SelfTuning PID Control

1
Self-Tuning PID Control

• 1999-2000
• Spring Presentation

2
Classic PID Algorithm
m controller output e error Kc controller
gain
Ti integral time Td derivative time Ts
sampling period
3
Our Approach

• Implement classic PID algorithm with integral
  windup provisions in Visual Basic on a PC,
  interfacing between PC and B3000 via Ethernet
• Implement self-tuning method
• Use an Athena controller as a benchmark to verify
  controller performance

4
Athena Control
5
Opto22 vs. Athena Control No Integral Windup
6
Opto22 vs. Athena Control With Anti-Windup
7
Self-Tuning Approach

• Process response curve method
• Gather data from open-loop plant response to unit
  step input
• Examine data set to find maximum slope
• Determine ? and ?
• Use tuning relations to generate PID constants
• Compare with Athena

8
Open Loop Response Method

• Approximates higher order plants as first order
  with pure delay
• Plot open loop plant response to step input
• Graphically derive ? and R
• PID settings Kc 1.2/R?
• Ti 0.5/?
• Td 0.5?

9
Athena Autotune Process
10

Self-Tuning Plant Response Approximation
11
Self-Tuning Results
12

Gain Scheduling

• Thermal Plants require different PID constants at
  different setpoints
• PID constants can be adjusted as a function of
  setpoint and system parameters, q and t
• Investigated variation of Athena PID constants as
  a function of setpoint and system parameters for
  multiple plants

13
Gain SchedulingPID Parameters Vs. Setpoint
14
Gain SchedulingCoefficients vs. q / t
15
Gain SchedulingCoefficients vs. q / t
16
Gain SchedulingFinal Equations
Given q, t, setpoint, ambient, max temp
17
Gain Scheduling Effects on Control
18

Gain Scheduling Results

• Gain Scheduling through this method is definitely
  possible, although it requires extensive data
  collection
• Data only covers 0.056
• Need more PID Values vs. q/t to cover 0
• Assumption Gain Schedule is related solely to
  q/t
• Gain Schedule based on Athena results - first
  principles?
• Possible Recommendations
• Automate data collection to perform autotune on
  Athena
• Investigate first principles involved in gain
  scheduling

19
Acknowledgements

• Team Members
• Jake Stern (Leader)
• Peter Scheidler
• Michael Messina (Spring)
• Michael Helfen (Spring)
• Adi Drost (Fall)
• Ari Moradi (Fall)

• Opto 22 Liaison
• Rene Gamero 99
• Faculty Advisor
• Prof. Carl Baumgaertner