Open loop PID tuning
for 1st order systems
Benjamin Bradu
CERN, EN-ICE
(Industrial Control Engineering)
June 2015
Regulation loops
Allow to regulate process variables given by sensors (current,
temperature, pressure, ).
Make use of actuators (valves, pumps, heaters) to fit different
objectives in real-time:
Stay between acceptable ranges.
Keep a constant set-point rejecting disturbances.
Follow a set-point (stairs or ramps).
2 main techniques :
1.
2.
Feed-Forward Control
Feedback control
Feedback Control
Regulation loop using a feedback (sensor) to compute the actuator position
y: Sensor to be controlled (process value)
u: Actuator to be manipulated (manipulated value)
r: Set-point to be respected (reference)
PID: most common algorithm
Compute the error
Proportional action
Integral action
Derivative action
Then, you have to Tune your PID
and find a compromise overshoot Vs time response.
PID tuning hypothesis
Process is a linear 1st order system (stable or unstable)
Be sure to make the test around operation point to limit non-linear effects.
Most of processes can be approximated by a 1st order system. If not, this method
cannot be applied.
Reminder about a linear system:
A valve step between 10% 20% has the same effect than
a step between 80%90%
-
A valve step of 10% at 20 C or at 30 C has the same effect
Conclusion: Nature is never linear (unfortunately for control)
PID tuning hypothesis
Tuning is done for a Single Input/Single Output (SISO) system
Try to minimize the other regulation impacts when you do tuning.
For cascade tuning:
For Split-Range tuning, use only one actuator during the tuning.
Actuator speed is infinite
First: Tune the internal loop with the external loop disabled
Second: Tune the external loop with the internal loop enabled and tuned
Be sure that your actuator can follow the controller output.
There is no disturbance on the measurement
Try to reduce external disturbances during the tuning.
If measurement is noisy, it is necessary to apply a first order filter on the
measured value.
PID tuning principle in open loop
1.
Identify your process dynamics
1.
2.
3.
4.
5.
6.
Open regulation loop (Controller in Positioning mode)
Put a constant PID output close to your operation point
Wait for stable situation (can be long)
Apply a step on the PID output (up or down)
Wait for stable situation (can be long)
Identify your process dynamics (curve analysis)
2.
Compute PID parameters (1 minute)
3.
Close the loop (Controller in regulation mode) to test
Open loop identification for stable system
1.
2.
Apply a step du on the PID output
When stable value, measure:
Gain: K=dy/du
Delay: tau
Time constant: T
PI tuning for stable systems
There are about thousands of tuning rules for a 1st order system. Here are 4 examples:
*This table is valid for a mixed PI structure (compatible with UNICOS)
Open loop identification for unstable system
1.
2.
Apply a step du on the PID output
When stable value, measure:
Gain: K=dy/dt/du
Delay: tau
PI tuning for unstable systems
There are about thousands of tuning rules for a 1st order system. Here are 4 examples:
*This table is valid for a mixed PI structure (compatible with UNICOS)
10
PID validity domains
Domains
Regulation type
T/tau > 20
On/Off Regulation
10 < T/tau < 20
P regulation
5 < T/tau < 10
PI regulation
2 < T/tau < 5
PID regulation
T/tau < 2
PID cannot achieve the regulation
11
UNICOS specificities
For all regulation loops:
There is a ramp on the set-point to limit the error when the regulation is
starting. This ramp has to be compatible with your tuning.
There is a ramp on the actuator to limit actuator jumps if PID output is too
fast. This ramp has to be compatible with your tuning.
PID regulation loops can be scaled or not:
No Scaling (CV): Engineering values are used
Input Scaling (CRYO): the set-point and the process value are scaled
between [0-100 %] according to the sensor range. So, the proportional gain
K should be scaled as well.
12
PID auto-tuning in UNICOS
Automatic tool to excite the process and find PID
parameters online
Define limits to be respected on process value and on actuator
Select PI or PID
Auto-Tune
3 methods available
Relay method: Open-loop auto-tuning.
IFT method: Close-loop auto-tuning (need an existing stable tuning).
SIMC method: Open-loop manual tuning (identification + tuning)
13
IFT and Relay methods
RELAY: Open loop auto tuning
IFT: Close loop auto tuning
14
SIMC method
Identification based on step response
Manual tuning based
on fast/robust IMC controller
15
PID auto-tuning: demo
Initial guess
Open loop method (relay)
PI = [300, 10]
PI = [169.76, 28.125]
Close loop method (IFT)
PI = [168.827, 22.43]
16
Conclusion
PID tuning is easy to be setup if :
The process can be excited enough.
The process is not too non-linear and not too disturbed.
Actuator dynamics and ranges are large enough.
It is impossible to tune a regulation loop without exciting the process.
You need to move your actuator significantly to evaluate its effects in term of
amplitude, time and frequency.
It is useless to try to change the PID parameters in closed loop when the loop
is regulating at a constant set-point.
It is impossible to compare controller gains between them if they are not
normalized (no scaling in the regulation loop).
The try and error method can work only if you are already experienced on
identical regulation loops having the same kind of gains, time constants
and delays.
17
References
18
