PIDControlfor the future Haruo TAKATSU Yokogawa Electric Corporation! 1!
Agenda! 2!
Discussion Items 1. Will PID control continue to be used in the future? Market Survey in Japan 2. When and why is derivative action used? Flexibility of PID Control block in DCS 3. What are alternatives? Self-Tuning / Adaptive Control, Internal Model Control 4. What is your favorite tuning method? PID Tuning in our system 5. Do we know everything about PID or are there any research needs? Introduce Monitoring & Diagnosis! 3!
Needs of PID Control SICE Report! 4!
1400 Control Technology Survey 90 Applications 80 1200 70 1000 60 O(Others) PW(Power plant) E(Engineering) CR(Cement and ceramic) 800 Applications 600 50 40 PP(Pulp and paper) CF(Chemical, fiber,film) PC(Petrochemical) R(Refinery) 400 30 20 200 10 0 0 I-PD and 2 degrees of freedom PID PID auto tuning Gain scheduling! 5! Model predictive control Fuzzy control Dead time compensation Decoupling PID Rule based control Optimization Kalman filter Neural network Optimal regulator Observer H infinity control Adaptive control Exact linearizarion Repetitive control Sliding mode control S(Steel and metal) Takatsu, H., Itou, T. : Future Needs for Control Theory in Industry Report of the Control Technology Survey in Japanese Industry, IEEE Transaction on Control Systems Technology, 07, 03 pp.298-305 (1999) Control Technologies
PID Control Structure Derivative Action! 6!
Control Block in DCS! 7!
PID Control DV PID SV - 1 1 T is T d s 100 PB MV Process PV DV PI-D SV - 1 1 T is - Tds α 100 PB MV Process PV DV I-PD 1 T ds 1 αt ds SV 1 - T is 1 Tds - 100 PB MV Process PV! 8! Page 8
PID Alternatives Adaptive / SelfTuning Control! 9!
Self-Tuning Control Estimated Model PID Tuning Model Estimation PID parameters Response Monitoring SV PID Control MV Process PV! 10!
Self-Tuning Control! Nonlinear Programming! Xmid Xmax Xbar Xref Xmin In case oscillating and small overshoot, Xexp Increase/Decrease P and decrease I. Reflection: Xref = (1a)Xbar! axmax In case No oscillation, slow convergence, Expansion: Decrease Xexp= P bxref(1!b)xbar &I. Xmid Xmax Xmid Xred Xmax Xcon Contraction: Xred In case large overshoot and Xcon= fast convergence, cxmax(1!c)xbar Xmin Increase/decrease P & I.! 11! Reduction In case slow oscillation and convergence, Decrease P and increase I.
PID Alternatives IMC Control! 12!
Internal Model Control DV Q(s) SV MV PV C(s) P(s)! - P M (s) P M (s) - DV C(s) SV MV PV Q(s) P(s)! P M (s) - Controller:C(s), Internal Model: P M (s), Process:P(s)! 13!
Internal Model Control SV Kc(1T P s) MV PV 1λT P(s) P s! K P e!lp s 1T P s! 14! 2012.1.6-20
PID Tuning Method IMC Method! 15!
PID Tuning Stable Process Gain Reset Derivative d PI τ - PID 2τ t K(2ε t d ) 2 2τ t d K(2ε t d ) τ t d t d 2 τt d 2τ t K=Process Gain, t d =Dead time τ ε =Time Constant =Desired Closed Loop Response d! 16! Integral Process Gain Reset Derivative P - - PI 2 - PID 1 Kε Kε td 2ε 2 td K( ε) 2 2 ε2 td 2ε 2 t d 2
PID Tuning The Advanced Tuning window is for user to perform detail analysis of each step test data and confirm the final suitable P, I, D parameters. Loop Details Pane most of the important parameters in loop configuration Step Pane the list of steps captured under the selected tag and the preview of the step data Loops Pane displays loops grouped by the Location defined in Loop Configuration the main window that user does the analysis for model and loop simulation Advanced Tuning Pane! 17!
PID Tuning is always correct?! 18!
Research & Development in the Future Plant Life Cycle Monitoring & Diagnosis! 19!
Loop Oscillation & Valve Stiction Dianosis Valve Stiction Results: Oscillating Only! 20!
Loop Oscillation & Valve Stiction Diagnosis Valve Stiction Results: Oscillating & Sticking! 21!
What is problem after installation? Industry has implemented many projects Shift from project work towards maintenance Staff is less appealed by maintenance work Shortage of skilled staff in process control and optimisation Maintenance can be tedious and time consuming Too much focus towards Uptime (Online factor) Applications do not deliver optimum performance Reference : - Recent survey of 20,000 loops performance Excellent 16% Acceptable 12% Fair 28% Poor 11% Manual mode 33%! 22!
Thank you for your attention! 23!