Practical Process Control

Using Control Station

Prof. Doug Cooper
Chemical Engineering Dept.
University of Connecticut (Storrs)

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 1. Fundamental Principles of Process Control
Motivation for Automatic Process Control

 Safety First:
– people, environment, equipment

 The Profit Motive:

– meeting final product specs
– minimizing waste production
– minimizing environmental impact
– minimizing energy use
– maximizing overall production rate

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 “Loose” Control Costs Money
Pro ces s : G rav i ty D rai n ed T an k C o n t ro l l er: M an u al M o d e

PM r oo rf ei t a Pb r l oe f iOt a pb el er a Ot i op ne r a t i o n
4 .470 operating constraint
Pro ces s : G rav i ty D rai n ed T an k C o n t ro l l er: M an u al M o d e

44 .2
.260

44 .0
.050

poor control means

33 .8
.840 large variability, so
the process must be
operated in a less
M o re

33 .6
.630
process variable profitable region

3 .4 60 80 100 120 140

Tim e (m in s)

 It takes 8more
0
material
100
toe1 2(m0make
Tim in s) a1 4product
0
thicker, so greatest
160

profit is to operate as close to the minimum thickness

constraint as possible without going under
 It takes more processing to remove impurities, so greatest
profit is to operate as close to the maximum impurities
constraint as you can without going over
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Tight Control = Most Profitable Operation

M o r e PM r o fr iet a Pb l re o Of i t pa eb r al e t i oO n p e r a t i o n
P ro ces s : G rav i t y D rai n ed T an k C o n t ro l l er: M an u al M o d e

4 .270 operating constraint

P ro ces s : G rav i t y D rai n ed T an k C o n t ro l l er: M an u al M o d e

44 .2
.060

43 .0
tight control permits
.850
operation near the
process variable
constraint, which
33 .8
.640
means more profit

33 .6
.430

3 .4 80 100 120 140 160

T im e (m in s)

80 100 120 140 160

 A well controlled process has less variability in the
T im e (m in s)

measured process variable, so the process can be

operated close to the profitable constraint

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Consider heating a house

thermostat
controller
set point
TC TT
temperature heat loss
sensor/transmitter (disturbance)
control
signal

fuel flow
furnace
valve

Terminology  Control Objective

Measured Process Variable
Set point
Controller Output
Manipulated Variable
Disturbances
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Automatic Control is
Measurement  Computation  Action

 Is house cooler than set point? ( TSetpoint  Thouse > 0 )

Action  open fuel valve

 Is house warmer than set point? ( TSetpoint  THouse < 0 )

Action  close fuel valve

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Components of a Control Loop
Home heating control block diagram

controller manipulated house

controller output fuel flow to temperature
error signal furnace
Set Point Home Heating
+ Thermostat Fuel Valve
- Process

Heat Loss
house temperature Disturbance
measurement
signal

Temperature
Sensor/Transmitter

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 General Control Loop Block Diagram

manipulated measured
controller controller process process
error output variable variable
Final
Set Point
+ Controller Control Process
- Element

Disturbance

feedback
signal

Measurement
Sensor/Transmitter

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Examples Used in This Workshop
 Measurement Sensors:
temperature, pressure, pressure drop, level, flow
density, concentration

 Final Control Element:

solenoid, valve, variable speed pump or compressor,
heater or cooler

 Automatic Controllers:
on/off, PID, cascade, feed forward, model-based
Smith predictor, multivariable, sampled data,
parameter scheduled adaptive control

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Workshop Goals
 learn why understanding the dynamic behavior of a process is
fundamental to controlling it
 practice methods of collecting and analyzing process data to gain
this all important understanding of process dynamics
 learn what "good" or "best" control performance means for a
particular process
 understand the computational methods behind the popular
controllers and learn when and how to use each
 learn how controller tuning parameters impact performance and
how to determine values for these parameters
 understand the limitations and pitfalls of the different controllers
and learn how to turn this to your advantage

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Thought Experiment: Cruise Control in a Car

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Cruise Control in a Car
 Control Objective: maintain car velocity
 Measured Process Variable: car velocity
 Manipulated Variable: pedal angle, flow of gas
 Controller Output: signal to actuator that adjusts
gas flow
 Set point: desired velocity
 Disturbances: hills, wind, passing trucks....

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 2. Hands-On Case Studies
Gravity Drained Tanks

manipulated variable
controller output
.

measured
process variable level sensor
& controller

disturbance
variable

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Hands-On Case Studies
Heat Exchanger

disturbance
variable cooling
flow exit
manipulated variable

controller
output

temperature sensor
& controller
measured
process variable

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 3. Graphical Modeling of Dynamic Process Data
Process Behavior and Controller Tuning
 Consider cruise control for a car vs a truck
– how quickly can each accelerate or decelerate
– what is the effect of disturbances (wind, hills, etc.)
 Controller (gas flow) manipulations required to maintain set point
velocity in spite of disturbances (wind, hills) are different for a car
and truck because the dynamic behavior of each "process" is different
Dynamic behavior 
how the measured process variable responds over time to changes in
the controller output and disturbance variables

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Understanding Dynamic Process Behavior
 To learn about the dynamic behavior of a process, analyze
measured process variable test data
 Process variable test data can be generated by suddenly
changing the controller output signal
 Be sure to move the controller output far enough and fast
enough so that the dynamic behavior of the process is clearly
revealed as the process responds
 The dynamic behavior of a process is different as operating
level changes (nonlinear behavior) so collect process data at
normal operating levels (design level of operation)

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Modeling Dynamic Process Behavior
 The best way to understand process data is through modeling
 Modeling means fitting a first order plus dead time (FOPDT)
dynamic process model to the data set:
dy(t )
P  y (t )  K P u (t   P )
dt
where:
y(t) is the measured process variable
u(t) is the controller output signal

 The FOPDT model is low order and linear so it can only

approximate the behavior of real processes

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Modeling Dynamic Process Behavior
 When a first order plus dead time (FOPDT) model is fit to
dynamic process data
dy(t )
P  y (t )  K P u (t   P )
dt

 The important parameters that result are:

– Steady State Process Gain, KP
– Overall Process Time Constant, P
– Apparent Dead Time, P

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 The FOPDT Model is All Important
 model parameters (KP, P and P) are used in correlations to
compute initial controller tuning values
 sign of KP indicates the action of the controller
(+KP  reverse acting; KP  direct acting)
 size of P indicates the maximum desirable loop sample time (be
sure sample time T  0.1P)
 ratio P /P indicates whether MPC (Smith predictor) would show
benefit (useful if P  P)
 model becomes part of the feed forward, Smith predictor,
decoupling and other model-based controllers

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Step Test Data and Dynamic Process Modeling
O peO npeLnoop S teSpteTpeTste st
L oop

V a r ia b le
V a r ia b le
P ro Pc eros sc :e sCsu: sCt ou m
s t oPmro Pc eros sc e s s C o nCt oronltlro
e r:l l M
e r:a nMu aanl uMa lo M
d eo d e

60 60
Step Test
C o n t r o l l e r O u t pP ur to c e s s
C o n t r o l l e r O u t pP ur to c e s s
55 55

50 50

60 60

55 55

50 50

0 0 5 5 10 10 15 15 20 20
T im Te im( me in( ms )in s )

 Process starts at steady state

 Controller output signal is stepped to new value
 Measured process variable allowed to complete response
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Process Gain From Step Test Data
 KP describes how much the measured process variable, y(t), changes
in response to changes in the controller output, u(t)
 A step test starts and ends at steady state, so KP can be computed
from plot axes

Steady State Change in the Measured Process Variable, y (t )

KP 
Steady State Change in the Controller Output, u (t )

where u(t) and y(t) represent the total change from initial to final
steady state
 A large process gain means the process will show a big response to
each control action

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 KP for Gravity Drained Tanks

V a r ia b le
G ra vi t y D ra i ne d T a nks - O pe n L oop S t e p T e s t
P r o c e s s : G r a v i t y D ra i n e d T a n k C o n t ro l l er: M a n u a l M o d e
3.0
2.8
2.6
y = (2.88 - 1.93) m
C o n t r o l l e r O u t Pp ru o t c e s s
2.4
2.2
2.0
1.8

60

55 u = (60 - 50) %

50

45
8 9 10 11 12 13 14 15 16 17 18
T im e ( m ins )

y 2.88  1.93 m m
KP    0.095
u 60  50% %

Steady state process gain has a:

size (0.095), sign (+0.095), and units (m/%)
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Overall Time Constant From Step Test Data
O peO npeLnoop S teSpteTpeTste st
L oop

V a r ia b le
V a r ia b le
P ro Pc eros sc :e sCsu: sCt ou m
s t oPmro Pc eros sc e s s C o nCt oronltlro
e r:l l M
e r:a nMu aanl uMa lo M
d eo d e

60 60
Step Test

C o n t r o l l e r O u t pP ur to c e s s
C o n t r o l l e r O u t pP ur to c e s s
55 55

50 50

60 60

55 55

50 50

0 0 5 5 10 10 15 15 20 20
T im Te im( me in( ms )in s )

Time Constant P describes how fast the measured process

variable, y(t), responds to changes in the controller output, u(t)
P is how long it takes for the process variable to reach 63.2% of
its total change, starting from when the response first begins

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 P for Gravity Drained Tanks
1) Locate where the measured process variable first shows a clear initial
response to the step change – call this time tYstart
From plot, tYstart = 9.6 min
V a r ia b le

G ra vi t y D ra i ne d T a nks - O pe n L oop S t e p T e s t
P r o c e s s : G r a v i t y D ra i n e d T a n k C o n t ro l l er: M a n u a l M o d e
3.0
2.8
2.6
C o n t r o l l e r O u t Pp ru o t c e s s

2.4
2.2
2.0
1.8

60

55

50

45
8 9 10 11 12 13 14 15 16 17 18
T im e ( m ins )

tYstart
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 P for Gravity Drained Tanks
2) Locate where the measured process variable reaches y63.2, or
where y(t) reaches 63.2% of its total final change
Label time t63.2 as the point in time where y63.2 occurs
V a r ia b le

G ra vi t y D ra i ne d T a nks - O pe n L oop S t e p T e s t
P ro ce s s : G rav i ty D rai n ed T an k C o n t ro l l er: M an u al M o d e
3.0
2.8
y63.2
2.6
C o n t r o l l e r O u t Pp ru o t c e s s

2.4
2.2
2.0
1.8

60

55

50

45
8 9 10 11 12 13 14 15 16 17 18
T im e ( m ins )

tYstart t63.2
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 P for Gravity Drained Tanks

V a r ia b le
G ra vi t y D ra i ne d T a nks - O pe n L oop S t e p T e s t
P r o c e s s : G r a v i t y D ra i n e d T a n k C o n t ro l l er: M a n u a l M o d e
3.0
2.8
y63.2 = 2.53 m
2.6
y = 0.95 m
C o n t r o l l e r O u t Pp ru o t c e s s
2.4
2.2
2.0
1.8

60

55

50

45
8 9 10 11 12 13 14 15 16 17 18
T im e ( m ins )

tYstart t63.2

 y(t) starts at 1.93 m and shows a total change y = 0.95 m

 y63.2 = 1.93 m + 0.632(y)
= 1.93 m + 0.632(0.95 m) = 2.53 m
 y(t) passes through 2.53 m at t63.2 = 11.2 min
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 P for Gravity Drained Tanks
- The time constant is the time difference between tYstart and t63.2
- Time constant must be positive and have units of time
From the plot: P = t63.2  tYstart = 11.2 min  9.6 min = 1.6 min
V a r ia b le

G ra vi t y D ra i ne d T a nks - O pe n L oop S t e p T e s t
P r o c e s s : G r a v i t y D ra i n e d T a n k C o n t ro l l er: M a n u a l M o d e
3.0
2.8
y63.2 = 2.53 m
2.6
y = 0.95 m
C o n t r o l l e r O u t Pp ru o t c e s s

2.4
2.2
2.0
1.8

60

55
P = 1.6 minutes
50

45
8 9 10 11 12 13 14 15 16 17 18
T im e ( m ins )

tYstart t63.2
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Apparent Dead Time From Step Test Data
 P is the time from when the controller output step is made until
when the measured process variable first responds
 Apparent dead time, P, is the sum of these effects:
– transportation lag, or the time it takes for material to travel
from one point to another
– sample or instrument lag, or the time it takes to collect analyze
or process a measured variable sample
– higher order processes naturally appear slow to respond

 Notes:
– Dead time must be positive and have units of time
– Tight control in increasingly difficult as P  0.7P
– For important loops, work to avoid unnecessary dead time

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 P for Gravity Drained Tanks

V a r ia b le
G ra vi t y D ra i ne d T a nks - O pe n L oop S t e p T e s t
P r o c e s s : G r a v i t y D ra i n e d T a n k C o n t ro l l er: M a n u a l M o d e
3.0
2.8
2.6

C o n t r o l l e r O u t Pp ru o t c e s s
2.4
2.2
2.0
1.8

60

55

P = 0.4 minutes

50

45
8 9 10 11 12 13 14 15 16 17 18
tUstep T im e ( m ins )

tYstart

P = tYstart  tUstep
= 9.6 min  9.2 min
= 0.4 min
Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Workshop 1:
Exploring Dynamics of Gravity Drained Tanks

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Processes Have Time-Varying Behaviors
 The predictions of a FOPDT model are constant over time

 But real processes change every day because

– surfaces foul or corrode
– mechanical elements like seals or bearings wear
– feedstock quality varies and catalyst activity drifts
– environmental conditions like heat and humidity change

 So the values of KP, P, P that best describe the dynamic

behavior of a process today may not be best tomorrow

 As a result, controller performance will degrade with time and

periodic retuning may be required

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Processes Have Nonlinear Behaviors
 The predictions of a FOPDT model are constant as operating level changes
 The response of a real process varies with operating level
V a r ia b le

E xa m pl e N onl i ne a r B e ha vi or
P ro c e s s : C u s t o m P ro c e s s C o n t ro l l er: M a n u a l M o d e
80
response shape is different
C
70 at different operating levels
C o n t r o l l e r O u t Pp ru o t c e s s

60
B
50
A
70
65
60
55 even though controller
50 output steps are the same
2 4 6 8 10 12 14 16 18 20
T im e ( tim e units )

Copyright © 2002
Douglas J. Cooper
All Rights Reserved
 Gravity Drained Tanks is Nonlinear

L e v e l (m )
N onl i ne a r B e ha vior of G ra vi t y D ra i ne d T a nks
M o d e l : F i rs t O r d e r P l u s D e a d T i m e ( F O P D T ) F i l e N a m e : T E S T .D A T
7
6 nonlinear process
u t e( %a )s u r e d

5
4 variable response
3
2 constant parameter
1 FOPDT model
C o n tro ll e r O u tp M

90
80
70
60
50
40
equal u’s

0 5 10 15 20 25 30 35 40 45
T im e

G a i n (K ) = 0 .0 7 5 , T i m e C o n s t a n t ( T 1 ) = 1 .1 5 , D e a d T i m e ( T D ) = 0 .5 3 S S E : 2 0 6 .0

A controller should be designed for

a specific level of operation!
Copyright © 2002
Douglas J. Cooper
All Rights Reserved