Supervisory

Control Engineering Modeling and simula6on HCDi- PhD program

FIT Sept 14th,2011 P. Millot

Modeling methodolgy
A model is a descrip.on or a representa.on of a phenomenon of any nature (physical, chemical, technological, biological, nancial, human ) and called process The purpose can be - to understand how it proceeds - to control or to reproduce it. Dierent ways of modeling regarding :
The size and the complexity of the process The knowledge (and its nature) we can have on it The accessibility for observing it The use we want to do with it (control, reproduce, understand)

The rst models were physical descrip.ons with equa.ons

Model of a Physical System Suspension of a car

Strengths: - Spring: -k (x-x0) - Damp: - f x - Inertia: m x Dieren6al equa6on of 2nd order: mx+ fx+ kx = u Normalized in: x + 2 z w x+ w2x = w2u
With: w: proper pulsa6on Z: damping coecient

Solving Through Laplace Transform

Proper6es: - Linearity: - Dieren6al equa6ons Algebric equa6ons: - Exponen6al func6ons Ra6onal frac6ons

Applica6on

Where: X is the output Laplace Transform, U is the sollicita6on of the spring Example when U = 1/s (step func6on) Solving this ra6onal func6on and applying the inverse Laplace tranform gives 2 cases according to Z values

Results
X(t)

Transitory period Permanent state X(t)

t
Transitory period Permanent state

Other example: rst order tranfer func6on

0.95k

0.63 k

3T

Transitory period

Permanent state

Func6onal diagram

Control in open loop:

u(s) x(s)

u(s) is the input x(s) the output

H(s) = x(s)
u(s)

Control in close loop: feedback

x0 is the set-value x is the output obtained through a sensor e is the error b is a disturbance C is a controller or a corrector and its output is a command value

Tranfer func6on of the system in close loop

Correc6on of a close loop system

Q: How to minimize the eect of the disturbance b ? A: In choosing C as an appropriate complex func6on C(s) Ie: if C = k (called proportional corrector) and if it is choosen as great as possible

But a\en6on. The system becomes oscillant (instable) . With this simple propor6onal corrector a compromise between precision and stability is needed !

PID corrector
- For reducing the oscilla6ons add a derivator as a corrector ks - For reducing the error add an integrator as a corrector k/s Then:

C (s) =
For instance with:

Let us see a simula6on

Results Without the noise b With the noise b

Remarks

The transitory period must be studied if: there is a risk of instability of the system (risk of breakdown) the transitory period can be cri6cal and its length is non negligible Otherwise, we can take only the system nal states into account ex: for an elevator the nal states are the stairs we want to reach

Iden6ca6on methods For the cases we dont have equa.ons then:

The automa.on scien.sts have dened catalogs of the responses of the most frequent transfer func.ons Then we have to observe the response of the system we want to model and try to match it with one of the system catalog responses. It is called an iden1ca1on Schema of an iden1ca1on through the model method

In any cases: Func6onal Analysis is needed to start modeling

Example: modeling the home hea6ng system with the components: - - - - - The master room The radiator The boiler The fuel pomp A thermometer

Func6onal Analysis to start modeling

Q rad
Temperature Room exchanges

Func6onal Analysis to start modeling

Q rad
Temperature Room exchanges

Radiator

Func6onal Analysis to start modeling

Qboil Boiler Q rad Radiator
Temperature

Room exchanges

Func6onal Analysis to start modeling

Vpomp Qfuel Fuel Pomp Qboil Boiler Q rad Radiator
Temperature

Room exchanges

Open loop hea6ng system for instance when the room is empty

Func6onal Analysis to start modeling

Desired temperature

Vpomp

Qfuel Fuel Pomp

Qboil Boiler

Q rad Radiator

Temperature

Room exchanges

Manual close loop hea6ng system

Func6onal Analysis to start modeling

Desired temperature

Vpomp

Qfuel Fuel Pomp

Qboil Boiler

Q rad Radiator

Temperature

Thermo stat

Room exchanges

Automa6c close loop hea6ng system

Func6onal Analysis to start modeling

Td Desired temperature

Error Td-Tm

Fuel speed Control

Vpomp

Room exchanges

Tm Measured temperature

Thermometer

Temperature

Thermostat principle

Func6onal Analysis to start modeling

Example of control laws of the fuel pomp: - If Error 0 then switch the fuel pomp on - If Error 0 then switch the fuel pomp o

Examples of func6onal models of the human

- Model of the human informa6on processing - Model of the human ac6vity regula6on - Model of the human solving and decision making processes - Model of the human in his/her socio-technical environment

Action sur Se

Informations passes Mmoire court terme action sur Me Hypothse du canal unique capacit limite Traitement dpendant de la tche vitesse v
Systmes dexcution effecteurs

CODAGE XE

Infos utiles XE Sommation

Ze Me

Zone utilise en traitement

Ze+Za-Cmax =C XE Comparaison

rponses

Se Signal visuel

Me

Action sur Ma

Mmoire long terme

action sur v

NOUVEAUTE XA Codage Signaux autres rcepteurs XA Ma S

Za

ma

XA Actions sur S

Gnrateur de puissance

Capacit affecte au Traitement des C perturbations Codage Perturbations PE

Cmax

Systme activateur

PE

MODEL of the human information processing (BEYAERT, MALVACHE 79)

Actions sur Se

Tasks demands Disturbances

+

Work capacity + allowed to the task

Choice of operating modes

Applic. of operating modes

Performance Workload

Human Operator state Maximum work capacity Cognitive state (expertise) Psycholog. State: Motivation,stress Physiolog.state: fatigue, health General individual parameters: Physical and mental capabilities Personal rythm of life Parameters linked to the task: Biological rythms, social environmt
-

Individual Parameters for coping the task: motivation, expertise

-

Model of the human 28 activity regulation (Millot,03)

HUMAN PROBLEM SOLVING

EVALUATE in relation to system constraints

INTERPRET consequences for system goals

Knowledge-based behavior Target goal diagnosis/prognosis (risk assessment) INDENTIFY (ANTICIPATE) system state (evolution)

DEFINE task

Hypothesis elaboration and test Rule-based behavior SEARCH for information explicity FORMULATE procedure

expectations DETECT abnormal conditions Skill-based behavior EXECUTE

29

Rasmussen s step-ladder revisited (after Hoc 96)

UNDERSTANDING THE HUMAN ACTION CONSTRAINTS

Functionnal Border for safety acceptability Prescribed Limit for safety acceptability Limit of acceptability of the financial losts (organization pressure)

Migration Margin

Degree of freedom for the human action

Limit of available ressources (human ressources)

Compromizes for Human action (adapted of Rasmussen 97, Amalberti 96)

30

Mul6purpose models - Supervisory control - Automobile in the road trac - Human driver

31

Supervisory control of automated systems

SUPERVISION
detection

Supervision H/M

Diagnosis/ prognosic

Decision/ planing

(1) maintenance (2) accommodation (3) reconfiguration

Objectives

Formalization (3)

+ _

control H/M (2)

Process (1)

Performance . Quantity, Quality of production . Safety

Low level control

OBJECTIFS : Analyse Car driving within the traffic

Other vehicles, Other users Objectives V2

driver V2

Interface

Vehicle V2
Trajectory V2 Fluidity, safety

Procedures Coordination Barriers

Interaction

Infrastructure

Objectives V1

Driver V1

Interface

Vehicle V1

Trajectory V1

33

Human-Machine task sharing

Trafic

Supervision Task sharing Coordination/ Traffic

Other vehicles

Decision Support System V1

Infrastructure

Objectives V1 Barriers

Driver V1

Interface

Vehicle V1

Safety

Interaction
Trajectory

34

Drivers tasks in 3 dimensions (after Summala, 1996)

35

Drivers tasks in 3 dimensions (after Summala, 1996)

Navigation Guidance Control

36

Drivers tasks in 3 dimensions (after Summala, 1996)

Conscious control Subconscious Control Navigation Guidance Control

37

- Complexity = (large size system) (highly interconnected variables) Ends

Part

Synthesis Modeling through dierent abstrac6on levels

Goal Level Function Level Engineering Sciences

Whole

{!

Behaviour Level Component Level Means

}!

Cognitives Sciences

- Means-End analysis through several hierarchized view points (Lind 03)

38

Exercice
Draw a func6onal schema of the car driving task in its dierent facets: - guidance on the lane through the steering wheel - control the car speed through the accelera6on and breaking pedals - the outputs of the system are: the car trajectory, its speed . Complete the func6onal schema according to several assistance tools: - Speed regulator - ACC automa6c Cruise Control - an ABS - a GPS