Black Boxes & White Noise - Automatic Control · 2012-09-28 · Servomechanism Theory Foundations...

LUND INSTITUTE OF TECHNOLOGY

Black Boxes & White NoiseBlack Boxes & White NoiseThe Evolution of Automatic Control

K. J. Åström

Lund Institute of TechnologyLund, Sweden


1. Introduction2. Early Ideas3. A Discipline Emerges4. The Second Wave5. Conclusions


Natural Science and Engineering ScienceNatural Science and Engineering ScienceMany similarities but also differencies

Natural PhenomenaAnalysis

Isolate phenomenaSimplicityBasic laws

Technical SystemsSynthesisInteractionComplexity

System principles


1.1.1. IntroductionIntroductionIntroduction2. Early Ideas3.3.3. A Discipline EmergesA Discipline EmergesA Discipline Emerges4.4.4. The Second WaveThe Second WaveThe Second Wave5. 5. 5. ConclusionsConclusionsConclusions


Industrial Process ControlIndustrial Process Control

Windmills Mead 1787

Steam enginesWatt, Boulton 1788Maxwell 1868Routh 1875

Water turbinesStodola 1893Hurwitz 1895


Accuracy & StabilityAccuracy & Stability

PID Control

u (t ) = k( t ) + e (s)ds +Tddedt∫

HoneywellTaylor InstrumentLeeds & NorthrupFoxboro

e ( ITi

t

o)


Flight ControlFlight ControlThe Wright Brothers 1903

Sperry 1912

Fully Automatictransatlantic flight 1947

Apollo 1969


Minorsky 1922Minorsky 1922

It is an old adage that a stable ship is difficult to steer.


TelecommunicationTelecommunication

SYSTEM DATECHANNELS

PER PAIRLOSS IN DB

(3000 MI)REPEATERS

(3000 MI)

1st Transcontinental 1914 1 60 3–6

2nd Transcontinental 1923 1–4 150–400 6–20

Open Wire Carrier 1938 16 1000 40

Cable Carrier 1936 12 12000 200

First Coaxial 1941 480 30000 600


The Feedback AmplifierThe Feedback Amplifier

V2V1

= −R2R1

⋅1

1+ 1A

1+R2R1

⎛

⎝ ⎜

⎞

⎠ ⎟

–A

R1

2R

V21V

Black's patent 1928

Granted 1937

“Singing” = Instability

Nyquist 1932

Bode 1945Network Analysis andFeedback Amplifier Design


1.1.1. IntroductionIntroductionIntroduction2.2.2. Early IdeasEarly IdeasEarly Ideas3. A Discipline Emerges4.4.4. The Second WaveThe Second WaveThe Second Wave5. 5. 5. ConclusionsConclusionsConclusions


Industrial Process ControlTelecommunications

Flight ControlMathematics

PrinciplesTheory

Design MethodologyApplications

A Discipline EmergesA Discipline Emerges


War PressuresWar Pressures

National Defense Research CommitteeMIT Radiation LaboratoryMIT Servomechanism LaboratoryMIT Instrumentation LaboratoryMIT Lincoln Laboratory


The Black Box ConceptThe Black Box Concept

OutputInput

AbstractionInformation hiding


The Black Box View ofThe Black Box View ofDynamical SystemDynamical System

LinearitySuperpositionSinusoidsOne pair sufficies

Input Output


2

0

–2

0 180° 360°

aϕ

The Notion of Transfer FunctionThe Notion of Transfer Function

u y

G

G (s ) =L y{ }L u{ }

(s ) = e−stg (t )dt = L (g)0

∞

∫


System PrinciplesSystem PrinciplesFeedforward Feedback

Combination

FF Process

FB Process

–1

Σ ΣProcess

FB Process

–1

Σ

FF


Two ParadigmsTwo Paradigms

Feedback

Open loopActs only on deviationsMarket drivenUnmeasurable disturbanceLess accurate model

Feedforward

Closed loopAct before deviations occurPlanningMeasurable disturbanceAccurate model


Servomechanism TheoryServomechanism Theory

FoundationsComplex variablesLaplace transforms

System ConceptsFeedbackFeedforward

Design methodologyFrequency responseGraphical methods

Analog simulation

Implementation


ConsequencesConsequences

EducationApplicationIndustrialization

OrganisationJournalsConferences


1.1.1. IntroductionIntroductionIntroduction2.2.2. Early IdeasEarly IdeasEarly Ideas3.3.3. A Discipline EmergesA Discipline EmergesA Discipline Emerges4. The Second Wave5. 5. 5. ConclusionsConclusionsConclusions


The Second WaveThe Second Wave

Feedback from applicationsChallenging problemsNew technologyNew ideas


Key ElementsKey Elements

Reexamination of fundamentalsVital interaction with other disciplinesTheory to match new technology


Two views of Dynamical SystemsTwo views of Dynamical Systems

External DescriptionElectrical engineeringInput/OutputBlack Box

Internal descriptionsMechanical EngineeringThe notion of state

y = g (x,u)

dxdt

= f (x,u)


““Modern” Control TheoryModern” Control Theory

Optimal controlComputer controlStochastic controlRobust control

CACESystem identificationAdaptive controlIntelligent control


Optimal ControlOptimal Control

EulerLagrangePontryaginHamiltonJacobiBellman

1707–17831736–181319621805–18651804–18511957

ATTITUDE AND THRUST CONTROL€ Swivel Outer Four Engines

€ Optimum to Desired End Conditions

€ Inertial Updating of Position and Velocity

GUIDANCE – SATURN

NAVIGATION

SII STAGEFive J-2 Engines

2, 500, 000N Thrust


Modeling DisturbancesModeling Disturbances

Power spectraWhite noiseInnovations


Stochastic Control TheoryStochastic Control Theory

Filtering and prediction

Merger of calculus ofvariations and theory ofrandom processes

Decision making underuncertainty

Industrial process control

••

•

•

Probability density

Set point for regulatorwith low variance

Set point for regulatorwith high variance

Testlimit

Process output


System IdentificationSystem Identification

Simplified diagram of a kraft paper machine

Thick stockdilution valve

Thick stockflow valve Beta ray

gauge

Moisturemeter

White water

MODEL OF PROCESS DYNAMICS AND DISTURBANCES


Control of Basis WeightControl of Basis Weight

Wet basis weight

Set point of stuff gate level


Adaptive ControlAdaptive Control

Design Estimator

Regulator Processr

uy

Regulatorparameters


Dual ControlDual ControlControl actions should be both directing and investigating

Consequences for decision making decisions under uncertainty

•

•


Two PrinciplesTwo Principles

Certainty Equivalence (H. Simon 1956)Make the best estimate act as if it was true.

Dual ControlControl should be investigating as wellas directing.


Computer Aided Control EngineeringComputer Aided Control Engineering

How to disseminate complicated technology?Conceptual simplicity computational sophisticationCombine human intuition with computational powerNice way to package theory


Computer ControlComputer Control


Control Design & Process DesignControl Design & Process Design


The Internal Model PrincipleThe Internal Model Principle

Controller Process

y

K� –LΣ

xC

B

–y–C

xΣ

uB Σ � Σ

A

v e

A


ApplicationsApplicationsEnergy generationEnergy transmissionProcess controlDiscrete manufacturingInstrumentationTelecommunication

TransportationHeating, ventilation, airconditionEntertainmentPhysicsBiologyEconomics


Mission CriticalMission Critical

Flight ControlSpace flightAutomotiveCD playerCamcorder manufacturing

DC-motor

photo diodes radial arm

Optical Pick-upUnit


The Mercedes AThe Mercedes A--classclassAutomatic control gives extra freedom to the designer

Unstable behavior improved byElectronic Stabilization Program (ESP)

ESP


Committee on Policy Optimisation HSO 1978

To consider the present state of developmentof optimal control techniques as applied to macro-economic policy. To make recommendationsconcerning the feasibility and value of applyingthes techniques within Her Majesty’s Treasury.

Control and EconomicsControl and Economics


1.1.1. IntroductionIntroductionIntroduction2.2.2. Early IdeasEarly IdeasEarly Ideas3.3.3. A Discipline EmergesA Discipline EmergesA Discipline Emerges4.4.4. The Second WaveThe Second WaveThe Second Wave5. Conclusions


Some ChallengesSome Challenges

The gap between theory and practice“Intelligent” systemsMan–machine interfacesTechnology / society interfacesAcademic positioning


Mathematical Modeling

Implementation

Commission

Operation

Real World Automatic Control

EnergyTransportationCommunicationManufactoringInstrumentationEntertainmentBiologyEconomics

AnalysisSimulationSynthesis


ConclusionsConclusions

A glimpse of an emerging fieldAutomatic control is pervasiveSome system principlesMany challenges ahead

Black Boxes & White Noise - Automatic Control · 2012-09-28 · Servomechanism Theory Foundations...

Documents

Transcript of Black Boxes & White Noise - Automatic Control · 2012-09-28 · Servomechanism Theory Foundations...