“The Six"

Fundamental Control Strategies Every

Process Control Developer Must Know
Standards
Certification
Education & Training
Publishing
Conferences & Exhibits
About the Presentation

• About the Presenter

• What do we mean by “Control Strategies”?
• Six Fundamental Control Strategies
– PID Feedback Control
– Feedforward Control
– Cascade Control
– Split-Range Control
– Ratio Control
– Override Select Control

• Going beyond these strategies

About the Presenter

• Dale E. Reed, CAP

– B.S. Chem. Eng., Rensselaer Polytechnic Institute, 1980.
– Current Position: Principal Engineer, PlantPAx System
Engineering, at Rockwell Automation in Mayfield Heights, Ohio
– With Allen-Bradley Company / Rockwell Automation since 1984.

• ISA / Cleveland Section Member since August 1980

– Currently serve as Delegate and Standards and Practices Chair
– Section Secretary, 2013-2014; Section President, 2015-2016
– Serve on ISA18, ISA101, ISA106 standards committees
– ISA S&P Board of Directors since January 2016

• ISA Certified Automation Professional since July 2005

What Do We Mean by
“Control Strategy”?

• Here’s a process!
– How are you going
to control the level?

“A control strategy is a planned

set of controls, derived from
current product and process
understanding, that assures
process performance and
product quality.”
U.S. FDA, ICH Q10
Adjust the Inflow to
Maintain Tank Level

• We could use a
float to sense the
tank level.
– If the level goes
low, increase the
inflow
– If the level goes
high, decrease
the inflow.
What If the Demand
Decreases?

• Outflow decreases
• Level starts to increase
• Float rises
• Inlet valve closes down
• Inlet flow decreases
• System stabilizes at
a reduced flow rate
What If the Demand
Increases?

• Outflow increases
• Level starts to decrease
• Float falls
• Inlet valve opens up
• Inlet flow increases
• System stabilizes at
an increased flow rate
The Essence of
Feedback Control

• The Process Variable (PV), in this case the level, has to

deviate from its target value, the Setpoint (SP), for any
control action to take place.
• The difference between the PV and SP is called the
Error (E).
• The sign of the Error depends on which way we have to
move the valve when the level is too high or too low:

E=PV-SP -or- E=SP-PV

Proportional-Only Control

• The control implemented on the tank level is called

Proportional-Only Control, or, simply, Proportional
Control.
• The Controlled Variable (CV), in this case, the valve
position, is directly proportional to the error.
• Slope is the Proportional Gain (Kp). CV

• There is also a Bias – because the B

valve position (and thus the flow)
is not zero when the error is zero.
E
𝐶𝑉 = 𝐾𝑝 𝐸 + 𝐵
Offset

• Notice the final levels at high demand and at low

demand:

• Proportional Control does not result in zero error

(except at one specific demand)! In general, there is an
Offset, an error value at which the loop stabilizes.
How Can We Eliminate
the Offset?

• For this mechanical example, we could move the fulcrum

of the lever up or down (adjust the Bias):

• If there is a sustained error, slide the fulcrum (Bias) up or

down to “mop up the error”.
Proportional-Plus-Integral
Control

• Determining the “sustained error” is done by integrating

(summing) E with respect to time.
• How quickly we move the bias (fulcrum) for a given
sustained error is the Integral Gain (Ki).
• Simply replace the Bias in the Proportional-Only
equation with the integrated error:
𝒕
𝐶𝑉 = 𝐾𝑝 𝐸 + 𝑲𝒊 𝑬 𝒅𝒕
𝒕=𝟎
• For convenience in tuning, we sometimes use this form:
𝑡
1
𝐶𝑉 = 𝐾𝑐 ( 𝐸 + 𝐸 𝑑𝑡
𝑡 𝑖 𝑡=0
What Puts the “D” in PID?

• What if the level starts rising (or falling) too fast?

– If the level is rising fast, start closing the valve…
– …even if the level is still below the setpoint!

• How fast is the level rising or falling?

– Take the derivative with respect to time (rate of change).
𝑡
𝒅𝑬
𝐶𝑉 = 𝐾𝑝 𝐸 + 𝐾𝑖 𝐸 𝑑𝑡 + 𝑲𝒅
𝑡=0 𝒅𝑻
– or
𝑡
1 𝑑𝐸
𝐶𝑉 = 𝐾𝑐 ( 𝐸 + 𝐸 𝑑𝑡 + 𝑡𝑑
𝑡𝑖 𝑡=0 𝑑𝑡

• Sorry, I couldn’t think of a mechanical analogue!

How It Looks in a Modern
Control System

LV
101

SP

LIC LT
101 101
Effects of Gain on PID Terms

• This animation shows the effects of increasing the

gain for each of the three terms of the PID algorithm.
– Note the offset before Integral is applied!
– Note the overshoot
with increasing P
and I gain!

• These are “ideal”

models – YMMV!
– In my experience,
Derivative makes a
less marked
improvement!

Courtesy: Automation Forum (automationforum.co, automationforum.in), see:

http://www.automationforum.co/2016/01/basics-of-pid-controller-proportional.html
Some Problems with
PID Control

• The underlying math depends on some assumptions

about “linear” variables.
– Things like pH are decidedly NON-linear!

• The CV has limits in what it can do.

– Can’t open a valve 120% open!

But the biggest problem:

• The PV must move away from the SP in order
for control action to be initiated.
– A disturbance must move the process from optimum to
less-than-optimum before the controller does anything!
What If…

• … you could measure the disturbance…

• … then account for it by changing the CV…
• … BEFORE the PV moves off setpoint?

Yes, you can!

Feedforward Control

• Suppose we measure the outflow from the tank…

LV
101

SP
FE
101
CV
LIC PV LT
101 101
FF
FT
101

• If we set the valve based on the outflow, we can adjust

inflow to the new demand BEFORE the level changes!
Feedforward Control

• The outflow is now a Measured Disturbance, or a

Disturbance Variable (DV).
– There’s usually some kind of Feedforward Model for how the DV
affects the process.
– The math goes between the DV measurement and the
Feedforward (FF) input to the controller.

• Here, the model is trivial: adjust the inflow to match the

outflow.
– The FF signal bumps the CV (output) directly, without going
through the PID (feedback) controller.
𝑡
𝑑𝐸
𝐶𝑉 = 𝐾𝑝 𝐸 + 𝐾𝑖 𝐸 𝑑𝑡 + 𝐾𝑑 + 𝑭
𝑡=0 𝑑𝑇
When to Use Feedforward
Control

• When you can measure the disturbance

– If you can’t measure it, you can’t account for it!

• When it’s important that the PV not change

– In our tank example, the tank might exist for the purpose of
letting the level vary! No big deal if it changes!

• When the disturbance dynamics have an impact

– Integral will “mop up” an unchanging FF input
to zero the Error.

• When the disturbance has a non-linear impact

– Feedforward can “remove” a non-linear disturbance from the
PID feedback control (which works best if linear).
 Damn the Distrubances!
I Can Control Everything!

• Nope! You need enough “degrees of freedom”.

– Controlling one variable means creating disturbances in another.
– Here, controlling tank level means disturbing the inflow.
– The disturbance in the level caused by changes in outflow is
transferred from the tank level to the inflow stream.

• The level example I’m using is a VERY BAD example!

A tank’s purpose is often to buffer disturbances in flow!

The level can float (within the capacity of the tank)

to absorb changes and differences in flow!

You cannot control both flows PLUS the level –

Disturbances can be transferred but not destroyed!*
* ”Rothenberg’s Law of Conservation of Disturbances”
When to Use Feedforward
Control

• More on “When the disturbance has a non-linear impact”

– PID works best for processes with a linear process response to
changes in the CV.
– Sometimes the PV measurement is non-linear – for example,
pH is a logarithm of a concentration.
– You learned the math for pH in high school chemistry! (Didn’t you?)
– Even if you can’t come up with an exact model:
– You can create a good enough model by approximation.
– You can create a good dynamic model based on recorded trends.
– Feedback control is there to mop up the unmeasured
disturbances…
– … but also to mop up leftover error in the model
Feedforward:
pH Control Example
• Neutralize the
FE
203 incoming stream

AE
• PID mops up the
AIT
203 FT unmeasured
203
203 disturbances
AY
202
pH FF Model! AV
202a
Acid
AV
202b
Base

AE
CV 202
AIC AIT
FF 202 PV 202
What’s With Those
Two Valves?

FE
• Low pH?
203 – Add base!

AE
• High pH?
AIT
203 FT – Add acid!
203
203
AY
202
• Just right?
pH FF Model! AV
202a
– Don’t waste
Acid expensive
AV chemicals!
202b
Base

AE
CV 202
AIC AIT
FF 202 PV 202
Split-Ranged Valves

• When the CV is less than 50% AV

202a
– Open one valve (let’s say Base) Acid
– Close the other valve (Acid)
– At 0% CV Base is wide open
AV
• When the CV is more than 50% 202b
Base
– Open the other valve (Acid)
– Close the other (Base)
– At 100% CV, Acid is wide open CV
AIC

• At 50% CV, both are closed 202

What should the CV do on a loop

failure (e.g., pH transmitter fail)?
Split-Range Use Cases

• Temperature Control TV
611h
– Heating when CV is more than 50% Hot
– Cooling when CV is less than 50% Water

• Others in your experience?

TV
611c
• Important considerations: Chilled
Water
– Be very careful about what to do on
system failures
– Be very careful about what to do near CV
TIC
the switchover point 611
Split-Range Considerations

• If you can, size the valves TV

so the loop gain doesn’t Hot

611h

change at switchover Water

– The switchover point doesn’t
have to be at 50% CV! TV
– The design may have more heat Chilled
611c

capacity than cooling, so the Water

switchover might be, say, 30%
– Or use adaptive gains CV
TIC
• Deadband or overlap 611

– May need to sequence devices

during switchover, for example:
switching chilled water and steam
Another Example

• Consider this Rumpus Room Temperature Control

temperature TIC TE
control strategy 611 611

– NOTE: this is a
closed-loop full
Return Air Discharge Air
analog control,
NOT an on-off
thermostat! Hot Water
TV
CV 611h

TV
611c
Chilled Water

• What happens if the outside air Fresh Air Fan

temperature suddenly drops?
What Are the Effects of
This Disturbance?

• The incoming Fresh

Air temperature falls. TIC PV TE
611 611
CV

• This results in cooler

Discharge Air into Return Air Discharge Air

the room.
Hot Water

• The room cools as a

TV
611h

result of thermal loss TV

611c
Chilled Water
through the walls.
• The controller calls Fresh Air Fan

for more hot water to

compensate.
So? We Can Measure
This Disturbance!

• We can
compensate TY FF TIC PV TE

with a 612 611

CV
611

Feedforward
Model… Return Air Discharge Air

• But what about Hot Water

OTHER TV
611h

disturbances? TV
611c
Chilled Water

TE
612
Fresh Air Fan

Ambient Temp
What About Other
Disturbances?

• The hot water (or

chilled water or TIC PV TE

steam) supply 611

CV
611

temperature or
pressure could Return Air Discharge Air

change.
Hot Water
• The damper TV
611h

positions could TV
611c

change the mix of Chilled Water

fresh air.
• The fresh air filter
Fresh Air Fan

(not shown) clogs

over time.
A Simple Way to Deal with
MOST of These Disturbances

• Consider two PID

loops, one stacked Room Temp
TE PV TIC CV SP TIC PV

on top of the other: 611 611 612

CV

– The Discharge Air

Temp loop Return Air
responds quickly to
disturbances like TE
Hot Water 612
supply variances, TV Discharge
611h
incoming fresh air TV
Air Temp

temp changes. 611c

Chilled Water
– The Room Temp
loop controls room
temperature by Fresh Air Fan

requesting a
discharge air
temperature.
Cascade Control

• When the output of

one loop becomes Room Temp
TE PV TIC CV SP TIC PV

the setpoint of 611 611 612

CV

another loop, the

loops are said to Return Air

be “cascaded”.
– The loop being
TE
Hot Water 612
TV Discharge
driven (SP set) is 611h
Air Temp
the “secondary” or TV
611c
“inner” loop. Chilled Water

– The loop driving

(CV sent to inner
Fresh Air Fan
loop) is the
“primary” or “outer”
loop.
Cascade Control
Considerations

• Generally, the inner

loop must be “much”
faster than the outer
Room Temp
loop. TE
611
PV TIC
611
CV SP TIC
612
PV

– About 3x faster or more CV

• Here, the discharge air Return Air

temperature loop Hot Water

TE
612

responds “much” faster TV

611h
Discharge
Air Temp

than the overall room TV

611c
Chilled Water
temperature
– Consider the thermal
Fresh Air Fan
mass of the room, plus
its insulation
Other Examples of
Cascade Control

TIC TE
Reactor Temp 510 510

TIC TE
Shell Temp 511 511

Steam
TV
511

• Reactor temperature (big, slow), over reactor shell

temperature (much less volume, faster)
When to Use
Cascade Control
• There are disturbances
that cause measurable
changes that you can Reactor Temp
TIC
510
TE
510

handle with this TIC TE

Shell Temp
separate loop. 511 511

– Suppose the steam valve

has lots of “sticktion”… Steam

– The inner loop keeps this

TV
511

or other non-linearity from

affecting the outer loop.

• That inner loop responds “faster” than the overall control.

• Useful: You can apply constraints on the inner loop…
– … such as Setpoint clamping, to prevent undesirable excursions.
Test Your Knowledge!

• Is this an example of Cascade Control?

 FT
101
FE
101

Effluent

HINT #2 FE FT
102 102 FV
102

Reagent

PV
SP FFIC CV
102

HINT #1
Where’s the Loop?

• NO! There’s not an outer control loop!

• Effluent flow is  FT
101
FE
101

measured, but Effluent

not controlled
FE FT
102 102 FV
• The two flows 102

Reagent
have similar
time response
PV

• Reagent flow is SP FFIC

102
CV

controlled at a
ratio to the
uncontrolled (“wild”)
effluent flow
Ratio Control

• One variable (typically

 FT
101
FE
101
flow) is controlled at a
given ratio to an
Effluent
uncontrolled (“wild”)
FE FT variable.
102 102 FV
102

Reagent
• Both variables are
measured, but there
is only one Final
SP FFIC
PV
CV
Control Element
102
 Test Your Knowledge
(#2)
SP • Is this an example
Blend PV FIC
Master 101 CV of Ratio Control?
FT FE
101 101
• No!
FV
101

Ingredient 1 – Neither flow is “wild”,

both are controlled
FE FT
102 102 FV – The Blend Master
102
drives setpoints to
Ingredient 2
both flow loops
PV

FIC
SP 102 CV • But Digital Blending
is a great way to
precision blend
ingredients!
Ratio Control Considerations

• The controlled flow

 FT FE

naturally lags 101 101

– Measurement lag Effluent

– Loop processing time FE FT

102 102 FV
102

• Ratio control Reagent

cannot make up for

“historical” inaccuracy PV
FFIC
– It is an “instantaneous”
SP CV
102

ratio
– Digital Blending uses
the flow rates AND the
accumulated flow totals
to “integrate out” any
accumulated variance
from the target ratio
And Now Something
Completely Different
• Consider this oil Main Line Unit

pipeline pump
station
• We measure: M
SC
M
60
– suction pressure
– discharge pressure PSL PSH

– motor current 43 71

• Let’s control
PT PT

41 73
discharge pressure M
“Previous” “Next”
using the variable- Station Station

speed drive on the

pump!
But There Are Constraints…

• We want to control Main Line Unit

the discharge
pressure, but…
• If the suction M
SC
M
60
pressure goes low,
the pump cavitates PSL PSH

– Big pump = big $$$! 43 71

PT PT
• If the motor current 41 73
goes too high, we M
“Previous” “Next”
trip the drive Station Station

– Pressure upset gets

sent down the line
Lucky For Us!

• Both constraints act Main Line Unit

“in the same direction”

– If the suction pressure
goes low, slow down
SC
the pump until it M M
60
recovers.
– If the motor current PSL PSH
goes high, slowing 43 71
down the pump
PT PT
reduces the
41 73
power, and so M
reduces the “Previous” “Next”
Station Station
motor current.
Override Select Control

• The Primary loop is the

Main Line Unit
station discharge
pressure.
• Suction pressure and
motor current are Speed Motor
Ref. SC Amps
Override loops. M
60
M

• The low-select picks

the lowest CV to
PIC
41
CV
 CV PIC
73

send to the drive PT CV IIC PT

speed reference. 41 60 73
M
“Previous” “Next”
Station Station
 How Override Select
Control Works (DEMO!)

• Suction pressure and

Main Line Unit motor current loops’
setpoints are set to the
constraint threshold
(where to start acting)
Speed Motor
M Ref. SC Amps M • When a constraint is
60
reached, that loop’s
error changes sign, its
PIC
41
CV
 CV PIC
73
output drops, and
PT CV IIC PT it is selected.
41 60 73
LOW! M • The selected CV is fed
“Previous” “Next”
Station Station back to all three loops
Topics for Further Study

• Tuning
• Tuning! (what do Kp, Ki, and Kd MEAN anyway?)
• Tuning!!! (seriously, there are books just on tuning!)
• Control for interacting process variables
• Control for deadtime-dominant processes
– Smith Predictor, among many

• Advanced process modeling and model-based control

– Tuning of model based controllers MAKES SENSE! Lag time,
dead time, process gain! Stuff you can read off a trend chart!

and so much more!

Questions?

• Ask now!
• Contact the presenter!
– [email protected]

• Ask other experts at this meeting! (There are several!)

• Get a great book for more details:
– Wade, Harold L., Basic and Advanced Regulatory Control:
System Design and Application (2nd Edition), ISA, 2004.
– ISBN: 978-1-55617-873-3
– Many more topics: The ones covered here, plus extensive
information on tuning, decoupling multiple variables, deadtime
compensation, basics of model-based control, plus applying!