Control System Design

School of Engineering and Information Technology

Semester 2, 2016

Changing Study Timetable

We are behind schedule.

Assignment
Computer test 2 (Week 13)
Group Presentation (Week 14)

Value

Due Date

20 % 27 10 2016
5 % 3 11 2016

Report 2

10 % 10 11 2016

Final examination (closed-book

without computer)

35 % Exam period (12/11)

Outline
Introduction of control system design
Control objectives
Inventory control

Selection of
Controlled variables
Manipulated variables
Measured variables

Control-loop interactions
The remaining independent variables
Mass and energy inventory control
Optimisation example

Plant-wide control fundamental

A typical chemical plant flowsheet has multiple units

connected in serries or parallel.

Two main sections are

Reaction: reactors
Separation: distillation, evaporation, crystallisation

Others are for chemical and physical transformation,

e.g. heat exchanger network.

Integrated processes
Effect of material recycle
Effect of energy integration
Need to account for chemical inventories
4

Material recycle

Six basic essential reasons

Increase conversion as shown in an example, e.g. from 16% (no
recycle) to 100% (with recycle)
Improve economics: cheaper to build a reactor with incomplete
conversion
Improve yields, e.g. ABC Need to keep A low B is also
low Need recycle.
Provide thermal sink: Need to feed excess material so that
temperature is not too high.
Prevent side reactions
Control properties: e.g. Polymer chain, viscosity

Other process design techniques

Heat integration
Overhead vapour from one distillation column provides
energy for vaporizing liquid in the reboiler of another
column.

Chemical component inventories

Three types of chemical species in a plant: reactants, products
and inerts.
Material balance for each components must be satisfied.
Reactants fed to reactor must be consumed by reactions.
Un-reacted reactants must be recycled but must avoid building
up in the recycle stream
Avoid loss through purging.

Downstream method

Plant production rate established with exit stream flow

LC

LC

LC

FC

Set production rate

Individual unit will see disturbances coming from both

its downstream and upstream neighbours

Upstream method

Plant production rate established with inlet stream flow

LC

LC

LC

FC

Set production rate

Individual unit will only see disturbances coming from its

upstream neighbour Production rate is not fixed.

Example: Simple recycle problem

Reaction: A B

Process
A well-mixed CSTR
A perfect separator which completely separates A and B.

Information

Feed = Stream 1: 100 moles/hour

CSTR volume: 200 moles
Reaction rate: 0.6hr-1
Conversion: 16% A is converted B (single run)
Recycle: ratio = 5:1 recycle/feed

Mole balance for single run

Take 100 mol/hr of product B as a basis of calculation

Find the feed rate of A (mol/hr).

Composition of A in the reactor

100 knT X A 0.6 * 200 * X A X A 100 / 120 0.833

Freacted
Conversion
F fed

100
0.16
Fo

Freacted
mol
Fo
625
0.16
hr

To make 100 mol/hr of B, 625 mol/hr of A is fed into the

reactor but 525 mol/hr of A did not react.
F0, XA0= 1

knTXA

F0, XA

(5)

500 A

Distillation
column
(1)

(2)

(3)

100 A 600 A
Reactor

500 A
100 B

(4)

100 B

A simple recycle problem

1

Snowball effect
If the fresh feed stream is increased from 100
to 110 moles/hour
The recycle rate is almost doubled, from 500 to 1210
moles/hour

If the fresh feed stream is decreased from 100

to 90 moles/hour
The recycle rate is almost halved, from 500 to 270
moles/hour

(5)

1210 A

Distillation
column

HR
(1)
110 A

(2)

(3)

1320 A
Reactor

1210 A
110 B

(4)

110 B

A simple recycle problem

Inventory problem
1

Mole balance for recycle stream

Molar rate of component B leaving the reactor

(1 X A )( F0 R) knT X A

knT X A
F0 R
1 X A
knT X A F0 X A F0
knT X A
R
F0
1 X A
1 X A
F0 knT X A

F0
XA
knT

Mole balance for recycle stream (i)

F0
XA
knT

F0
(knT F0 ) F0
knT
R
F0
1
knT

Define Damkohler number

Da 1 R

knT
Da
F0

F0
R
Da 1

F0
R
knT F0

F0
R
knT
1
F0
nT = 200

k = 0.6

F0

R
50

35.7

90

270

100

500

110

1210

Control System Design

(Snowball effect)

Condenser

Reflux drum
LC

LC

F0, z0

HD
D, xD

V
R

Distillation
column

HR

F, z
Reactor
1. Fresh feet flow (valve)
2. Reactor level (reactor
effluent flow)
3. Bottom product purity
4. Distillate purity)

HB
LC

B, xB

Control System Design

Condenser

Reflux drum
LC

HD
D, xD

F0, z0

V
R

Distillation
column

HR
LC

F, z
FC

Reactor
1. Fresh feet flow (valve)
2. Reactor level (reactor
effluent flow)
3. Bottom product purity
4. Distillate purity)

HB
LC

B, xB

Control objectives

Control System Design (CSD) provides some insight in

designing a control system for either part or all of a
process plant.
Job of a process engineer is helping a control engineer.
The approach is based on some useful rules-of-thumb
= heuristics.
In the 1400's a law was set forth in England that a man was
allowed to beat his wife with a stick no thicker than his thumb.
Hence we have the rule of thumb.

Why
Safe operation: alarms, trips and interlocks
Steady state operation
Optimal operation: On- and off-line optimisation
1

Inventory control
All process plants are
Dynamic in nature
Operating according to dynamic M & E balances
Subject to disturbances

The plant operation can be mathematically

expressed in a lumped-parameter form, where
y is a vector of inventories (mass and energy)
u is a vector of manipulated variables (MV)
d is a vector of disturbances (DV)

dy
f ( y, u, d )
dt

Notes

Complete control of every yi is impossible.

Unable to measure or too costly to measure, e.g. in a 100
stage distillation column, every component hold-up in
every tray would have to be measured!!!
Lack of degree of freedom not enough ui.
Impractical control loops
Low sensitivity,
Slow dynamics,
Interactions

Effects of disturbances d and any unused manipulated

variables u should be minimised.
2

Regulatory design procedure

Select the controlled variables, a subset of y
from yi.
Select a manipulated variable ui for each yi
Check for interactions between yi and other yis,
and between yi and uis.

Reduce the effect of other ui and di

Maintain mass inventories first, then energy
inventories.
2

Selection of CV
Guideline 1
Always select state variables representing
inventories, which are not self-regulatory.
F1

F1

F2

Self-Regulatory
System

F2

Non Self-Regulatory
System
22

Selection of CV (i)
Guideline 1
A self-regulatory system has in-built regulating
behaviour.
dh
A F1 k h
dt
A non self-regulatory system
dh
Has no in-built regulating behaviour,
A
F1 F2
dt
Is known as pure integrator
Notes: most energy inventories are self-regulatory in
nature.

Selection of CV (ii)
Guideline 2
Select state variables which, although self-regulatory,
may exceed equipment or process constraints, e.g.
tanks may drain or overflow before new steady state
is attained.

Guideline 3
Select state variables which, although self-regulatory,
may seriously interact with other state variables, e.g.
a steam boiler supplying a large plant.
Steam pressure is self-regulatory
Fluctuation would affect all steam users

Selection of CV & MV
Guideline 4
Select state variables that are direct measures of
product quality or strongly affect it. Provide product
to the satisfaction of the market!!!

Guideline 5
The manipulated variables should affect the state
variable directly rather than indirectly.

Selection of MV

Vapour

Guideline 5
Liquid
Liquid

Steam

PV

MV

Vaporiser

d(mass accumulate d)
inflow outflow vaporized
dt
d(E accum. in coil)
E lost to vapour E gained from steam
dt

(E accum. in steam)
E lost to coil E gained from incoming steam
dt

Selection of MV (i)
Guideline 6
Sensitivity: The value of the gain between the
manipulated and controlled variables should be as
large as possible, e.g. A or B to control the level?
500 kg/hr

A 480 kg/hr
B 20 kg/hr

Level Control Example

Selection of MV (ii)
Guideline 7
Speed of response: Any delays or lags associated
with a possible MV should be small compared to the
state variable time constant, e.g. inflow or outflow to
control the level of liquid accumulated at the bottom of
the column?

Packed Tower

Selection of MV (iv)
Guideline 8
Interactions with other balances: The extent of
interactions with other balances should be minimised,
e.g. applying Relative Gain Array method (Bristol,
1966)

Guideline 9
Avoiding recycling of disturbances: Choose an outlet
or utility stream if possible - pass disturbances
downstream, e.g. R or D to control the level of the
overhead collector?

Selection of measured variables

It is essential that CVs or variables closely related to CVs

can be measured, e.g. weight, level, pressure are relatively
easy to measure; concentration is not!

FF control can only be done if disturbances are measured

Guideline 10
Sensitivity: The selected measured variable must be
sensitive to underlying changes in the state variable.

Guideline 11
Select measurement points that minimise time delays and
time constants.

Control system for a recycle process

Feed
HCL/Water

LC

Receiving
Drum

HCL Vapour
Water
Vapour

Stripping
Column
Steam

LC

Steam
LC

Evaporator

Control-loop interactions (i)

The system cannot be stabilized because it exhibits severe

recycling! Why?

Does the control system for each unit in isolation obey all
guidelines?

Consider an increase in the inlet flowrate.

Receiving drum: Feed level outflow

Stripping column: Inflow level outflow

Evaporator: Inflow level recycle

Control-loop interactions (ii)

The documentation of the control system for each is on

separate P&ID. It is not easy to review the whole system.

A slight modification to the control system will allow a stable

operation.

Consider an increase in the inlet flowrate.

Receiving drum: Feed level outflow

Stripping column: Inflow level outflow

Evaporator: Inflow level steam

Modified control system

Feed
HCL/Water

LC

Receiving
Drum

HCL Vapour
Water
Vapour

Stripping
Column
Steam

LC

Steam
LC

Evaporator

FC

Remaining independent variables

Desirable actions to reduce possible

sources of disturbances
Equipment design or arrangements can be
modified.
Control loops can be installed. Advanced
control techniques can be applied, e.g.
feedforward, cascade, ratio controls.
Operating procedures may be specified.

Mass inventory control

Selection of controlled variables for mass
inventory control is influenced by
Using mechanical energy rather than gravity to
move materials. non self-regulatory so must be
controlled
Minimising of vessel size self-regulatory +
constraints so must be controlled
Relatively small design margins on pressure
vessels self-regulatory + constraints so must
be controlled

Energy inventory control

Energy inventory control should be developed
after appropriate mass inventories are
controlled.
Energy inventories are self-regulatory, but
equipment and process constraints may dictate
their selections as controlled variables.
Selection of manipulated variables for energy
inventory control is influenced by the previously
discussed guidelines.

Control system for product quality

It is the key operating objectives of a process.
There are no guidelines.
Controlling product quality can only achieved by
adjusting appropriate set-points of inventory
controllers because there is no MV left after
controlling inventories.
Controlling quality depend pre-conditions
Ability to reduce the effects of disturbances
Good control of inventories within the system

Control system for product quality (ii)

Four approaches to measuring product quality
Direct measurement, e.g. on-line analysers to
measure composition
Inferred measurement, e.g. the salt content or sugar
contents in food products are measured instead of
tasting
Model-based measurement, e.g. using a mathematical
model to predict the quality of the product
Faith, e.g. avoidance of disturbances, good inventory
control and experience (a prayer, perhaps!)

Example - CSD

This is optimisation!!!!

41

Special optimisations (i)

Dynamic optimisation = Multiobjective optimisation

So far only single OF has been dealt with and the problem runs
in steady state mode (no differential equation, no time is
involved).

In dynamic optimisation, the problem runs in dynamic mode and

multivariable OF must be dealt with or multivariable solution must
be obtained.

Example (MATLAB, optimisation section)

Using function lsqnonlin to find the optimal values of the control

parameters of a PID controller.

OF is to minimise the errors between input and output signals.

42

Special optimisations (ii)

Example (MATLAB, optimisation section)

Using function lsqnonlin to find the optimal values of the control

parameters of a PID controller.

OF is to minimise the errors between input and output signals.

Dynamic problem must be modelled in SIMULINK or dealt with

odes in MATLAB.

Multiobjective function because of minimising the errors at many

time steps (not a single time step).

Example of tuning a controller

g ( s)

6
(2s 1)(4s 1)(6s 1)

Three tanks in series having the transfer function

Implementing a PID controller with the initial parameters.

K P 0.3, K I 0.7, K D 0.1

Tuning the controller, using a Simulink model with:

lsqnonlin or fminimax
43

Exam (09)

A company has two grades of inspectors, 1 and 2 who are to

be assigned to a quality control inspection. It is required that at
least 1800 valves be inspected per 8-hr day. Grade-1
inspectors can check at the rate of 25 valves per hour, with an
accuracy of 98%. Grade-2 inspectors can check at the rate of
15 valves per hour, with an accuracy of 95%. The wage of a
grade-1 inspector is $4.00/h, while that of a grade-2 inspector
is $3.00/h. Each time an error is made by an inspector, the
cost to the company is $2.00. The company has available for
the inspection job 8 grade-1 inspectors and 10 grade-2
inspectors. The company wants to determine the optimal
assignment of inspectors that will minimize the total cost of the
inspection.
44

Exam (ii)

Formulate a complete optimisation problem. The problem should

be in a standard format for MATLAB code. (10 marks)
Do not solve the problem but select the MATLAB function used to
solve the problem and provide the reason for selection. (2 marks)

In a more general problem, let the wage of a grade-2 inspector

be Z and the rate of inspection be S. There is a relationship
between S and Z shown below, where Z 2.5 and Z 2.4.
S 0.19e1.22Z

Formulate a complete optimisation problem. The problem should

be in a standard format for MATLAB code. (6 marks)
Do not solve the problem but select the MATLAB function used to
solve the problem and provide the reason for selection. (2 marks)
Fmincon, fminunc, fsolve, lsqnonlin
45