CHAPTER 15: FEEDFORWARD CONTROL
When I complete this chapter, I want to be able to do the following.
Identify situations for which feedforward is a good control enhancement Design feedforward control using the five design rules Apply the feedforward principle to other challenges in life
CHAPTER 15: FEEDFORWARD CONTROL
Outline of the lesson.
A process challenge - improve performance Feedforward design rules Good features and application guidelines Several process examples Analogy to management principle
CHAPTER 15: FEEDFORWARD CONTROL
F 1 L 1
feed
T 1
product
Discuss this stirred tank heat exchanger.
TC 2
PID controller
T 3
F 2
heating stream
CHAPTER 15: FEEDFORWARD CONTROL
Class exercise: What do we do?
F 1 L 1
76
TC
IAE = 237.6971 ISE = 758.425
74 temperature
feed
T 1 Disturbance = feed temperature Control performance not acceptable!
72
minimum
20 40 60 80 100 120 140 160 180 200
70 0
TC 2
Lets use cascade
T 3
F 2
heating stream
CHAPTER 15: FEEDFORWARD CONTROL
CASCADE DESIGN CRITERIA FOR T1
Cascade is desired when 1. 2. Single-loop performance unacceptable A measured variable is available
A secondary variable must 3. 4. 5. Indicate the occurrence of an important disturbance Have a causal relationship from valve to secondary Have a faster response than the primary
CHAPTER 15: FEEDFORWARD CONTROL
CASCADE DESIGN CRITERIA FOR T1
Cascade is desired when
OK 1.
2.
Single-loop performance unacceptable A measured variable is available
A secondary variable must 3. 4. 5. Indicate the occurrence of an important disturbance Have a causal relationship from valve to secondary Have a faster response than the primary
CHAPTER 15: FEEDFORWARD CONTROL
CASCADE DESIGN CRITERIA FOR T1
Cascade is desired when
OK 1. OK 2.
Single-loop performance unacceptable A measured variable is available
A secondary variable must 3. 4. 5. Indicate the occurrence of an important disturbance Have a causal relationship from valve to secondary Have a faster response than the primary
CHAPTER 15: FEEDFORWARD CONTROL
CASCADE DESIGN CRITERIA FOR T1
Cascade is desired when
OK 1. OK 2.
Single-loop performance unacceptable A measured variable is available
A secondary variable must
OK 3.
4. 5.
Indicate the occurrence of an important disturbance Have a causal relationship from valve to secondary Have a faster response than the primary
CHAPTER 15: FEEDFORWARD CONTROL
CASCADE DESIGN CRITERIA FOR T1
Cascade is desired when
OK 1. OK 2.
Single-loop performance unacceptable A measured variable is available
A secondary variable must
OK 3.
4. 5.
Indicate the occurrence of an important disturbance Have a causal relationship from valve to secondary Have a faster response than the primary
NO!
Cascade not possible. We need another enhancement!
CHAPTER 15: FEEDFORWARD CONTROL
Lets think about the process behavior. Causal relationship from T1 disturbance to T2 (without control) How can we manipulate valve to compensate? v (valve) T0 (Feed temperature) Q
T 3 F 1
feed
T 1
L 1
product
TC 2
F 2
heating stream
TC
CHAPTER 15: FEEDFORWARD CONTROL
76
We want to adjust the valve to cancel the effect of the disturbance.
74
72 70 68 66 0 20 40 60 80 100 120 140 160 180 200
Time
T0
0 20 40 60 80
Dm(t) = T0
100 Time 120 140 160 180 200
CHAPTER 15: FEEDFORWARD CONTROL
76
We want to adjust the valve to cancel the effect of the disturbance.
74
72 70 68 66 0 20 40 60 80
CVA(t) = disturbance effect
100 120 140 160 180 200
Time
T0
0 20 40 60 80
Dm(t) = T0
100 Time 120 140 160 180 200
CHAPTER 15: FEEDFORWARD CONTROL
CVB(t) = compensation effect
76
We want to adjust the valve to cancel the effect of the disturbance.
74
72 70 68 66 0 20 40 60 80
CVA(t) = disturbance effect
100 120 140 160 180 200
Time
T0
0 20 40 60 80
Dm(t) = T0
100 Time 120 140 160 180 200
CHAPTER 15: FEEDFORWARD CONTROL
CVB(t) = compensation effect
76
CVA + CVB = no deviation
We want to adjust the valve to cancel the effect of the disturbance.
74
72 70 68 66 0 20 40 60 80
CVA(t) = disturbance effect
100 120 140 160 180 200
Time
T0
0 20 40 60 80
Dm(t) = T0
100 Time 120 140 160 180 200
CHAPTER 15: FEEDFORWARD CONTROL
CVB(t) = compensation effect
76
CVA + CVB = no deviation
We want to adjust the valve to cancel the effect of the disturbance.
74
72 70 68 66 0 60 58 56 54 52 50 0 20 40 60 80 100 Time 120 140 160 180 200 20 40 60 80
CVA(t) = disturbance effect
100 120 140 160 180 200
MV(t) = v
T0
0 20 40 60 80
Dm(t) = T0
100 Time 120 140 160 180 200
CHAPTER 15: FEEDFORWARD CONTROL
We use block diagram algebra to determine the form of the calculation [Gff(s)] to achieve the desired performance.
Measured disturbance, T0
Dm(s)
CV A(s)
Gd(s)
+
Controlled variable, T
Feedforward controller
Gff(s)
MV (s)
CV (s)
How do we measure CVA?
Gp(s)
Manipulated variable
CV B(s)
CHAPTER 15: FEEDFORWARD CONTROL
CV ( s ) = CV A ( s ) + CVB ( s ) = 0
= Gd ( s ) + G ff ( s )G p ( s ) Dm ( s ) = 0
Not a PID algorithm! Why?
??
Gd ( s ) MV ( s ) = G ff ( s ) = Dm ( s ) G p (s)
This is general!
CHAPTER 15: FEEDFORWARD CONTROL
Gd ( s ) MV ( s ) = G ff ( s ) = Dm ( s ) G p (s)
Special case of Gp(s) and Gd(s) being first order with dead time
Please verify.
Tld s + 1 ff s MV (s ) G ff (s ) = e = K ff Tlg s + 1 D m (s )
Gain Lead-lag Dead time
CHAPTER 15: FEEDFORWARD CONTROL
G ff ( s ) = K ff
Lead-lag FF controller gain controller dead time Lead time Lag time
Tld s + 1 ff s e Tlg s + 1
= (Tlds+1)/Tlgs+1) = Kff = - Kd/Kp = ff = d - p 0 = Tld = p = Tlg = d
How do we get values for these parameters?
CHAPTER 15: FEEDFORWARD CONTROL
G ff ( s ) = K ff Tld s + 1 ff s e Tlg s + 1
Tld / t + 1 ( Dm ) N ( MV ff ) N 1 + K ff ( MV ff ) N = Tlg / t + 1 Tlg / t + 1 Tld / t ( Dm ) N 1 K ff Tlg / t + 1 Tlg / t ( MV ff ) N = a( MV ff ) N 1 + b( Dm ) N + c( Dm ) N 1
Digital implementation is straightforward. Its derived in textbook.
CHAPTER 15: FEEDFORWARD CONTROL
Typical dynamic responses from the lead-lag element in the feedforward controller. It synchronizes the compensation and disturbance effects.
Results for several cases of Tlead/Tlag :
a. 0.0 b. 0.5 c. 1.0 d. 1.5 e. 2.0
CHAPTER 15: FEEDFORWARD CONTROL
F 1
feed
L 1
How do we combine feedback with feedforward?
FF highlighted in red
T 1 FF TC 2
TY 1
MVfb
TY 2 T 3 F 2
MVff
heating stream
CHAPTER 15: FEEDFORWARD CONTROL
Control Performance Comparison for CST Heater Single-Loop
IAE = 237.6971 ISE = 758.425 76 76 75 temperature temperature 74 74 73 72 71 70 0 20 40 60 80 100 120 140 160 180 200 70 0 20 40 60 80 100 120 140 160 180 200
Feedforward with feedback
IAE = 27.772 ISE = 8.0059
72
Much better performance! WHY?
CHAPTER 15: FEEDFORWARD CONTROL
IAE = 27.772 ISE = 8.0059 75.4 75.2 temperature 75 74.8 74.6 74.4
TC
The MV changed before T deviated from its set point!
0 20 40 60 80 100 120 140 160 180 200 SAM = 11.4394 SSM = 774.0613
60 heating valve (% open) 58 56 54 52 50
Valve adjustment not too aggressive
0 20 40 60 80 100 Time 120 140 160 180 200
T1
Disturbance occurred at this time
Why wait after disturbance?
CHAPTER 15: FEEDFORWARD CONTROL
What have we gained and lost using feedforward and feedback? For each case, is FF with FB better, same, worse than single-loop feedback (TC2 v)??
F 1 feed T 1 TC 2 + TY 2 T 3 F 2 L 1
TY 1
FF
heating stream
A disturbance in feed inlet temperature A disturbance in heating medium inlet pressure A disturbance in feed flow rate A change to the TC set point
CHAPTER 15: FEEDFORWARD CONTROL
What have we gained and lost using feedforward and feedback? For each case, is FF and FB better, same, worse than single-loop feedback (TC2 v)??
F 1 feed T 1 TC 2 + TY 2 T 3 F 2 L 1
TY 1
FF
heating stream
A disturbance in feed inlet temperature A disturbance in heating medium inlet pressure A disturbance in feed flow rate A change to the TC set point
FF/FB better Both the same Both the same Both the same
CHAPTER 15: FEEDFORWARD CONTROL
FEEDFORWARD DESIGN CRITERIA Feedforward is desired when 1. 2. 3. 4. 5. Single-loop performance unacceptable A measured variable is available Indicate the occurrence of an important disturbance NOT have a causal relationship from valve to measured disturbance sensor Not have a much faster affect on the CV than the MV (when combined with feedback)
A measured disturbance variable must
CHAPTER 15: FEEDFORWARD CONTROL Feedforward and Feedback are complementary
Advantages Feedforward Compensates for disturbance before CV is affected Does not affect the stability of the control sysytem (if Gff(s) stable) Cannot eliminate steadystate offset Requires a sensor and model for each disturbance Feedback Provides zero steadystate offset Effective for all disturbances
Disadvantages
Does not take control action until the CV deviates from its set point Affects the stability of the control system
CHAPTER 15: FEEDFORWARD CONTROL
CLASS EXERCISE: SOME QUESTIONS ABOUT FEEDFORWARD CONTROL Why do we retain the feedback controller? When would feedforward give zero steady-state offset? Why does the feedforward controller sometimes delay its compensation? Dont we always want fast control? What is the additional cost for feedforward control? How can we design a strategy that has two controllers both adjusting the same valve? What procedure is used for tuning feedforward control?
CHAPTER 15: FEEDFORWARD CONTROL
heating stream
Discuss this packed bed reactor.
F 1
F 2
T 2 T 3
feed
T 1 A 2
packed bed reactor
Notes: 1. A1 measures reactant concentration 2. Circle is shell & tube heat exchanger 3. Feed valve is adjusted by upstream process 4. Increasing temperature increases reaction rate A 1
product
CHAPTER 15: FEEDFORWARD CONTROL
Class exercise: Design feedforward control to improve the performance. Performance not acceptable for feed composition disturbance
F 1
heating stream
F 2
What about cascade?
T 2 T 3
feed
T 1 A 2
AC
0.2 0.15 CV1 0.1 0.05 0 -0.05 0
IAE = 22.9349 ISE = 3.0248
packed bed reactor
maximum
Disturbance in feed composition
AC 1
100
200
300
400
500
product
CHAPTER 15: FEEDFORWARD CONTROL
Class exercise: Design feedforward control to improve the performance.
Feedforward design criteria 1. Single-loop not acceptable 2. Disturbance variable is measured 3. Indicates a key disturbance 4. No Causal relationship, valve Dm 5. Disturbance dynamics not much faster than compensation A2 F1 F2 T1 T2 T3
Lets use the feedforward design rules!
Remember: The disturbance is the feed composition.
CHAPTER 15: FEEDFORWARD CONTROL
Class exercise: Design feedforward control to improve the performance.
Feedforward design criteria 1. Single-loop not acceptable 2. Disturbance variable is measured 3. Indicates a key disturbance 4. No Causal relationship, valve Dm 5. Disturbance dynamics not much faster than compensation
A2 Y Y Y Y Y
F1 Y Y N Y
F2 Y Y N N
T1 Y Y N Y
T2 Y Y N Y
T3 Y Y N N
N/A N/A N/A N/A N/A
A2 satisfies all of the rules and can be used as a feedforward variable.
CHAPTER 15: FEEDFORWARD CONTROL
heating stream
F 2
MVff MV
AC 1
F 1
T 2 T 3
feed
T 1 FF
A 2
Dm
MVfb packed bed reactor
AC 1
AY 2
CV1
SP1 from person product
CHAPTER 15: FEEDFORWARD CONTROL
Control Performance Comparison for Packed Bed Reactor Single-Loop
IAE = 22.9349 ISE = 3.0248 0.2
Feedforward and feedback
IAE = 2.1794 ISE = 0.017852
0.15
AC
AC
0.1 CV1
0.05
-0.05 0 100 200 300 400 500
0.01 0 -0.01 -0.02 -0.03 0
100
200
300
400
500
Much better performance! WHY?
Little model error, most experimental feedforward not this good!
CHAPTER 15: FEEDFORWARD CONTROL
What have we gained and lost using feedforward and feedback? How does the system respond to the following? A disturbance in T2 A disturbance inT1 A disturbance to feed composition, A2 A change to the AC-1 set point
Both the same
feed AY 2 heating stream F 1 F 2 MV T 2 T 3 MVff AC + 1
T 1 FF
A 2 Dm packed bed reactor
MVfb
AC 1
Both the same
CV1
SP1 from person product
A disturbance in heating medium inlet pressure
Both the same
Both the same
FF/FB better
CHAPTER 15: FEEDFORWARD CONTROL
We can combine cascade and feedforward to gain the advantages of both.
heating stream
F 2 MV2 MVff
F 1
T 2 TC 3 secondary MV1 packed bed reactor primary CV1 AC 1 SP1 from person TY 3
feed
T 1 AY 2
A 2
CV2
CHAPTER 15: FEEDFORWARD CONTROL
Ratio control is a simple and frequently used feedforward application. In ratio control, the dynamics are negligible. Uncontrolled (wild) flow
F 2 FY 1 FC 1
Desired F1/F2 = R
SPF1 = F2*R
Blended flow
Goal is to keep F1/F2 constant.
Manipulated flow
CHAPTER 15: FEEDFORWARD CONTROL
CLASS EXERCISE: Use analyzer in automatic control while retaining the good aspects of ratio control. Uncontrolled (wild) flow
F 2
Desired F1/F2 = R
SPF1 = F2*R
x FY
1 FC 1
A 1
Blended flow
Goal is to keep A1 constant.
Manipulated flow
CHAPTER 15: FEEDFORWARD CONTROL
CLASS EXERCISE: Use analyzer in automatic control while retaining the good aspects of ratio control. Uncontrolled (wild) flow
F 2
Feedback PID
AC 1
R
SPF1 = F2*R
x FY
1 FC 1
Blended flow
Goal is to keep A1 constant.
Manipulated flow
CHAPTER 15: FEEDFORWARD CONTROL
In many organizations, we take actions on inputs to prevent large disturbances to outputs. Sometimes, these are called pre-actions. What would you do if? Number of births per year increases by 10% in your country A drought occurs in in the most fertile area of your country
After you have measured the change, you have some time to react before it hits you
New legislation will impose stricter emissions regulations in three years Do we need feedback? What is your algorithm? What would you do if the measurement were noisy?
CHAPTER 15: FEEDFORWARD WORKSHOP 1 Evaluate feedforward control for a disturbance in the heating medium inlet temperature. You may add a sensor but make no other changes to the equipment.
F 1 feed T 1 TC 2 L 1 product
T 3
F 2 heating stream
CHAPTER 15: FEEDFORWARD WORKSHOP 2 Prepare a flowchart for the calculations performed by the packed bed feedforward controller. Show every calculation and use process variable symbols (e.g., A1), not generic symbols (CV1). Report the equations for digital control.
heating stream F 2 MV F 1 T 2 T 3 MVff AC + 1
feed
T 1 FF
A 2 Dm packed bed reactor
MVfb
AY 2
AC 1 CV1
SP1 from person
CHAPTER 15: FEEDFORWARD WORKSHOP 3 Answer each of the following questions true or false 1. 2. 3. The feedback controller tuning does not change when combined with feedforward compensation. The feedforward controller has no tuning parameter. The feedforward controller should react immediately when the measured disturbance is measured. Feedforward could be applied for a set point change.
4.
CHAPTER 15: FEEDFORWARD WORKSHOP 4 Identify a process that would benefit from ratio control. You may select from examples in your summer/co-op jobs, engineering laboratories, and course projects. Draw a sketch of the process with ratio control. Explain the advantages and any disadvantages of the design.
CHAPTER 15: FEEDFORWARD
When I complete this chapter, I want to be able to do the following. Identify situations for which feedforward is a good control enhancement Design feedforward control using the five design rules Apply the feedforward principle to other challenges in life
Lots of improvement, but we need some more study! Read the textbook Review the notes, especially learning goals and workshop Try out the self-study suggestions Naturally, well have an assignment!
CHAPTER 15: LEARNING RESOURCES
SITE PC-EDUCATION WEB - Instrumentation Notes - Interactive Learning Module (Chapter 15) - Tutorials (Chapter 15) The Textbook, naturally, for many more examples
CHAPTER 15: SUGGESTIONS FOR SELF-STUDY 1. Suggest some methods for fine-tuning a feedforward controller. 2. Program a feedforward controller for one of the processes modelled in Chapters 3-5. 3. Explain why the feedforward compensation should not be much slower than the disturbance. Why doesnt this guideline apply when no feedback is implemented? 4. Discuss whether you would recommend more than one feedforward controller on the same process. 5. Write a memorandum explaining feedforward compensation for a company with non-technical employees
CHAPTER 15: SUGGESTIONS FOR SELF-STUDY 6. A friend asks whether the general sketch for feedback, textbook Figure 1.4, applies to feedforward. Answer completely, including any changes to the sketch. 7. Discuss why the feedforward controller dead time must be positive.
