Multiple Input, Multiple Output I: Numerical Decoupling By Peter Woolf University of Michigan Michigan Chemical Process Dynamics and.

Slides:



Advertisements
Similar presentations
Dynamic Behavior of Closed-Loop Control Systems
Advertisements

Control Architectures: Feed Forward, Feedback, Ratio, and Cascade
MATH 224 – Discrete Mathematics
MIMO systems. Interaction of simple loops Y 1 (s)=p 11 (s)U 1 (s)+P 12 (s)U 2 (s) Y 2 (s)=p 21 (s)U 1 (s)+p 22 (s)U 2 (s) C1 C2 Y sp1 Y sp2 Y1Y1 Y2Y2.
Modeling Basics: 4. Numerical ODE Solving In Excel 5. Solving ODEs in Mathematica By Peter Woolf University of Michigan Michigan Chemical Process Dynamics.
Introductory Statistics By Peter Woolf University of Michigan Michigan Chemical Process Dynamics and Controls Open Textbook version.
Control of Multiple-Input, Multiple-Output Processes
1 Outline Control structure design (plantwide control) A procedure for control structure design I Top Down Step 1: Degrees of freedom Step 2: Operational.
CHE 185 – PROCESS CONTROL AND DYNAMICS PID CONTROL APPLIED TO MIMO PROCESSES.
Chapter 9 Gauss Elimination The Islamic University of Gaza
Using process knowledge to identify uncontrolled variables and control variables as inputs for Process Improvement 1.
Bayesian Networks I: Static Models & Multinomial Distributions By Peter Woolf University of Michigan Michigan Chemical Process Dynamics.
Comparing Distributions I: DIMAC and Fishers Exact By Peter Woolf University of Michigan Michigan Chemical Process Dynamics and Controls.
Comparing Distributions II: Bayes Rule and Acceptance Sampling By Peter Woolf University of Michigan Michigan Chemical Process Dynamics.
Dynamical Systems Analysis III: Phase Portraits By Peter Woolf University of Michigan Michigan Chemical Process Dynamics and Controls.
Enhanced Single-Loop Control Strategies
Comparing Distributions III: Chi squared test, ANOVA By Peter Woolf University of Michigan Michigan Chemical Process Dynamics and Controls.
Andrew Kim Stephanie Cleto
Dynamical Systems Analysis I: Fixed Points & Linearization By Peter Woolf University of Michigan Michigan Chemical Process Dynamics.
Control of Multiple-Input, Multiple- Output (MIMO) Processes 18.1 Process Interactions and Control Loop Interactions 18.2 Pairing of Controlled and Manipulated.
Multiple Choice Explanation Chlorine Ross Bredeweg Ryan Wong.
Dynamical Systems Analysis II: Evaluating Stability, Eigenvalues By Peter Woolf University of Michigan Michigan Chemical Process Dynamics.
Dynamical Systems Analysis IV: Root Locus Plots & Routh Stability
Course AE4-T40 Lecture 5: Control Apllication
Modeling Basics: 1. Verbal modeling By Peter Woolf University of Michigan Michigan Chemical Process Dynamics and Controls Open Textbook version 1.0 Creative.
Multivariable systems Relative Gain Array (RGA)
Chemical Process Controls: PID control, part II Tuning
Multiple Input, Multiple Output II: Model Predictive Control By Peter Woolf University of Michigan Michigan Chemical Process Dynamics.
Improving Stable Processes Professor Tom Kuczek Purdue University
Lecture II-2: Probability Review
PID Controllers Applied to MIMO Processes
Algebra Problems… Solutions
College Algebra Fifth Edition James Stewart Lothar Redlin Saleem Watson.
Proportional/Integral/Derivative Control
MIMO Multiple Input Multiple Output Communications © Omar Ahmad
Chapter 5. Loops are common in most programming languages Plus side: Are very fast (in other languages) & easy to understand Negative side: Require a.
Control Loop Interaction
RELATIVE GAIN MEASURE OF INTERACTION We have seen that interaction is important. It affects whether feedback control is possible, and if possible, its.
Ch. 6 Single Variable Control
ME451 Kinematics and Dynamics of Machine Systems Numerical Solution of DAE IVP Newmark Method November 1, 2013 Radu Serban University of Wisconsin-Madison.
Process Control: Designing Process and Control Systems for Dynamic Performance Chapter 20. Multiloop Control – Relative Gain Analysis Copyright © Thomas.
CSC 4510 – Machine Learning Dr. Mary-Angela Papalaskari Department of Computing Sciences Villanova University Course website:
IT253: Computer Organization Lecture 3: Memory and Bit Operations Tonga Institute of Higher Education.
College Algebra Acosta/Karwoski. CHAPTER 1 linear equations/functions.
Kinematic Redundancy A manipulator may have more DOFs than are necessary to control a desired variable What do you do w/ the extra DOFs? However, even.
Topic 5 Enhanced Regulatory Control Strategies. In the last lecture  Feedforward Control –Measured Vs Unmeasured Loads –Purpose of feedforward control.
Introduction to Linear Algebra Mark Goldman Emily Mackevicius.
Hank Childs, University of Oregon Isosurfacing (Part 3)
Chapter 9 Gauss Elimination The Islamic University of Gaza
1 Decentralized control Sigurd Skogestad Department of Chemical Engineering Norwegian University of Science and Tecnology (NTNU) Trondheim, Norway.
1 Decentralized control Sigurd Skogestad Department of Chemical Engineering Norwegian University of Science and Tecnology (NTNU) Trondheim, Norway.
Statistical Data Analysis 2010/2011 M. de Gunst Lecture 10.
2.1 Notes – Represent Relations and Functions
1 II. Bottom-up Determine secondary controlled variables and structure (configuration) of control system (pairing) A good control configuration is insensitive.
The Mechanical Simulation Engine library An Introduction and a Tutorial G. Cella.
CSE 425: Industrial Process Control 1. About the course Lect.TuTotal Semester work 80Final 125Total Grading Scheme Course webpage:
Camera Calibration Course web page: vision.cis.udel.edu/cv March 24, 2003  Lecture 17.
Central limit theorem - go to web applet. Correlation maps vs. regression maps PNA is a time series of fluctuations in 500 mb heights PNA = 0.25 *
Multi-Variable Control
بسم الله الرحمن الرحيم وبه نستعين
Transfer Functions Chapter 4
Fredrik Bengtsson, Torsten Wik & Elin Svensson
Control of Multiple-Input, Multiple-Output Processes
Outline Control structure design (plantwide control)
Enhanced Single-Loop Control Strategies
The Simplex Method The geometric method of solving linear programming problems presented before. The graphical method is useful only for problems involving.
Numerical Analysis Lecture 2.
Example regulatory control: Distillation
Example “stabilizing” control: Distillation
Outline Control structure design (plantwide control)
Presentation transcript:

Multiple Input, Multiple Output I: Numerical Decoupling By Peter Woolf University of Michigan Michigan Chemical Process Dynamics and Controls Open Textbook version 1.0 Creative commons

Inputs and Outputs Up to now, we have relied on our intuition about a process to decide how to connect sensors to actuators, but there are a few problems that can arise: (1)Sometimes these parings can be difficult to determine (2)Sometimes it is difficult to distinguish between pairings (3)Sometimes the parings we think will work don’t work (4)Sometimes there are no pairings that will work Questions: Are there other, data driven ways to see if a system can be decoupled? How can we evaluate the best pairing in a more objective way?

Four terms: 1) SISO: Single input, single output. Simplest to design, uses data from one sensor to control one thing. PID controller TC v1 2) SIMO: Single input, multiple output. Uses data from one sensor to control multiple things. #1 PID controller TC v1 #2 PID controller TC v2 3) MISO: Multiple input, single output. More complex as it uses data from multiple sensors to control one thing. E.g. cascade control #1 PID controller TC v1 FC #2 PID controller Set point

4) MIMO: Multiple Input, Multiple Output. Hardest to design as it integrates multiple sensor data to coordinate multiple actuators MIMO controller TC v2 FC v1 v3 General strategy: MIMO controllers are more complex, and as such designers often try to avoid them. Process design attempts to minimize cross-talk between sensors and actuators if possible. Even if there is significant cross talk, can we get away with a simpler controller by finding approximate pairings? Note: MIMO controllers are generally not PID, and as such are often designed for each particular case.

Example Process Schematic Process C AO Q CATCAT Simplified Process schematic Simplified Process Model

gain array Elements of the array: Note that this is not the jacobian.. Experimental alternative Change each single input at a time and observe outputs Calculate local change in outputs and these are entries in the gain array

Intuitive evaluation of gain arrays: Good or bad?

Intuitive evaluation of gain arrays: Best case because each manipulated variable exactly controls only one output. Here Cao controls Ca and Q controls T. Good or bad?

Intuitive evaluation of gain arrays: Good or bad?

Intuitive evaluation of gain arrays: Good or bad? Worst case. Manipulated variables can’t individually control outputs. System is fully coupled. How can we measure the degree of coupling?

Condition number (CN) Tells how near to a linearly dependent system we are based on a single value decomposition (SVD)

Condition number (CN) Tells how near to a linearly dependent system is based on a singular value decomposition (SVD) of the gain array Ratio of largest to smallest value=CN =5.46/0.36= 15.1 In Mathematica Terms Map[MatrixForm,{u, w, v}=SingularValueDecomposition[G]]

CN limits

The rule of thumb is that CN>50 is too hard to decouple.. What does CN=50 mean in this matrix above?

CN limits The rule of thumb is that CN>50 is too hard to decouple.. What does CN=50 mean in this matrix above?

In mathematica terms In matrix terms Intuitively and mathematically RGA is a normalized form of the gain matrix that describes the impact of each control variable on the output relative to the control variables impact on other variables.

A few features of an RGA matrix: The nearer all of the entries are to 1 the more decoupled the system is. The best pairing is found by taking the max of the RGA matrix for each row. Each row and each column of the RGA sum to 1

Example: Gain array from a system where the pressure and temperature of a reactor can be controlled by two valves v1 and v2. CN=402 Interpretation: Poor decoupling The same as Interpretation: All values far from 1, so poor decoupling. However, best coupling would link v2 to T and v1 to P. Interpretation: v2 influences T a little bit more than P, but both v1 and v2 strongly influence T and P.

What pairing would work best here? Therefore can be decoupled. Best pairings Note that the pairings are easier to identify using the RGA vs. G matrix.

Material from previous lectures

Take Home Messages It is possible to numerically determine if a system can be decoupled using the condition number RGA helps determine optimal pairings Not all systems can be effectively decoupled Even if a system can be decoupled, it may or may not perform well