Greg McMillan shares how to reduce tuning time for near integrating processes. Recorded video version available for viewing at: http://www.screencast.com/t/NmUxZTBiNTg

1. Interactive Opportunity Assessment Demo and Seminar (Deminar) Series for Web Labs – PID Tuning for Near-Integrating Processes June 23, 2010 Sponsored by Emerson, Experitec, and Mynah Created by Greg McMillan and Jack Ahlers www.processcontrollab.com Website - Charlie Schliesser (csdesignco.com)

5. Time (seconds) % Controlled Variable (CV) or % Controller Output (CO)  CO  CV  o  p K p =  CV  CO  CV CO CV self-regulating process time constant Self-regulating process gain (%/%) Response to change in controller output with controller in manual observed process deadtime Self-Regulating Process Response Most temperature loops have a process time constant so much greater than the deadtime, the response is a ramp in the allowable control error about setpoint and are thus termed “ near- integrators”

6. Lambda Tuning for Self-Regulating Processes Self-Regulation Process Gain: Controller Gain Controller Integral Time Lambda (Closed Loop Time Constant)

7. Near Integrator Gain Approximation For “Near Integrating” gain approximation use maximum ramp rate divided by change in controller output The above equation can be solved for the process time constant by taking the process gain to be 1.0 or for more sophistication as the average ratio of the controlled variable to controller output Tuning test can be done for a setpoint change if the PID gain is > 2 and the PID structure is “ PI on Error D on PV” so you see a step change in controller output from the proportional mode

8. Fastest Possible Tuning for Maximum Disturbance Rejection For max load rejection set lambda equal to deadtime Substitute Into Tuning for max disturbance rejection (Ziegler Nichols reaction curve method gain factor would be 1.0 instead of 0.5) For setpoint response to minimize overshoot

9. Reduction in Tuning Test Time The near integrating tuning test time (3 deadtimes) as a fraction of the self-regulating tuning test (time to steady state) is: If the process time constant is greater than 6 times the deadtime Then the near integrating tuning test time is reduced by 90%: For our example today: The near integrator tuning time is reduced by 97%!

11. Rapid Process Model Identification and Deployment Opportunity For the manipulation of jacket temperature to control vessel temperature, the near integrator gain is Since we generally know vessel volume (liquid mass), heat transfer area, and process heat capacity, We can solve for overall heat transfer coefficient (least known parameter) 4 CV SUB First Principle Parameters = f ( K i ) CO n-1 Value of controller output (%) from last scan ∆ CO θ o K P K i ∆ CV Switch ODE ( K i ) ∆ CV ∆ CV Sum CV n-1 Value of controlled variable (%) from last scan K P = CV o / CO o process gain approximation  P = K P /K i negative feedback time constant  P + = K P /K i positive feedback time constant Methodology extends beyond loops to any process variable that can be measured and any variable that can be changed CO  P K P  P + 1 2 3 ∆ CV

12. Rapid Process Model Identification and Deployment Opportunity The observed deadtime ( θ o ) and integrator gain (K i ) are identified after a change in any controller output (e.g. final control element or setpoint) or any disturbance measurement. The identification of the integrator gain uses the fastest ramp rate over a short time period (e.g. 2 dead times) at the start of the process response. The models are not restricted to loops but can be used to identify the relationship between any variable that can be changed and any affected process variable that can be measured. The models are used for processes that are have a true integrating response or slow processes with a “near integrating” response (  P  θ o ). The process deadtime and integrating process gain can be used for controller tuning and for plant wide simulations including but not limited to the following types of models: Model 1: Hybrid ordinary differential equation (ODE) and experimental model Model 2: Integrating process experimental model Model 3: Slow self-regulating experimental model Model 4: Slow non-self-regulating positive feedback (runaway) experimental model Patent disclosure filed on 3-1-2010

14. Values at Start of Output Change

15. Values at End of Deadtime

16. Values at End of 1 st Deadtime Interval

17. Values at End of 2 nd Deadtime Interval

18. Tuning for Today’s Example For setpoint response to minimize overshoot Lambda tuning equations for integrating processes would give similar results if Lambda (arrest time) is set equal to the observed deadtime (see next Deminar for more details)

20. Help Us Improve These Deminars! WouldYouRecommend.Us/105679s21/

22. QUESTIONS?