PID to Model Predictive Control Yurong Kimberly Wang Adjunct Professor - OIT Standards Certification Education & Training Publishing Conferences & Exhibits Objectives • Industrial process control challenge • PID control limitation • Model predictive control development procedures • MPC for operation advantage • MPC vendors and reference materials Tyco Electronics / Precision Interconnect Coax Manufacturing Processes • Dielectric layer – Taping or Extrusion • Shield layer – Braiding or Serving • Jacket layer – Taping or Extrusion Coax Property • Formulae 16 .95 k C ln( D ) d Zo 59.96 D ln( ) d k Td 1.016 k • • • • • • C: capacitance (pF/Foot) Td: time delay (ns/Foot) Z0: impedance (Ohm) k: dielectric constant D: outer diameter (Mil) d: center conductor diameter (Mil) Process Control Challenge • Multiple Outputs – Capacitance, Diameter, Time delay, Impedance, … • Multiple Inputs – Screw speed, line speed, barrel temperatures, tape tensions, … • Long and Variable Time Delays – Variable line speeds and sensor to actuator distances • Input and Output Constraints – Input and output upper and lower spec limits • Nonlinearity – Variety of operating conditions • Disturbances – Center conductor variation, tape thickness variation, … PID Control Limitation • • • • • • Multiple-loop PID with decoupling Cascade PID loops Gain scheduling PID Anti-windup for input constraints Difficult to control large time delay processes Difficult to control non-minimum phase processes Model Predictive Control (MPC) • • • • • Model-based multi-variable control Optimal control law with I/O constraints Nonlinear control with model mismatch Long and variable time delay processes Non-minimum phase processes MPC System and Optimization MPC Sampling Instants Tuning parameters: prediction horizon and control horizon Process Modeling Tools • Models based on first principals – Mechanics, thermodynamics, heat transfer, fluid dynamics, … – S or Z domain or state space models • Models based on system identification – – – – Step response method: TF, FIR, BJ, ARX, ARMAX,… PRBS method: TF, FIR, BJ, ARX, ARMAX,… MatLAB System Identification Toolbox LabVIEW System Identification Toolkit Simulation Tools • MatLAB and Simulink • LabVIEW Control Design and Simulation Module • Tuning parameters – – – – – Output weights: the higher the weight, the closer the output to setpoint Input weights: the higher the weight, the closer the input to setpoint Input change weights: the higher the weight, the slower the response Predictive horizon: up to plant settling time Control horizon: case specific for each control objective PID and MPC Setpoint Following Comparison of MPC and PID with Smaller Time Delay 0.7 PVPID PVMPC SP 0.6 Case 1: time delay is 41 steps y 0.0025 41 z u z 0.35 0.5 0.4 0.3 0.2 0.1 0 0 50 100 150 200 PID scrap MPC scrap 250 300 350 400 450 PID and MPC Setpoint Following Comparison of MPC and PID with Larger Time Delay 0.9 PVPID PVMPC SP 0.8 Case 2: time delay is 82 steps y 0.0025 z 82 u z 0.35 0.7 0.6 0.5 0.4 0.3 0.2 Same tuning parameters 0.1 0 0 100 200 300 400 500 600 700 800 900 MPC with Predictive SP Comparison of MPC with Predictive SP and PID 0.6 PVPID PVMPC SP 0.5 Case 3: setpoint profile is used y 0.0025 41 z u z 0.35 0.4 0.3 0.2 0.1 0 -0.1 0 50 100 150 200 250 300 350 400 450 LabVIEW MPC System Architecture Quality Engineers Plant Managers Process Engineers Productio n Productio n Productio n Manufacturing Engineers Productio n Report Program for Data Analysis Business Network Internet Manufacturing Information Server Remote Users Productio n OPC Client & Server for Data Logging Business Network LabVIEW HMI & MPC Control OPC Client & Server for Data Sharing Productio n Control Network Local Control Module Local Control Module Local Control Module Figure 1. System Architecture Local Control Module LabVIEWTM MPC Project LabVIEW MPC Block Diagram MatLABTM MPC Block Diagram LabVIEW MPC Application MPC for Operation Advantage • Six Sigma process performance and optimal product quality control – Multi-variable auto-controlled product quality with constraints • Productivity improvement – Unmanned auto production overnight run and throughput ramp up • Equipment cost reduction – Inner diameter gauge elimination • Sensor fault detection – Controller acting up with sensor fault readings • Labor cost reduction – Coax off-line test and operator/machine ratio reduction MPC Vendors • • • • • • • • Aspen Technology Honeywell Emerson Process Management Siemens Shell Global MathWorks National Instruments … Reference Material • A survey of industrial model predictive control technology, Control Engineering Practice, 2003 by S. J. Qin, and T. A. Badgwell • Advanced Control Unleashed, ISA, 2003 by Terrence L. Blevins, Gregory K. McMillan, Willy K. Wojsznis, and Michael W. Brown • LabVIEW Model Predictive Control Module User Manual, 2009 by National Instruments • MatLAB Model Predictive Control Toolbox User Manual, 2009 by MathWorks • MatLAB System Identification Toolbox User manual, 2009 by MathWorks Conclusion • MPC for complicated process controls • MPC development procedures • MPC for operation advantage • MPC vendors and reference materials Question and Answer • Contact information yurongwang1@yahoo.com • “A survey of industrial model predictive control technology” website: http://cepac.cheme.cmu.edu/pasilectures/darciodolak/Review_ article_2.pdf • “Advanced Control Unleashed” website: http://www.isa.org/Template.cfm?Section=Books1&template =Ecommerce/FileDisplay.cfm&ProductID=6087&file=Adv .ControlUnleashed_TableofContents.pdf