powerpoint

Report
ChE / MET 433
4 Apr 12
1
Feedback Controller Tuning:
(General Approaches)
1) Simple criteria; i.e QAD via ZN I, tr, etc
• easy, simple, do on existing process
• multiple solutions
2) Time integral performance criteria
• ISE
integral square error
• IAE
integral absolute value error
• ITAE
integral time weighted average error
3) Semi-empirical rules
• FOPDT (ZN II)
• Cohen-Coon
4) ATV, or Autotuning
5) Trial and error
6) Rules of thumb
2
Trial and Error (field tuning)*
• Select the tuning criterion for the control loop.
• Apply filtering to the sensor reading
• Determine if the control system is fast or slow responding.
– For fast responding, field tune (trail-and-error)
– For slow responding, apply ATV-based tuning
• Turn off integral and derivative action.
• Make initial estimate of Kc based on process knowledge.
• Using setpoint changes, increase Kc until tuning criterion is
met
ys
c
a
b
Time
* J.B. Riggs, & M.N. Karim
Chemical and Bio-Process
Control, 3rd ed. (2006)
3
Trial and Error (field tuning)*
Decrease Kc by 10%.
Make initial estimate of tI (i.e., tI=5tp).
Reduce tI until offset is eliminated
Check that proper amount of Kc and tI are used.
c
b
ys
•
•
•
•
a
Time
* J.B. Riggs, & M.N. Karim
Chemical and Bio-Process
Control, 3rd ed. (2006)
4
Kc and
Kc
tI
levels good?
tI
5
Feedback Controller Tuning:
(General Approaches)
1) Simple criteria; i.e QAD via ZN I, tr, etc
• easy, simple, do on existing process
• multiple solutions
2) Time integral performance criteria
• ISE
integral square error
• IAE
integral absolute value error
• ITAE
integral time weighted average error
3) Semi-empirical rules
• FOPDT (ZN II)
• Cohen-Coon
4) ATV, or Autotuning
5) Trial and error
6) Rules of thumb
6
Rules of Thumb
*
sec
** •
•
•
•
sec
Flow Loops: typically PI controllers; PB ~ 150; t I  0.1 min
Level Loops: PI for tight control; P for multiple tanks in series;
Pressure Loops: can be fast or slow (like P control by controlling condenser)
Temperature Loops: typically moderately slow; typically might use PID
controller; PB fairly low (depends on gains); integral time on order of
process time constant, with faster process t I can be smaller
derivative time ~ ¼ the process time constant.
* D.A.Coggan, ed., Fundamentals of Industrial
Control, 2nd ed., ISA, NC (2005)
** W.L.Luyben, Process Modeling, Simulation and Control for
Chemical Engineers, 2nd ed., McGraw-Hill (1990)
7
Higher Order Process
8
Feedback Controller Tuning:
(General Approaches)
1) Simple criteria; i.e QAD via ZN I, tr, etc
• easy, simple, do on existing process
• multiple solutions
2) Time integral performance criteria
• ISE
integral square error
• IAE
integral absolute value error
• ITAE
integral time weighted average error
3) Semi-empirical rules
• FOPDT (ZN II)
• Cohen-Coon
4) ATV, or Autotuning
5) Trial and error
6) Rules of thumb
9
Feedback Control
•Design
Disturbances:
•Load
•Setpoint
Questions:
•Type of controller to use?
•How select best adjustable parameters?
•Performance criteria?
Guidelines:
•Define performance.
•Obtain “best” parameters, for K C ,t I ,t D
•Select controller with “best” performance.
10
Feed Back Control
•PID controllers
Proportional:
•Accelerates response
•Offset
Controller with “best” performance.
•P – only if can
•PI – eliminate offset
•PID – speed up response of sluggish systems
(T, comp, control; multi-capacity systems)
Integral:
•Eliminates offset
•Sluggish responses
•If increase Kc, more oscillations -> unstable?
Derivative:
•“Anticipates” future error
•Stabilizing effect
•Noise problem
11
Examples:
12
13
14
15
16
E s 
Controllers:
P-Only:
mt   m  K c r (t )  c(t )  M s   K
c
E (s)
P-I Controller:
P-I-D Controller:
tD 
mt   m  K c e(t ) 
Kc
tI
Gc
M s 

 e(t ) dt
M s   
1 
 
  K c 1 
E (s)   t I s  

1
d e(t ) 
mt   m  K c e(t )   e(t ) dt  t D

t
dt
I


Derivative (rate) time [=] time
Chapter 5 ~ p 183

M s   
1
  K c 1 
 t D s  
E (s)   t I s

17
Derivative Action:
P-I-D Controller:
R(t )

1
d e(t ) 
mt   m  K c e(t )   e(t ) dt  t D

t
dt
I


A
C (t )
e(t )
C (t )
R(t )
A
e(t )
t
slope 
d e(t )
dt
C (t )
slope 
d C (t )
dt
t

1
d C (t ) 
mt   m  K c e(t )   e(t ) dt  t D

t
dt
I


18
Derivative Action:
Another potential problem: noise
R(t )
A
filter
C (t )
e(t )
C (t )
e(t )
t
slope  
t Ds
t D s  1
 small
0.05    0.2
19
PID
Derivative action:
Advantages:
• Reduces overshoot
• Reduces oscillations
• Recommended for slow/sluggish processes (speed up control)
Disadvantages:
• Susceptible to noise
• Filtering (or averaging PV) introduces delay
• 3rd tuning parameter
20
PID Control
PID Tuning
• Tune for PI
• Derivative:
•Add in tD
•Minimum response time
•tD initial = Tu/8
•Adjust Kc and tI by same factor (%)
•Check response has correct level of integral action
PS: Try PID for HE process on Loop Pro Developer
21
Rules of Thumb
*
** •
•
•
•
Flow Loops: typically PI controllers; PB ~ 150; t I  0.1 min
Level Loops: PI for tight control; P for multiple tanks in series;
Pressure Loops: can be fast or slow (like P control by controlling condenser)
Temperature Loops: typically moderately slow; typically might use PID
controller; PB fairly low (depends on gains); integral time on order of
process time constant, with faster process t I can be smaller
derivative time ~ ¼ the process time constant.
* D.A.Coggan, ed., Fundamentals of Industrial
Control, 2nd ed., ISA, NC (2005)
** W.L.Luyben, Process Modeling, Simulation and Control for
Chemical Engineers, 2nd ed., McGraw-Hill (1990)
22
ChE / MET 433
23

similar documents