diagnosis.

Automation of semiconductor manufacturing in the

future will consist of meeting a range of technological

challenges. These include the needfor fasteryield ramp,

increasing cost pressures that compel productivity im-

provements, environmental safety and health concerns,

and shrinking device dimensions and chip size. The de-

velopment of 300 mm platforms in the last few years

has spawned equipment with new software systems and

capabilities. These systems will allow smart data col-

lection, storage, and processing on the equipment, and

transfer of data and information in a more efﬁcient

manner. Smart data management implies that data are

collected as needed and based upon events and metrol-

ogy results. As a result of immediate and automatic

processing of data, a larger fraction of data can be ana-

lyzed, and more decisions aredata driven.Newsoftware

platforms provide the biggest opportunity for a control

paradigm shift seen in the industry since the introduc-

tion of statistical process control.

31.6 Automation and Process Safety

In modern chemical plants, process safety relies on the

principle of multiple protection layers. A typical conﬁg-

uration is shown in Fig.31.8. Each layer of protection

consists of a grouping of equipment and/or human ac-

tions. The protection layers are shown in the order of

activation that occurs as a plant incident develops. In the

inner layer, the process design itself provides the ﬁrst

level of protection. The next two layers consist of the

basic process control system (BPCS) augmented with

two levels of alarms and operator supervision or in-

tervention. An alarm indicates that a measurement has

exceeded its speciﬁed limits and may require operator

action.

The fourth layer consists of a safety interlock

system (SIS), which is also referred to as a safety

instrumented system or as an emergency shutdown

(ESD) system. The SIS automatically takes corrective

action when the process and BPCS layers are unable

to handle an emergency; for example, the SIS could

automatically turn off the reactant pumps after a high-

temperature alarm occurs for a chemical reactor. Relief

devices such as rupture discs and relief valves pro-

vide physical protection by venting a gas or vapor if

overpressurization occurs. As a last resort, dikes are

located around process units and storage tanks to con-

tain liquid spills. Emergency response plans are used

Part D 31.6

542 Part D Automation Design: Theory and Methods for Integration

Community emergency response

Plant emergency response

Physical protection (dikes)

Physical protection

(relief devices)

Automatic action SIS or ESD

Critical alarms, operator

supervision, and manual intervention

Basic controls

process alarms, and

operator supervision

Process design

Note:

Protection layers for a typical process are shown in the order of activation

expected as a hazardous condition is approached

ESD = Emergency shutdown; SIS = Safety interlock system

Fig. 31.8 Typical layers of protection in a modern chemical

plant [31.11]

to address emergency situations and to inform the

community.

The functioningof the multiple layer protection sys-

tem can be summarized as follows [31.11]:

Most failures in well-designed and operated chem-

ical processes are contained by the ﬁrst one or two

protection layers. The middle levels guard against

major releases and the outermost layers provide

mitigation response to very unlikely major events.

For major hazard potential, even more layers may

be necessary.

It is evident from Fig.31.8 that automation plays an

important role in ensuring process safety. In particu-

lar, many of the protection layers in Fig. 31.8 involve

instrumentation and control equipment. The SIS oper-

ation is designed to provide automatic responses after

alarms indicate potentially hazardous situations. The

objective is to have the process reach a safe condition.

The automatic responses are implementedvia interlocks

and automatic shutdown and start-up systems. Distinc-

a) Low-level interlock

b) High-pressure interlock

Liquid

storage

tank

Gas

out

flare stack

Gas

storage

tank

LSL

PSH

LSL = Level switch low

= Pressure switch high

= Solenoid switch

Fig. 31.9a,b Two interlock conﬁgurations [31.1]

tions are sometimes made between safety interlocks and

process interlocks; process interlocks are used for less

critical situations to provide protection against minor

equipment damage and undesirable process conditions

such as the production of off-speciﬁcation product.

Twosimple interlock systemsare shown in Fig.31.9.

For the liquid storage system, the liquid level must stay

above a minimum value in order to avoid pump damage

such as cavitation. If the level drops below the speciﬁed

limit, the low-level switch (LSL) triggers both an alarm

and a solenoid (S), which acts as a relay and turns the

pump off. For the gas storage system in Fig. 31.9b, the

solenoid-operated valve is normally closed. However,

if the pressure of the hydrocarbon gas in the storage

tank exceeds a speciﬁed limit, the high-pressure switch

(PSH) activates an alarm and causes the valve to open

fully, thus reducing the pressure in the tank. For inter-

lock and other safety systems, a switch can be replaced

by a transmitter if the measurement is required. Also,

transmitters tend to be more reliable.

The SIS in Fig.31.9 serves as an emergency backup

system for the BPCS. TheSIS automaticallystarts when

a critical process variable exceeds speciﬁed alarm limits

that deﬁne the allowable operating region. Its initiation

Part D 31.6

Process Automation References 543

results in a drastic action such as starting or stopping

a pump or shutting down a process unit. Consequently,

it is used only as a last resort to prevent injury to people

or equipment.

It is very important that the SIS function indepen-

dently of the BPCS; otherwise emergency protection

will be unavailable during periods when the BPCS

is not operating (e.g., due to a malfunction or power

failure). Thus, the SIS should be physically sepa-

rated from the BPCS and have its own sensors and

actuators. Sometimes redundant sensors and actua-

tors are utilized; for example, triply redundant sensors

are used for critical measurements, with SIS actions

based on the median of the three measurements. This

strategy prevents a single sensor failure from crip-

pling SIS operation. The SIS also has a separate

set of alarms so that the operator can be notiﬁed

when the SIS initiates an action (e.g., turning on an

emergency cooling pump), even if the BPCS is not

operational.

31.7 Emerging Trends

Main emerging trends in process automation have to do

with process integration, information integration, and

engineering integration. The theme across all of these

is the need for a greater degree of integration of all

components of the process. A thorough discussion on

the implications of these trends and their challenges on

process automation can be found in this Handbook in

Chap.8.

31.8 Further Reading

•

A. Cichocki, H.A. Ansari, M. Rusinkiewicz,

D. Woelk: Workﬂow and Process Automation: Con-

cepts and Technology, 1st edn. (Springer, London

1997)

•

P. Cleveland: Process automation systems Control

Eng. 55(2), 65–74 (2008)

•

S.-L. Jämsä-Jounela: Future trends in process au-

tomation Annu. Rev. Control, 31(2), 211–220

(2007)

•

J. Love: Process Automation Handbook: A Guide

to Theory and Practice, 1st edn. (Springer, London

2007)

References

31.1 D.E. Seborg, T.F. Edgar, D.A. Mellichamp: Process

Dynamics and Control, 2nd edn. (Wiley, New York

2004)

31.2 T.F. Edgar, C.L. Smith, F.G. Shinskey, G.W. Gass-

man, P.J. Schafbuch, T.J. McAvoy, D.E. Se-

borg: Process control. In: Perry’s Chemical

Engineering Handbook, ed. by R.H. Perry,

D.W. Green, J.O. Maloney (McGraw-Hill, New York

2008)

31.3 C.R. Cutler, B.L. Ramaker: Dynamic matrix con-

trol – a computer control algorithm, Proc. Jt.

Auto. Control Conf., paper WP5-B (San Francisco

1980)

31.4 J. Richalet, A. Rault, J.L. Testud, J. Papon:

Model predictive heuristic control: applications

to industrial processes, Automatica 14,413–428

(1978)

31.5 S.J. Qin, T.A. Badgwell: A survey of industrial model

predictive control technology, Control Eng. Pract.

11, 733–764 (2003)

31.6 D. Bonvin: Optimal operation of batch reactors –

a personal view, J. Process Control 8, 355–368

(1998)

31.7 T.G. Fisher: Batch Control Systems: Design, Appli-

cation, and Implementation (ISA, Research Triangle

Park 1990)

31.8 J. Parshall, L. Lamb: Applying S88: Batch Control

from a User’s Perspective (ISA, Research Triangle

Park 2000)

31.9 T.F. Edgar, S.W. Butler, W.J. Campbell, C. Pfeif-

fer, C. Bode, S.B. Hwang, K.S. Balakrishnan,

J. Hahn: Automatic control in microelectronics

manufacturing: practices, challenges and possibil-

ities, Automatica 36, 1567–1603 (2000)

31.10 J. Moyne, E. del Castillo, A.M. Hurwitz (Eds.): Run to

Run Control in Semiconductor Manufacturing (CRC,

Boca Raton 2001)

31.11 AIChE Center for Chemical Process Safety: Guidelines

for Safe Automation of Chemical Processes (AIChE,

New York 1993)

Part D 31

“This page left intentionally blank.”