IFAC2002 konferansen
Barcelona, 20.26. juli 2002

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Monday 22. July:
Plenary session:
Luenberger D.G.: Systems Concepts in Finantial Pricing Theory

*Demonstrated how to apply system theory concepts to pricing.
*Finance theory: 1) Pricing (determination of a fair price of an asset)
2) Portfolio design (determination of the best portifolio
given existing prices or estimates of future prices)
Relation between system theory and pricing:
system theory concepts contribute to the developement of the pricing theory
pricing theory change the way important systems and controlled problems
are formualted, e.g. changed objective function.
Session: Measurementbased Optimization
Lee K.S., H. Ahn, I.S. Chen, J.H. Lee, D.R. Yang:
Optimal Iterative Learning Control Of Wafer Temperature Uniformity In Rapid
Thermal Processing


Welz C., B. Srinivasan, D. Bonvin:
Evaluation Of Measurementbased Optimization Schemes For Batch Distillation

*measurementbased optimization=use measurements to compensate for the
uncertainty
measurementbased optimization schemes are classified and compared to
a batch distillation column with terminal quality constraint
Cruse A.*, W. Marquardt, A. Helbig, J.S. Kussi:
Optimizing Adaptive Calorimetric Model Predictive Control of A Benchmark
SemiBatch Reaction Process

(optimizing adaptive calorimetric MPC scheme applied to solve the combined
temperature control and feed rate optimization problem of the benchmark
semibatch reaction process.)
Govatsmark M.S., S. Skogestad:
Selection of Controlled Variables And Robust Setpoints


Cheng J.H., E. Zafiriou:
Results Analysis for Iterative Feedback Steadystate Optimization

*Conventional RTO/online optimization:
repetitive online optimization is used to improve process operation
by determining the setpoints for the lower level feedback systems
optimalization results analysed before implementation to find out if
the inherent variability is caused by measurement error or model
uncertainty/disturbances
*iterative feedback optimization=similar to numerical optimization, where
the iteration in numerical optimization is replaced by steadystate operation
period in process operation. Process measurements are used to update the
gradient information directly without requiring model parameter updating.
(similar idea as Tyssedal presented at PROST annual meeting 2001)
Francois G., B. Srinivasan, D. Bonvin*:
Runtorun OPtimization Of Batch Emulsion Polymerization

Pcontrol used to control the terminal constraint for repeted batch
operation
Session: Production Planning & Optimization Models
Four of six presentations cancelled!
Tian Y., T. Sawaragi, Q. Wei, N. Sannomiya:
An Application of DBR And Buffer Manangement In Solving Real World Problems

operational research problem
Latre L.G., M.T.M. Rodrigues:
Sequential Approach To Production Planning in Multisite Environments

present a sequential approach to production plaining
solved by using MILP
typical operational research solution
Plenary Lecture:
Blasi A.: Conditions for Successful Automation in Industrial Applications

how to apply automation successfully in the industry
Panel Discussion: Future Directions of Automatic Control

Chairman (R. Izerman) does not understand why so many are dealing with
feedback control (2/3) when feedforward control is the future: Feedback is
always coming too late and we get improved models
(Just comments from the chairman).
Proporsal to replace 'signal approach' by some 'energyminimizing approach'
which gives a simplified description of the system. Otherwise the problem
formulation is too complex and the resulting control performance is too
conservative.
IFAC in the future:
*More turned against application and less against theory
*How to detect and include new areas faster?
What is the future vision of control?
E.g. medicine community: solve the cancer problem.
Want to try to make a summary of what was said in the discussion 
check homepage
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Tuesday 23. July:
Plenary Lecture:
Futura K.: SuperMechano Systems

*Supermechano systems (SMS) = new mechanical systems with selforganising
capabilities of its structure and functions
Idea: simulatan design of objectiveconfigured mechanism and appropriate
controller  fusion of mechanic and control to attain high system performance
Session: Power Plant Modelling And Simulation
Erlich I., U. Bachmann:
Dynamic Behavior of Variable Speed Pump Storage Units in The German Electric
Power System

simulation studies with respects to a new pump storage power plant Goldistal
in the German power company VEAG
studied the dynamic behavior of the power plant in interaction with the power
system  voltage variation
Matthias H.B., P. Angerer, E. Doujak, B. List:
Modelling The Influences of Hydro Power Refurbishing Projects

developing a knowlegde based system to evaluate the technical conditions
by analysing the effect of different component on selected parameters
=> results in a list of possible improvements (prioritied)
Weber H.W., M. Hladky, T. Haase, S. Spreng, C.N. Moser:
High Quality Modelling of Hydro Power Plants for Restoration Studies

verification studies of existing plans for network restart after blackouts
in the European electrical energy systems require detailed models of
electrical system, especially the power plants
here: high quality models of different types of hydro power plants in
Switzerland
Checked: Many details, but no nonlinear model presented.
Sulc, Dlouhy, Hrdlicka: Simulated Boiler Pressure Pulsation in Comparison
with Experimental Measurement

two approaches (detailed + simplified model) to study the effect of
the geometry on the flue gas side of the reboiler with resepct to pressure
variations => Use: boiler design
FerrariTrecate G., E. Gallesty, A. Stothert, G. Hovland, P. Letizia,
M. Spedicato, M. Morari, M. Antoine:
Modelling And Control of CoGeneration Power Plants Under Consideration
of Lifetime Consumption: A Hybrid System Approach

model of combined cycle power plant in the Mixed Logical Dynamic framework
(include startup, switching of turbines, etc.)
economic optimization is done by using MPC (economic objective, not setpoint
objective).
+no detailed physical model of the plant, but many interesting references and
interesting problem formulation. Check
Session: Advanced Control Concepts for Power Plants I
Liu G.P., S. Daley, G.R. Duan: Application of Optimaltuning PID Control
to Industrial Hydraulic Systems


Aleotti L., C. Aurora, P. Colombo, L. Magni, F. Pretolani, R. Scattolini,
G. Villa: Multivariable Predictive Control of A Thermal Power Plant

Apply MPC (industrial MPC) to a power plant (here: detailed, nonlinear
simulator) to improve the effiency in the operation (motivated:
liberation of the energy market).
Clear problem definition  Check.
Session: Nonlinear Process Control I
Motivation:
Many people claim that nonlinear control are the future.
When should nonlinear control be applied?
Other nonlinear controllers than MPC based on nonlinear model,
gain scheduling, adaptive control.
Guay M., T. Zhang:
Adaptive Extremum Seeking Control of Nonlinear Dynamic Systems With
Parametric Uncertainties

*extremum seeking control = find the operating setpoints that maximize or
minmize the an objective function
(motivation: some applications the control objective is the optimization
of some objective function which depend on unknown plant parameters or the
selection of the desired states to keep a performance function at its
extremum value.)
Selfoptimizing control is mentioned (no references):
"Selfoptimizing control and extremum seeking control are two methods to
handle these kinds of optimization problems."
presented toy example: simple selfoptimizing solution (x1/theta1)
Hernjak N., F.J. Doyle, R.K. Pearson:
Controlrelevant Characterization of Classes of Process Systems

define class of nonlinearities and corresponding properties, try to differ
between controlrelevant and not controlrelevant nonlineraities
two examples (isotherm CSTR and nonisoterm reactor) are used to demonstrate
the different types.
interesting  check more: relevant selfoptimizing control nonlinearities
Monningmann M., W. Marquardt:
Bifurcation Placement of Hopf Points for Stabilization of Equilibria


Session: Nonlinear Process Control II
Boskovic D.M., M. Krstic:
Boundary Control Of Chemical Tubular Reactor

Stabilizing plug flow reactor by using modelbased, feedback control:
Controlling a set of compositions and temperatures inside the reactor
by manipulating on the inlet temperature and inlet composition
Brown J., D. Dochain, M. Perrier, F. Forbes:
Modal Decomposition of A Nonlinear Tubular Reactor Model: A Control Perspective

modal decomposition to a nonlinear convectionreactiondiffusion
distributed parameter system: dynamic model of an industrial pulp
bleaching tubular reactor
a discretized, finitedimensional dynamic model which contains the dominate
states of the process dynamics, is used to controller design
Waller J.B., H.T. Toivonen: A Neurofuzzy Model Predictive Controller
Applied to A Phneutralization Process

nonlinear MPC based on neurofuzzy process model
good: strong nonlinearities depend strongly on the operation point
(general wrt. nonlinear control)
applied to pHneutralization (both reference tracking and disturbances)
Wu W.: Output Regulation of Reactor Systems with Actuator Constraints


Longhi L.G.S., P.R. Barrera:
Disturbance Attenuation of An Experimental Ph Neutralization System

Motivation (nonlinear control): The process gain shows large variation
within the expected operating range.
Nonlinear Hinfinity control (based on solving HamiltonJacobi equations)
is implemented and compared with PIDcontrol wrt. pHcontrol around
neutralization point in a CSTR (strong acid  strong base)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Wednesday 24. July:
Plenary Lecture:
Goodwin G.C.: Inverse Problems With Constraints

*The inverse problem = given an output (e.g. a measurement) determine the
input which lead to that output.
Difficult: illconditioned and/or constrained (input / output)
Signal processing, telecommunication and control are all dealing with inverse
problems, but with different solutions and aspects:
Control  stability
Telecommunications  performance
Signal processing  implementation
Learn of each other!
Session: Process Modelling and Analysis Tools
Scharwaechter H., L. von Wedel, A. Yang, W. Marquardt:
A Tool Integration Framework For Dynamic Simulation in Process Engineering

Motivation: Integration of different simulation tools for the design and
operation in process plants may be necessary in complex simulation tasks
in a technically and economically manner
propose a "simulation framework approach" towards systematic support to
tool integration (Prototype CHEOPS developed)
identified and analyzed key issues connected to simulation tool integration
Dadhe K., S. Engell, R. Gesthuisem, S. Pegel, M. Volker:
Control Structure Selection for A Reactive Distillation Column

Motivation: Integrated processes (e.g. reaction and separation) reduce the
degrees of freedom and require tight control to operate the process at
optimal conditions.
Presented a systematic control structure selection consisting of
prescreening of structures and calculation of the attainable performance
based on available measurements and input variables.
applied to heterogous catalyzed esterfication of methyl acetate in a batch
reactive distillation column.
Wittenmark B., M.E. Salgado:
Hankelnorm Based Interaction Measure For InputOutput Pairing

presented a new interaction measurement for stable multivaiable system
(Hankel norm index)
dynamic extention of idea in the classic RGA
advantages: frequency dependent interactions and appliable in inputoutput
pairing
(What about frequency dependent RGA?)
Petersson M., K.E. Årzen, H. Sandberg, L. de Mare:
Implementaion of A Tool for Control Structure Selection

Present a method which indicate if a considered SISOloop will benefit
from an addition of feedforward control from a measured disturbance
Extended to feedback control in the future?
Eskinat E.: Consistency Relations in Process Modeling

Mass and energy balances in a process give dependent elements in a transfer
function matrices => consistency relation
Important wrt data reconcilation, model reduction, model uncertainty.
Horowitz B., J. Liebman, C. Ma, T.J. Koo, T.A. Henzinger, A. Sangiovanni
Vincentelli, S. Sastry: Embedded Software Design And System Integration for
Rotorcraft Uav Using Platforms

Helicopter control
Skipped afternoon and evening sessions  sightseeing in Barcelona.
Missed plenary lecture of J. Orasanu: Risk Perception: A Critical Element
of Aviation Safety and panel discussion, e.g. Playing with Automation:
Toys and/or Science.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Thursday 25. July:
Plenary Session:
Kopetz K.: Timetriggered Realtime Computing

presented the basic principle for the realtime computing system
realtime computing systems concern with both value and temporal correctness
*task="computational or communication process which is started whenever the
environment delivers a service request".
*eventtriggered task = task started by an external event
*timetriggered task = task started when the global timereaches a specific
value
Session: Linearization Of Nonlinear Systems
Motivation:
Initial: similar idea as selfoptimizing control?
Simplified controller design?
Deutscher J.: Approximate InputOutput Linearization Of Nonminimum Phase
Nonlinear Systems With Linear Unforced Dynamics

(presented an approximate inputoutput linearization approach for
nonminimum phase nonlinear system with linear unforced dynamic)
Ge S.S., Z. Sun, T.H. Lee:
Criterion For Nonregular Feedback Linearization With Application

(presented a new criterion for nonregular feedback linearization for nonliner
systems with 2 inputs)
Session: History of Control: 19551975
Motivation: Always good to know the history.
Owens D.: A Historical View Of Multivariable Frequency Domain Control

19601970: new frequency domain methods in the analysis and design of
multivariable systems based on Nyquist and root locus theory.
british/persional review of ideas, concepts and technics.
Dorato P.: A History Of Analytic Feedback Systems

A review of the history of analytic/ syntesis method for linear feedback
system design.
*Analytic design=design methods where an existence theorem is available
(alt:synthesis) and computable algorithm for computing a solution is
available when the solution exists (e.g. LQG, poleplacement)
Classes of methods: analytic design methods, ad hoc design methods (usually
when the controllers have fixed structure or lower order), numerical
design methods
Session: Applications of Robust Control I
Motivation: Numerical robustness? Industrial application?
Castano F., M.G. Ortega, F.R. Rubi:
A Multivariable Hinfinity Controller For A Rotary Dryer

design and implemented a multivariable Hinfinity controller for drying
process (cocurrent rotary dryer to evaporate moisture from a twophase cake)
readjusting the weight function until achieved good control
(Comment: Multivariable Frequency Analysisbook is very good)
One more???
Session: Linear Model Predictive Control
Motivation: Objective criterium? Advantages? Compared to online dyn.
optimization.
McAvoy T.: Model Predictive Statistical Process Control  Handling Step Upsets


Kakalis N.M.P., V. Dua, V. Sakizlis, J.D. Perkins, E.N. Pistikopoulus:
A Parametric Optimization Approach For Robust MPC

presented an algorithm for deriving a explicit robust modelbased controller
robust optimal controller is derived offline as function of the states
Advantage: avoid repetitive solution of online optimization, online
implementation is reduced to a sequence of simplefunction evaluations.
applied to the wellknown evaporization example
Ramos C., J.S. Senent, X. Blasco, J. Sanchis:
Lpdmc Control Of A Chemical Plant With Integral Behavior

describe the implementation of a hierarchical distributed control system
from a chemical plant using industrial standard components:
regulatory control (PID), multivariable control (MPC), optimization and
economic layer (LP)
Question: What is new compared to industrial practice?
Shows all details  not done by the industry.
Schafer J., A. Cinar:
Multivariable Mpc Performance Assessment, Monitoring and Diagnosis

presented a method for determine a benchmark and monitoring MPC performance
online
Performance measure:
Monitoring: Ratio of historical and achieved performance
(something wrong happend?)
Diagnosis : Ratio of design and achieved performance
(acceptable control performance?)
ex: linear and a nonlinear model of the wellknown evaporation process
Session: Large Scale System I
Labibi B., B. Lohmann, A.K. Sedigh, P.j. Maralani:
Decentralized Quantitative Feedback Design Of LargeScale Systems

Cancelled
Session: Nonlinear Model Predictive Control
Martinsen F.*, L.T. Biegler, B.A. Foss:
Application Of OPtimization Algorithms To Nonlinear Mpc

evaluated use of different discretization methods and SQPoptimization
algorithms in nonlinear model predictive control through a case study of a
CSTR.
Bloemen H.H.J., V. Verdult, T.J.J. van den Boom, M. Verhaegen:
Bilinear Versus Linear Mpc: Application To A Polymerization Process

compare a linear modelbased and a bilinear modelbased identification
and predictive control:
1. Blackbox identification of inputoutput data from a nonlinear,
firstprinciple model
2. Blackbox model applied in MPC to control the nonlinear process
(ex: free radical polymerization of methyl methacrylate)
3. Bilinear MPC better than linear MPC
Session: Control Applications of Optimization
Motivation: Some good casestudies for applying selfoptimizing control?
Few people at this session: 7 including chairman and the speaker
One speaker had probably relative relaxed preparation to the presentation:
The transparents was just copied from his article!
Tarasyev A.M.*, C. Watanabe: Optimal Trends In Models Of Economic Growth

(applied optimal control in composition of production, technology stocks
and rates in a nonlinear model of growth)
Tigrek T., S. Dasgupta, T.F. Smith: Nonlinear Optimal Control Of Hvac Systems

applied optimal control to HVACsystems (Heating, Ventilation and Air
Conditioning) (1/3 of the energy consumption in the US)
objective: energy effiency vs user comfort
controlled variables already selected  no real objective function defined
Session: Advanced Control Concepts For Power Plants and Power Systems II
Carpanzano E., L. Ferrarini, C. Maffezzoni: A Bottomup Methodology For
Testing Complex Control Functions Of Process and Power Plants

presented a modular simulationbased technique for the automatic verification
of logic control functions => define a structured bttomup methodlogy for
testing of the overall control functions of process and power plants
Watanabe T.: Robust Decentralized Turbinegovernor Compressor Control
Subject To Constraint On Turbine Output


Banquett dinner on the old railway station
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Friday 26. July:
Plenary Session:
Dormido S.: The Learning Of Control: Present and Future


Session: Fundamental Control Performance Limitation and Design Tradeoff
Skogestad, S., K. Havre, T. Larsson: Control Limitaions For Unstable Plant


Metha, G.H., A. Banaszuk, C.A. Jacobsen: Framework For Studying Limitations
Of Achieveable Performance In Control Of Nonlinear Combustion Processes


Chen J., S. Hara, G. Chen:
Best Tracking And Regulation Perfromance Under Control Effort Constraints:
Twoparameter Controller Case


Perez T., G.C. Goodwin*, M.M. Seron:
Cheap Control Fundamental Limitation Of Input Constrained Linear Systems

