01 Systems

Systems Systems Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
General concepts General concepts
– 1
– 1
Modelling, Modelling,
identification and identification and
analysis analysis
A system
20
20
Control,
optimisation and
Systems Control,
optimisation and
validation validation
Fault diagnosis
Linear systems Fault diagnosis
A system can be defined as a set of elements (or components) that coop-
Advanced topics in machine learning erate in order to perform a specific functionality that would be otherwise
Classification Classification
impossible to attain for the individual components alone
FC 2,
FC 2,
Representation Representation
Input-output Input-output
representation representation
Francesco Corona
– 2
– 2
State-space State-space This is definition is very fine, but it does not highlight an important fact
The model The model • The dynamical behaviour of the system
Department of Computer Science
G
G
Examples Federal University of Ceará, Fortaleza Examples
System System For us a central paradigm is that a system is subjected to external stimuli
properties properties
• Stimuli influence the temporal evolution of the system itself
U
U
Dynamical v Dynamical v
Instantaneous Instantaneous
Linear v Nonlinear Linear v Nonlinear
A
A
Stationary v Stationary v
non-stationary non-stationary
Proper v improper Proper v improper
With v without With v without
delay delay
Systems Systems (cont.) Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
Modelling,
identification and
A system (reloaded) Modelling,
identification and
analysis analysis
20
20
Control, A system is a physical entity, typically consisting of different interact- Control,
optimisation and optimisation and
validation ing components, that responds to external stimuli producing a deter- validation
Fault diagnosis
mined/specific dynamical behaviour Fault diagnosis
Classification Classification General concepts

FC 2,
FC 2,
Input-output We study how to mathematically model a broad variety of systems Input-output Systems
– 2
– 2
State-space State-space
representation Our scope is to analyse their dynamical behaviour representation
The model The model
! We want to operate them appropriately
G
G
Examples Examples
! The design of control devices
System System
properties ! Under external stimuli properties
U
U
A
A
Stationary v The methodological approach shall be formal and system independent Stationary v
delay delay
Systems General topics Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis analysis
20
20
Control, Control,
There is a wide spectrum of problems that spin around systems theory
Fault diagnosis
Classification
Fault diagnosis
Classification
Modelling, identification and
analysis
FC 2,
FC 2,
Representation
! System modelling, identification and analysis Representation
representation
• System control, optimisation and verification representation
General concepts
– 2
– 2
The model The model
• System diagnosis
G
G
Examples Examples
System System
U
U
A
A
delay delay
Systems Modelling Systems Modelling (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
Example
– 1
– 1
Modelling,
identification and
Modelling,
identification and
Consider the electric circuit consisting of two serially arranged resistors
analysis analysis
20
20
Control,
optimisation and
Control,
optimisation and
The current flow i(t) [A] thru the system depends on tension v (t) [V]
Fault diagnosis
To study a system, the availability of a mathematical model is crucial Fault diagnosis
Classification Classification R1
! A quantitative description of the behaviour of the system
FC 2,
FC 2,
• R1 = 1 [Ω]
The model is often constructed on the knowledge of the component devices • R2 = 3 [Ω]
– 2
– 2
State-space State-space v(t) R2
• A knowledge of the laws the system obeys to must be available, too i(t)
The model The model
G
G
Examples Examples
System System
Both resistors can be assumed to follow Ohm’s law1
U
U
Linear v Nonlinear Linear v Nonlinear ! v (t) = (R1 + R2 )i(t) = 4i(t)
A
A
Proper v improper Proper v improper "
delay delay
1
The potential difference (‘voltage’) across an ideal conductor is proportional to
the current that flows through it. The proportionality constant is called ‘resistance’.
Systems Identification Systems Identification (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
Example
8
General concepts At times, we only have an incomplete knowledge on the system’s devices General concepts
– 1
– 1
Consider the electric circuit consisting of two serially arranged resistors
identification and
analysis
• The model must be constructed from observations identification and
analysis
20
20
Control,
optimisation and
• (Observations of the system behaviour) Control,
optimisation and
The current flow i(t) [A] thru the system depends on tension v (t) [V]
Fault diagnosis Fault diagnosis
Classification Case A) We have a knowledge on the type/number of component devices Classification R1
FC 2,
FC 2,
Representation • Not all of their parameters are known Representation
• R1 = ?
representation • System observations are available representation
R2 • R2 = ?
– 2
– 2
State-space State-space v(t)
representation
! Parametric identification representation
i(t)
The model The model
! White-box identification
G
G
Examples Examples
System System
properties properties Both resistors can still be assumed to follow Ohm’s laws,
U
U
Dynamical v Case B) We have no knowledge on the components and their parameters Dynamical v
Linear v Nonlinear • Observations of the system are available Linear v Nonlinear v (t) = (R1 + R2 )i(t) = Ri(t)
A
A
non-stationary ! Black-box identification non-stationary
R is now an unknown system parameter
delay delay
! It can be identified from data
Systems Identification (cont.) Systems Analysis

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
! "N
8
8
General concepts
We can observe the system by collecting N pairs of measurements (vk , ik ) k=1 General concepts
– 1
– 1
analysis 4 4 analysis
20
20
Control,
3
(a) 3
(b) Control,
Fault diagnosis i 2 i 2 Fault diagnosis Systems analysis is about forecasting the future behaviour of a system
1 1 ! Based on the external stimuli it is subjected to
FC 2,
FC 2,
0 0
Input-output 0 0.5 1 0 0.5 1 Input-output
v v
– 2
– 2
State-space
representation
State-space
representation
The availability of a mathematical model of the system is fundamental
The model The model • Needed to approach the problem in a quantitative manner
G
G
Examples Often, such points will not be perfectly aligned along a line with slope R Examples
System System
properties
! Disturbances alter the behaviour to the system properties
U
U
Dynamical v
Instantaneous
! Measurement errors are always present Dynamical v
Instantaneous
A
A
non-stationary We choose R corresponding to the line that best approximates the data non-stationary
With v without " With v without
delay delay
Systems Analysis (cont.) Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
General concepts Example General concepts
– 1
– 1
identification and
analysis
The marine ecosystem is described thru time evolution of its fauna and flora identification and
analysis
20
20
Control, Control,
optimisation and • Birth-growth-dead processes optimisation and
Fault diagnosis
Classification
The behaviour of the system is influenced by many factors
Fault diagnosis
Classification
Control, optimisation and
validation
FC 2,
FC 2,
Representation • Climate conditions, availability of food, ... Representation
• Human predators, pollutants, ...
General concepts
– 2
– 2
representation • ..., and so on representation
The model G The model
G
Examples Examples
System
They recently spoke of reducing CO2 emissions by injecting it into the sea System
• CO2 dissolves in sea water
U
U
Linear v Nonlinear
The lack of a valid model limits our understanding about the system Linear v Nonlinear
A
A
non-stationary
• We do not know the response of the ecosystem non-stationary
delay " delay
Systems Control Systems Control (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
Example
8
8
Consider a conventional network for the distribution of drinking water
– 1
– 1
analysis analysis • Water pressure must be kept constant throughout the net
20
20
Control, Control,
optimisation and The objective of control is to impose a desired behaviour on a system optimisation and
We can measure pressure at various network locations
Classification We need to explicitly formulate what we mean by ‘desired behaviour’ Classification

• Locations have nominal (target) pressure values
FC 2,
FC 2,
Representation ! The specifications that such behaviour must satisfy Representation
Input-output Input-output Specs suggest instantaneous variations within ±10% of nominal value
– 2
– 2
State-space
representation
We need to design a device for implementing this task, the controller State-space
representation
The model The model
We identify two stimuli that act on the system (and modify its behaviour)
! The scope of a controller is to stimulate the system
G
G
Examples Examples • The flow-rate of water withdrawn from the network
! Drive its evolution toward the desired behaviour
System System • The pressure imposed by the network pumps
U
U
Linear v Nonlinear Linear v Nonlinear We cannot control water withdrawals, they are understood as disturbances
A
A
Pump pressures can be manipulated to meet specifications
With v without With v without • This manipulation is done by the controller
delay delay
"
Systems Optimisation Systems Optimisation (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
identification and
analysis
identification and
analysis
Example
20
20
Control,
optimisation and
Control,
optimisation and
Consider a conventional suspension system of a conventional car
validation Achieve a certain system’s behaviour, while optimising a performance index validation
Fault diagnosis Fault diagnosis It is designed to satisfy two different needs
Classification
• Optimisation can be understood as a special case of control Classification
! Appropriate passengers’ comfort
FC 2,
FC 2,
! Good handling in all conditions
We impose a desired behaviour, while optimising a performance index
– 2
– 2
The model
• The index measures the quality of the behaviour of the system
G The model Modern cars have suspensions based on ‘semi-active’ technology
G
Examples • (Economic or operational terms) Examples • A controller (dynamically) changes the dumping factor
System System
• It guarantees (a compromise between) the two needs
U
U
The optimiser takes into account of cabin and wheel oscillations
A
A
non-stationary non-stationary "
delay delay
Systems Validation Systems Validation (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
analysis analysis
20
20
Control, Control,
optimisation and optimisation and Example
Fault diagnosis Fault diagnosis Consider a conventional elevator
Suppose that a mathematical model of a system under study is available
The system is controlled to guarantee that it responds correctly to requests
FC 2,
FC 2,
Representation
• Suppose that a set of desired properties can be formally expressed Representation
• The controller is an abstract machine
– 2
– 2
representation Validation allows to see whether the model satisfies the properties representation
• Programmable logic controller (PLC)
The model The model
G
G
Examples Examples
System System Formal verification can be used to guarantee the correct functioning
U
U
Dynamical v Dynamical v "
A
A
delay delay
Systems Systems Fault diagnosis
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis analysis
20
20
Control, Control,
Systems deviate from nominal behaviour because of occurrence of faults
Classification Fault diagnosis Classification

! We need to detect the presence of an anomaly
FC 2,
FC 2,
Input-output General concepts Input-output ! We need to identify the typology of fault
– 2
– 2
State-space State-space ! We need to devise a corrective action
G
Examples Examples Fault diagnosis
System System
U
U
A
A
delay delay
Systems Fault diagnosis (cont.) Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
analysis analysis
20
20
Control, Control,
optimisation and
validation
Example optimisation and
validation
Fault diagnosis The human body is a complex system subjected to many potential faults Fault diagnosis
Classification
• We conventionally call them diseases Classification Classification
FC 2,
FC 2,
Systems
– 2
– 2
State-space Consider the presence of fever, or another anomalous condition State-space
The model • Symptoms reveal the presence of a disease The model
G
G
Examples Examples
System A doctor, once identified the pathology, prescribes a therapy System

U
U
Dynamical v " Dynamical v
A
A
delay delay
Systems Classification Systems Time-evolving systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
Modelling,
identification and
The diversity of systems leads to a number of methodological approaches Modelling,
identification and
analysis analysis Time-evolving systems
20
20
Control, • Each approach pertains a particular class of models Control,
validation validation The system/model behaviour is described with functions, or signals
Fault diagnosis
Conventional methodological approaches and model/system classification Fault diagnosis
• The independent variable is time (t or k )
FC 2,
FC 2,
Representation By typology Representation
Input-output
representation
! Time-evolving systems Input-output
representation
!Continuous time-evolving systems
– 2
– 2
State-space
• Discrete-event systems State-space • The time variable varies continuously
The model
• Hybrid systems The model
! Discrete time-evolving systems

G
G
Examples Examples
System By representation System • The time variable takes discrete values

U
U
Dynamical v • Input-output models Dynamical v
Instantaneous Instantaneous The signal can take values in a discrete set, digital time-evolving systems
Linear v Nonlinear ! State-space models Linear v Nonlinear
A
A
delay delay
Systems Systems Time-evolving systems (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
analysis analysis
20
20
Control, Control,
Classification Time-evolving systems Classification

The evolution of such systems is completely based on the passage of time
FC 2,
FC 2,
Input-output Systems by typology Input-output
The functions of system behaviour satisfy differential/difference equations
– 2
– 2
State-space
representation
State-space
representation
• A specification of the relation between functions and their changes
The model The model
G
G
Examples Examples
System System
U
U
A
A
delay delay
Systems Time-evolving systems(cont.) Systems Time-evolving systems(cont.)
UFC/DC UFC/DC
CK0255|TIP8244 Example CK0255|TIP8244
2018.2 2018.2
Continuous time-evolving systems Example
8
Consider a tank whose volume of liquid V (t) [m3 ] varies in time Discrete time-evolving systems
– 1
– 1
analysis
• Variation is due to input and output flow rates analysis
Consider a tank whose volume of liquid V (t) [m3 ] varies in time
20
20
Control, Control,
validation
• (By two externally operated pumps) validation
• Measurements are not continuously available
• Acquisitions only every ∆t units of time
Classification q1 (t) Classification
FC 2,
FC 2,
Representation Representation We are interested in the system behaviour at 0, ∆t, 2∆t, . . . , k ∆t, . . .
Input-output Tank cannot be emptied or filled Input-output
– 2
– 2
State-space State-space q1 (t)
representation hmax d representation
! V (t) = q1 (t) − q2 (t)
The model G dt The model We consider discrete-time variables
G
Examples Examples
V (t) h(t)
System System hmax • V (k ) = V (k ∆t)
properties hmin • Output flow q2 (t) ≥ 0 [m3 s−1 ] properties
U
U
Dynamical v
• Input flow q1 (t) ≥ 0 [m3 s−1 ]
Dynamical v • q1 (k ) = q1 (k ∆t)
V (t) h(t)
Linear v Nonlinear q2 (t) Linear v Nonlinear • q2 (k ) = q2 (k ∆t)
A
A
Stationary v Stationary v hmin
Proper v improper
The differential equation is a relation between continuous-time functions
Proper v improper
For k = 0, 1, 2, . . .
delay delay q2 (t)
V (t), q1 (t), q2 (t)
"
Systems Time-evolving systems(cont.) Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
We approximate the derivative with the difference quotient
– 1
– 1
analysis analysis
20
20
Control, d ∆V V (k + 1) − V (k ) Control,
optimisation and V (t) ≈ = optimisation and
validation d ∆t ∆t validation
Classification
∆V
Classification Discrete-event systems
FC 2,
FC 2,
Representation Multiplying both sides of = q1 (k ) − q2 (k ) by ∆t yields Representation
Input-output ∆t Input-output Systems by typology
– 2
– 2
representation
! V (k + 1) − V (k ) = q1 (k )∆t − q2 (k )∆t representation
The model The model
G
G
Examples Examples
System The difference equation is a relation between discrete-time functions System

U
U
Instantaneous V (k ), q1 (k ), q2 (k ) Instantaneous
A
A
non-stationary " non-stationary
delay delay
Systems Discrete-event systems Systems Discrete-event systems (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
Example
– 1
– 1
analysis analysis Discrete-event systems
20
20
Control, Control,
validation validation Consider a depot where parts are awaiting to be processed by some machine
Discrete-event systems • The number of parts awaiting to be processed cannot be larger than 2
FC 2,
FC 2,
Dynamic systems whose states take logical or symbolic values (not numeric) • The machine can be either healthy (working) or faulty (stopped)
representation
The behaviour is characterised by the occurrence of instantaneous events representation
– 2
– 2
representation
! [At irregular (perhaps unknown) times] representation
L0 a L1 a L2
The model G The model The state of the system
G
Examples Examples • Number of awaiting parts
System System p p
• Status of the machine
U
U
Dynamical v Dynamical v r g r g r g
The events of the system
A
A
Stationary v Stationary v a a • Changes in state
G0 G1 G2
delay delay
Systems Discrete-event systems (cont.) Systems Discrete-event systems (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
L0 a L1 a L2
– 1
– 1
identification and identification and L0 a L1 a L2
analysis
p p
Six possible states analysis
20
20
Control, Control,
• L0 , L1 and L2
optimisation and
validation r g r g r g
optimisation and
validation p p
Four possible events
Fault diagnosis • G0 , G1 and G2 Fault diagnosis
• a and p
Classification Classification r g r g r g
a a • g and r
FC 2,
FC 2,
Input-output G0 G1 G2 Input-output a a
– 2
– 2
representation representation G0 G1 G2
The model The model
• L0 , the machine is working and the depot is empty
G
G
Examples Examples
System
• L1 , the machine is working and there is one part in the depot System • a, a new part arrives to the depot
• L2 , the machine is working and there are two parts in the depot
U
U
Dynamical v Dynamical v • p, the machines takes one part from the depot
Linear v Nonlinear Linear v Nonlinear • g, the machine gets faulty
A
A
Stationary v
non-stationary
• G0 , the machine is not working and the depot is empty Stationary v
non-stationary • r , the machine gets fixed
Proper v improper
• G1 , the machine is not working and there is one part in the depot Proper v improper
delay
• G2 , the machine is not working and there are two parts in the depot delay
Systems Discrete-event systems (cont.) Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
L0 a L1 a L2
8
– 1
– 1
analysis
p p analysis
20
20
Control, Control,
optimisation and r g r g r g optimisation and
Classification
a a Classification Hybrid systems
FC 2,
FC 2,
G0 G1 G2 Systems by typology
– 2
– 2
representation Event a can only occur when the depot does not have two parts representation
The model The model
#
G
G
Examples Li → Li+1 Examples
a!
System Gi → Gi+1 System
U
U
Instantaneous Event p can only occur when the deport is not empty Instantaneous
A
A
Stationary v
$ Stationary v
non-stationary p ! Li → Li−1 non-stationary
delay Event g and r determine the switches Li → Gi and Gi → Li , respectively delay
"
Systems Hybrid systems Systems Hybrid systems (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2 Example
8
8
Hybrid systems
– 1
– 1
identification and
analysis
identification and
analysis
Consider a modern sauna, a cabin where the temperature is regulated
20
20
Control, Control,
• A thermostat controls a stove used as heat generator
• Keep the temperature between 80◦ C and 90◦ C
A model that combines time-evolving dynamics and discrete-event dynamics
FC 2,
FC 2,
Representation Representation T ≥ 90◦ C
Input-output ! Thus, they are the most general class of dynamical systems Input-output
– 2
– 2
representation representation d d
T (t) = k [Ta − T (t)] T (t) = k [Ta − T (t)] + q(t)
The model The model dt dt
G
G
Examples Examples T ≤ 80◦ C
System System OFF ON
U
U
The thermostat can be represented using a discrete-event model
A
A
Stationary v Stationary v • Switch the heater ON
Proper v improper Proper v improper • Switch the heater OFF
delay delay
The cabin can be represented using a time-evolving model

• Temperature T (t)
Systems Hybrid systems(cont.) Systems Hybrid systems(cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
T ≥ 90◦ C
– 1
– 1
analysis T ≥ 90◦ C analysis
d d
20
20
Control, Control, T (t) = k [Ta − T (t)] T (t) = k [Ta − T (t)] + q(t)
optimisation and optimisation and dt dt
Fault diagnosis d d Fault diagnosis T ≤ 80◦ C
T (t) = k [Ta − T (t)] T (t) = k [Ta − T (t)] + q(t) ON
Classification dt dt Classification OFF
FC 2,
FC 2,
Representation T ≤ 80◦ C Representation
Input-output
OFF ON Input-output When the thermostat is ON
– 2
– 2
representation representation T (t) in the cabin increases, heat generated by the stove q(t)
The model When the thermostat is OFF
G The model
G
Examples Examples d
T (t) in the cabin decreases, heat exchanged with the outside [Ta < T (t)] T (t) = k [Ta − T (t)] + q(t)
System System dt
U
U
Dynamical v d Dynamical v
The state of the system is x = (l , T )
Instantaneous T (t) = k [Ta − T (t)], with k > 0 Instantaneous
Linear v Nonlinear dt Linear v Nonlinear
A
A
non-stationary non-stationary • A logical variable l ∈ {ON, OFF}, representing the discrete state
• A real function T (t) ∈ R, representing the continuous state
delay delay
"
Systems Systems Representation

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
analysis analysis
20
20
Control,
optimisation and
Control,
optimisation and
We provide fundamental concepts for the analysis of time-evolving systems
• Evolution generates from the passing of time
Classification Representation Classification • Focus on continuous-time systems
FC 2,
FC 2,
Input-output Systems Input-output
We introduce the two main forms that are used for describing such systems
– 2
– 2
The model The model
! Mathematic formulations and example(s)
G
G
Examples Examples
System System
properties properties We conclude with a classification, based on some system/model properties
U
U
A
A
delay delay
Systems Representation (cont.) Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis analysis
20
20
Control, Control,
optimisation and
validation
Fundamental step to use formal techniques to study time-evolving systems optimisation and
validation
Fault diagnosis
! We describe the system behaviour in terms of functions Fault diagnosis
Classification Classification Input-output representation
FC 2,
FC 2,
Input-output There are two possible such model/system descriptions Input-output Representation
– 2
– 2
State-space • Input-output (IO) representation State-space
The model ! State-space (SS) representation
G The model
G
Examples Examples
System We define the mathematical elements and properties of these representations System
U
U
A
A
delay delay
Systems Input-output representation Systems Input-output representation (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
General concepts The quantities involved in the input-output (IO) representation of a system General concepts
– 1
– 1
analysis analysis
20
20
Control, Causes Control,
validation • Quantities that are generated outside the system validation
• Their evolution influences the system behaviour A system
FC 2,
FC 2,
Representation
• They are not influenced by the system behaviour Representation The system S can be seen as an operator or a processing/computing unit
representation representation • It assigns a specific evolution to the output variables
Effects
– 2
– 2
representation representation • One for each possible evolution of the input variables
The model
• Quantities whose behaviour is influenced by the causes The model
G
G
Examples • Their evolution is influenced by nature of the system Examples
System System
U
U
Dynamical v
Instantaneous
By convention, #
Dynamical v
Instantaneous
Linear v Nonlinear Causes ! Inputs Linear v Nonlinear
A
A
Stationary v
! Stationary v
non-stationary Effects ! Outputs non-stationary
delay delay
Systems Input-output representation (cont.) Systems Input-output representation (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
Example
A system can have more than one (r ) input and more than one (p) output
8
General concepts General concepts A car (IO representation)
• Both inputs and outputs are assumed to be measurable/observable
– 1
– 1
analysis analysis Consider a car
20
20
Control, Control,
optimisation and
validation
A graphical IO system representation optimisation and
validation Let its position and speed be the output variables
r inputs (in Rr ) • They are both measurable
FC 2,
FC 2,
% &T
Representation
u1(t) y1(t) u(t) = u1 (t) · · · ur (t) Representation As input variables, we can consider wheel and gas position
Input-output .. .. Input-output
representation . S . representation • They are both measurable and manipulable
– 2
– 2
State-space
ur(t) yp(t) p outputs (in Rp ) State-space
The model % &T The model
y(t) = y1 (t) · · · yp (t) By acting on the input variables, we influence the output behaviour
G
G
Examples Examples
System System • The changes depend on the specific system (car)

Manipulable inputs
U
U
• (More precisely, by its dynamics)
• They can be used for control
A
A
Stationary v Stationary v Wind speed could be considered as an additional input variable
Proper v improper
Non-manipulable inputs Proper v improper
• It may be measurable but it is hardly manipulable
With v without • They are called disturbances With v without
delay delay
r = 3 inputs and p = 2 outputs (A MIMO system)
"
Systems Input-output representation (cont.) Systems Input-output representation (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
Suppose that flow-rates q1 and q2 can be set to some desired value (pumps)
– 1
– 1
identification and
analysis
Two tanks (IO representation) identification and
analysis
Also suppose that q3 depends linearly on the liquid level in the tank
20
20
Control, Control,
optimisation and
validation
Consider a system consisting of two cylindric tanks, both of base B [m2 ] optimisation and
validation
• q3 = kh2 [m3 s−1 ], with k [m2 s−1 ]
Fault diagnosis • Output flow-rate from tank 1 is the input flow-rate to tank 2 Fault diagnosis
• k appropriate constant
! q2
FC 2,
FC 2,
Input-output Input-output Inputs, q1 and q2
First tank
– 2
– 2
State-space State-space ! Measurable and manipulable
representation q1 q2 • Input flow-rate q1 [m3 s−1 ] representation
The model The model ! They influence the liquid levels in the tanks
• Output flow-rate q2 [m3 s−1 ]
G
G
Examples Examples
System • h1 is the liquid level [m] System Output, d = h1 − h2

properties d properties
U
U
Dynamical v h1 Dynamical v
! Measurable, but not manipulable
Instantaneous h2 Second tank Instantaneous
Linear v Nonlinear q3=k⋅h2 Linear v Nonlinear
! It is influenced indirectly only through inputs
• Input flow-rate q2 [m3 s−1 ]
A
A
• Output flow-rate q3 [m3 s−1 ] "
(a)
With v without
delay
• h2 is the liquid level [m] With v without
delay
Systems Systems State-space representation
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis analysis
20
20
Control, Control,
Classification State-space representation Classification

For a given behaviour of the inputs, S defines the behaviour of the outputs
FC 2,
FC 2,
Input-output Representation Input-output ! The output at time t is not only dependent on the inputs at time t
– 2
– 2
State-space State-space ! It also depends on the past behaviour (evolution) of the system
G
Examples Examples
System System
U
U
A
A
delay delay
Systems State-space description(cont.) Systems State-space representation (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 Example 2018.2
8
8
General concepts
Two tanks (SS representation) General concepts We can take this into account by introducing an intermediate variable
– 1
– 1
identification and Consider a system consisting of two cylindric tanks, both of base B [m2 ] identification and
A variable that exists between inputs and outputs
analysis analysis
20
20
Control, Control,
optimisation and Let d0 = h1,0 − h2,0 be the output variable at time t0 optimisation and
• The state variable of the system
Fault diagnosis • (h1,0 and h2,0 are the liquid levels at time t0 ) Fault diagnosis
r inputs (in Rr )
FC 2,
FC 2,
Representation Representation u(t) = [u1 (t) · · · ur (t)]T
Input-output Suppose that all input variables (q1 and q2 ) are zero at time t0 Input-output x1(t)
representation representation u1(t) y1(t)
n states (in Rn )
– 2
– 2
State-space State-space .. .. ..
representation q1 q2 representation . . .
The model The model
ur(t) yp(t) x(t) = [x1 (t) · · · xn (t)]T
G
G
Examples Examples xn(t)
• q1,0 = 0 p outputs (in Rp )
System d System ingressi stati uscite
• q2,0 = 0
U
U
h1
Dynamical v
Instantaneous
h2 Dynamical v
Instantaneous
y(t) = [y1 (t) · · · yp (t)]T
q3=k⋅h2
A
A
Proper v improper (a) Proper v improper The state condenses information about past and present of the system
With v without
delay
Output d(t) at any time t > t0 depends on input values q1 (t) and q2 (t) With v without
delay
• Over the entire interval [to , t]
"
Systems State-space representation (cont.) Systems State-space description (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
u1(t) x1(t) y1(t)
.. .. ..
– 1
– 1
Modelling,
identification and
. . . Modelling,
identification and
analysis analysis
20
20
Control,
ur(t) yp(t) Control,
optimisation and
xn(t) optimisation and
stati uscite
In general, it is possible to select different physical entities as state variables
The state vector x(t) = [x1 (t) · · · xn (t)]T has n components
FC 2,
FC 2,
Representation Representation • The state variable is neither univocally defined, nor it is determined
Input-output ! We say that n is the order of the system Input-output
representation representation • It is anything that can be seen as an internal cause of evolution
– 2
– 2
State-space • (In this representation) State-space
• (In general)
G
Examples Examples
Definition
System System
properties State variable properties
U
U
Linear v Nonlinear
The state of a system at time t0 is a variable that contains the necessary Linear v Nonlinear
A
A
Stationary v information to univocally determine the behaviour of output y(t) for t ≥ t0 Stationary v
Proper v improper • Given the behaviour of input u(t) for t ≥ t0 and the state itself at t0 Proper v improper
delay delay
"
Systems State-space representation (cont.) Systems State-space representation (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
Example
8
8
Two tanks (SS representation)
– 1
– 1
analysis
Consider a system consisting of two cylindric tanks, both of base B [m2 ] analysis
20
20
Control, Control,
Fault diagnosis First tank Fault diagnosis
We shall see that we are able to evaluate the output d(t) for t > t0
Classification • Input flow-rate q1 [m3 s−1 ] Classification
q1 q2
FC 2,
FC 2,
Representation • Output flow-rate q2 [m3 s−1 ] Representation
! Need to know the initial state, V1,0 and V2,0 , at t0
representation • h1 is the liquid level [m] representation
! Need to know the input, q1 (t) and q2 (t), in [t0 , t]
– 2
– 2
representation
d representation
The model
h1 Second tank The model "
h2
G
G
Examples Examples
q3=k⋅h2 • Input flow-rate q2 [m3 s−1 ]
System System
properties • Output flow-rate q3 [m3 s−1 ] properties
U
U
Instantaneous (a) • h2 is the liquid level [m] Instantaneous
A
A
non-stationary
Let d0 = h1,0 − h2,0 be the output variable at time t0 non-stationary
With v without • h1,0 and h2,0 are the liquid levels at time t0 With v without
delay delay
As state variable, select the volume of liquid in the tanks, V1 (t) and V2 (t)
Systems State-space representation (cont.) Systems State-space description(cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis Common to choose as state those variables that characterise system energy analysis
20
20
Control, Control,
Fault diagnosis • For a cylindric tank of base B and liquid level h(t), the potential energy Fault diagnosis
Classification at time t is Ep (t) = 1/2ρgV 2 (t)/B , with ρ the density of the liquid and Classification Consider a system in which there is energy stored, its state is not zero
V (t) = Bh(t). V (t) or equivalently h(t) can be used as state variable
FC 2,
FC 2,
Representation Representation • The system can evolve even in the absence of external inputs
– 2
– 2
State-space • For a spring with elastic constant k , the potential energy at time t is State-space The state can be understood as a possible (internal) cause of evolution
The model
Ek (t) = 1/2kz 2 (t) with z (t) the spring deformation with respect to an The model
equilibrium position. z (t) can be used as state variable
G
G
Examples Examples
System • For a mass m moving with speed v (t) on a plane, the kinetic energy at System
U
U
Dynamical v time t is Em (t) = 1/2mv 2 (t). v (t) can be used as state of the system Dynamical v
A
A
delay delay
Systems State-space representation (cont.) Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 Example 2018.2
8
8
General concepts Two tanks (SS representation, reloaded) General concepts
– 1
– 1
identification and
analysis
Consider any of the tanks in the two cylindric tanks system, base B [m2 ] identification and
analysis
20
20
Control, Control,
validation First tank validation
Classification q1 q2
Classification Mathematical model
FC 2,
FC 2,
Input-output • h1 is the liquid level [m] Input-output
Representation
– 2
– 2
State-space
representation
d State-space
representation
The model
h1 Second tank The model
h2
G
G
Examples q3=k⋅h2 Examples
System • Output flow-rate q3 [m3 s−1 ] System

U
U
Dynamical v (a) • h2 is the liquid level [m] Dynamical v
A
A
Stationary v
non-stationary
Each of the tanks can store a certain amount of potential energy Stationary v
non-stationary
Proper v improper • The amount depends on the liquid volumes (levels) Proper v improper
delay delay
The entire (two-tank) system has order 2

"
Systems Mathematical model Systems Mathematical model (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis analysis
Input-output model
20
20
Control, Control,
optimisation and System analysis studies the relations between the system inputs and outputs optimisation and
validation
Fault diagnosis (IO) or, alternatively, between the system inputs, states and outputs (SS)
validation
Fault diagnosis
The relationship between the system output y(t) ∈ Rp and its derivatives,
the system input u(t) ∈ Rr and its derivatives (a differential equation)
FC 2,
FC 2,
Representation We are given certain input functions Representation
representation ! Interest in understanding how states and outputs evolve in time representation
State-space model
– 2
– 2
The model We need a model to describe quantitatively the system behaviour The model It describes how the evolution ẋ(t) ∈ Rn of the system state depends on the
state x(t) ∈ Rn itself and on the input u(t) ∈ Rr (the state equation)
G
G
Examples • The relations between inputs (states) and outputs Examples
System System
properties properties It describes how the system output y(t) ∈ Rp depends on system state
U
U
Dynamical v Dynamical v x(t) ∈ Rn and on system input u(t) ∈ Rr (the output transformation)
A
A
delay delay
Systems Systems Input-output model

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
General concepts General concepts The IO model of a SISO system is given as a differential equation
– 1
– 1
identification and identification and ' ,
analysis analysis
h y(t), ẏ(t), . . . , y (n) (t), u(t), u̇(t), . . . , u (m) (t), t =0
20
20
Control, Control,
( )* + ( )* + ()*+
validation validation output input time
Classification Input-output model Classification

d d dn
FC 2,
FC 2,
Representation Representation • ẏ(t) = y(t), . . . , ẏ 2 (t) = y(t) and . . . , y (n) (t) =
y(t)
Input-output Mathematical model Input-output
dt dt 2 dt n
d d d m
– 2
– 2
State-space
representation
State-space
representation
• u̇(t) = u(t), . . . , u̇ 2 (t) = 2 u(t) and . . . , u (m) (t) = m u(t)
dt dt dt
The model The model
G
G
Examples Examples
System System h is a multi-parametric function that depends on the system

U
U
• n is the maximum order of derivation of the output
• m is the maximum order of derivation of the input
A
A
The order of the system (model) is n
delay delay
Systems Input-output model (cont.) Systems Input-output model (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2 The IO model of a MIMO system with p outputs and r inputs
8
⎧ % (m1,1 ) (m1,r ) *
– 1
– 1
Modelling, Modelling, (n1 )
identification and Consider a system model given by the differential equation identification and ⎪h1 y1 (t ), ẏ1 (t ), . . . , y1
⎪
⎪
⎪
(t ), u1 (t ), u̇1 (t ), . . . , u1 (t ), . . . , ur (t ), . . . , ur (t ), t
analysis analysis ⎪
⎪ & '( ) & '( ) & '( )
20
20
⎪
Control, Control,
⎪
⎪
⎪ output 1 input 1 input r
optimisation and
h optimisation and
⎪
⎪
* +( ) ⎪
⎪ =0
validation - validation ⎪
⎪ % *
(m ) (m )
⎪h2 y2 (t ), ẏ2 (t ), . . . , y (n2 ) (t ), u1 (t ), u̇1 (t ), . . . , u 2,1 (t ), . . . , ur (t ), . . . , ur 2,r (t ), t
⎪
⎪
Fault diagnosis
2 ẏ(t)y(t) +2 ()*+
t u(t)ü (t) = 0 Fault diagnosis ⎪
⎪
⎪
⎪ & '(
2
) & '(
1
) & '( )
Classification ( )* + ( )* + Classification ⎪
⎪
time
⎪
output
FC 2,
FC 2,
input ⎨
Representation Representation =0
⎪
⎪
Input-output Input-output ⎪ .
representation We have, representation
⎪
⎪ ..
⎪
⎪
– 2
– 2
⎪
⎪
State-space State-space ⎪
representation • Output order of derivation, n = 1 representation
⎪
⎪
⎪
⎪
⎪ % *
The model The model ⎪ (np ) (mp,1 ) (mp,r )
• Input order of derivation, m = 2 ⎪
⎪hp yp (t ), ẏp (t ), . . . , yp
⎪
⎪ (t ), u1 (t ), u̇1 (t ), . . . , u1 (t ), . . . , ur (t ), . . . , ur (t ), t
G
G
Examples Examples
⎪
⎪ & '( ) & '( ) & '( )
⎪
⎪
⎪
⎪
⎪ output p input 1 input r
System System
properties
Function h links y and ẏ, and u and ü, and t is the independent variable properties
⎩
=0
U
U
Linear v Nonlinear The relationship explicitly depends on the independent variable (time) Linear v Nonlinear hi (i = 1, . . . , p) are multi-parametric functions depending on the system
A
A
non-stationary √ non-stationary • ni , max order of derivation of the i-th component of output yi (t)
Proper v improper ! t Proper v improper
With v without With v without • mi , max order of derivation of the i-th component of input ui (t)
delay delay
"
A total of p differential equations
Systems Systems State-space model

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
The SS model of a SISO system is NOT a differential equation of order n
8
8
State equation
– 1
– 1
analysis analysis ⎧ % &
20
20
Control, Control, ⎪
⎪ ẋ1 (t) = f1 x1 (t), . . . , xn (t), u(t), t
optimisation and optimisation and ⎪ % &
validation validation ⎨ẋ2 (t) = f2 x1 (t), . . . , xn (t), u(t), t
⎪
..
Classification State-space model Classification
⎪
⎪
⎪
⎪ .
FC 2,
FC 2,
⎩ % &
Representation Representation ẋn (t) = fn xn (t), . . . , xn (t), u(t), t
Input-output Mathematical model Input-output
– 2
– 2
State-space State-space It links the derivative of each state with the other states and the input
The model The model
G
G
Examples Examples Output transformation
System System % &
properties properties y(t) = g x1 (t), . . . , xn (t), u(t), t
U
U
It further links the output with each state variable and the input
A
A
Proper v improper Proper v improper fi with i = 1, . . . , n and g are multi-parametric functions
delay delay • They depend on (are) the dynamics of the system
Systems State-space model (cont.) Systems State-space model (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis analysis
20
20
Control, Control,
validation Let ẋ(t) be the vector whose components are the derivatives of the state validation
Fault diagnosis Fault diagnosis State-space model
⎡ ⎤
Classification d Classification # % &
FC 2,
FC 2,
⎢ dt x1 (t) ⎥ ẋ(t) = f x(t), u(t), t
⎡⎤
Representation ẋ1 (t) Representation ! % &
d ⎢ ⎥
Input-output ⎢ . ⎥ ⎢ .. ⎥ Input-output y(t) = g x(t), u(t), t
representation ẋ(t) = x(t) = ⎣ .. ⎦ = ⎢ . ⎥ representation
– 2
– 2
State-space dt ⎢ ⎥ State-space
representation ẋn (t) ⎣d ⎦ representation
The model xn (t) The model
dt f is a vectorial function whose i-th component is fi , with i = 1, . . . , n
G
G
Examples Examples
System System
U
U
A
A
delay delay

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 The SS model of a MIMO system with r inputs and p outputs 2018.2
8
8
General concepts State equation General concepts
– 1
– 1
identification and ⎧ % & identification and
analysis
⎪ẋ1 (t) = f1 %x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t &
⎪ analysis
20
20
Control,
⎪ Control,
optimisation and
⎨ẋ2 (t) = f2 x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t
⎪
optimisation and
Fault diagnosis ⎪ .. Fault diagnosis
⎪
⎪
⎪ . State-space model
Classification ⎩ % & Classification
ẋn (t) = fn xn (t), . . . , xn (t), u1 (t), . . . , ur (t), t
FC 2,
FC 2,
# % &
ẋ(t) = f x(t), u(t), t
Input-output
representation Output transformation Input-output
representation
! % &
y(t) = g x(t), u(t), t
– 2
– 2
representation ⎧ % & representation
The model ⎪
⎪ y1 (t) = g1 x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t The model
⎪ % &
⎨y2 (t) = g2 x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t
G
G
⎪
Examples Examples
System ⎪ .. System
properties ⎪
⎪ . properties
U
U
⎪
⎩ % &
Instantaneous yp (t) = gp x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t Instantaneous
A
A
Proper v improper Multi-parametric functions depending on the system, f and g Proper v improper
delay • fi with i = 1, . . . , n delay
• gi with i = 1, . . . , p
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis The state equation is a set of n first-order ordinary differential equations analysis
20
20
Control, Control,
optimisation and
• Regardless of the fact that the system is SISO or MIMO optimisation and
validation validation The state-space representation of a system is central in our methods
• It offers a consistent framework for analysing systems
Classification The output transformation is a scalar or vectorial algebraic equation Classification
• Analysis of systems of arbitrary degree of complexity
FC 2,
FC 2,
• Depending on the number p of output variables
– 2
– 2
State-space
representation
State-space
representation
Conversion of a scalar n-th order ordinary differential equation
The model G The model ! n first-order ordinary differential equations
G
Examples u(t) y(t) Examples
x! (t ) = f ( x (t ), u(t ), t ) x(t) y(t)=g(x(t),u(t),t) ! A first-order vector equation, dimension n
System System
U
U
A
A
delay delay

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
Example
– 1
– 1
analysis analysis The three scalar differential equations that must be solved
Consider the third-order scalar equation
20
20
Control, Control,
optimisation and optimisation and ⎡ ⎤ ⎡ ⎤
validation ... validation ẋ1 (t) x2 (t)
x (t) + c3 ẍ (t) + c2 ẋ (t) + c1 x (t) = bu(t)
ẋ(t) = ⎣ẋ2 (t)⎦ = ⎣% x3 (t) &
⎦
Classification Classification ẋ3 (t) − c3 x3 (t) − c2 x2 (t) − c1 x1 (t) + bu(t)
FC 2,
FC 2,
Representation We define, Representation ⎡f %x(t), u(t), t &⎤
⎡ ⎤ ⎡ ⎤ 1
Input-output
representation
x1 (t) x (t) Input-output
representation ⎢ % &⎥
– 2
– 2
State-space x(t) = x2 (t) = ẋ (t)⎦
⎣ ⎦ ⎣ State-space ⎢
= ⎢f2 x(t), u(t), t ⎥
⎥
The model
x3 (t) ẍ (t) The model
⎣ ⎦
% &
G
G
Examples Thus, Examples f3 x(t), u(t), t
System ! ẋ3 (t) + c3 x3 (t) + c2 x2 (t) + c1 x1 (t) = bu(t) System
properties properties A single vector (state) equation
U
U
"
Linear v Nonlinear
This equation can be first integrated to get x3 (t) [! ẍ (t)] Linear v Nonlinear
A
A
non-stationary • Two more integrations to get x2 (t) [! ẋ (t)] and x1 [! x (t)] non-stationary
delay delay
Systems Systems Examples
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
Example
8
– 1
– 1
identification and identification and First tank
analysis analysis
20
20
Control, Control, q1 q2 • Input flow-rate q1 [m3 s−1 ]
validation validation • Output flow-rate q2 [m3 s−1 ]
Classification Examples Classification d

• h1 is the liquid level [m]
FC 2,
FC 2,
Representation Representation h1
h2
q3=k⋅h2
Second tank
– 2
– 2
The model The model
(a)
G
G
Examples Examples • h2 is the liquid level [m]
System System
U
U
Dynamical v
Instantaneous
Dynamical v
Instantaneous
• ui = qi with i = 1, 2, the input variables
• y = d, the output variable
A
A
• x1 = V1 and x2 = V2 , the state variables
delay delay
We are interested in the IO and the SS representations of the system
Systems Examples (cont.) Systems Examples (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
analysis For an incompressible fluid, by mass conservation analysis
Moreover, we have y(t) = d(t) = h1 (t) − h2 (t)
20
20
Control, ⎧ Control,
validation ⎪
⎪ dV1 (t) validation ' 1 1 , ' 1 k ,
Fault diagnosis
⎨ = q1 (t) − q2 (t) Fault diagnosis ẏ(t) = ḋ(t) = ḣ1 (t) − ḣ2 (t) = q1 (t) − q2 (t) − q2 (t) − h2 (t)
dt (1) B B B B
Classification ⎪
⎪ dV 2 (t) Classification ( )* + ( )* +
= q2 (t) − q3 (t) = q2 (t) − kh2 (t)
FC 2,
FC 2,
⎩
Representation dt Representation ḣ1 (t ) ḣ2 (t )
representation representation 1 2 k
– 2
– 2
State-space We can set h1 = V1 /B , h2 = V2 /B , and q3 = kh2 State-space = q1 (t) − q2 (t) + h2 (t)
representation representation B B B
The model ⎧ The model
⎪ 1 1 1 2 k
= u1 (t) − u2 (t) + [h1 (t) − y(t)]
G
G
Examples ⎪
⎨ḣ1 (t) = q1 (t) − q2 (t) Examples
B B B B B
System ! System
properties ⎪
⎪ 1 1 1 k properties
⎩ḣ2 (t) = q2 (t) − q3 (t) = q2 (t) − h2 (t)
U
U
Dynamical v
Instantaneous
B B B B Dynamical v
Instantaneous
We used u1 (t) = q1 (t) and u2 (t) = q2 (t)
A
A
delay delay
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
General concepts 1 2 k k General concepts
k 1 2 k k
ẏ(t) = u1 (t) − u2 (t) + h1 (t) − y(t) ÿ(t) + ẏ(t) − u̇1 (t) + u̇2 (t) − u1 (t) + u2 (t) = 0
B2
– 1
– 1
Modelling,
identification and B B B B
Modelling,
identification and
B B B B
analysis analysis
20
20
Control, Control,
optimisation and
validation
By taking the second derivative of y(t), we have optimisation and
validation The obtained model is in the general IO form
1 2
k k ⎧ %
(n ) (m1,1 ) (m ,r )
*
Classification ÿ(t) = u̇1 (t) − u̇2 (t) +
ḣ1 (t) − ẏ(t) Classification ⎪
⎪h1 y1 (t ), ẏ1 (t ), . . . , y1 1 (t ), u1 (t ), u̇1 (t ), . . . , u1 (t ), . . . , ur (t ), . . . , ur 1 (t ), t
FC 2,
FC 2,
Representation
B BB B Representation
⎪
⎪
⎪
⎪
⎪
& '( ) & '( ) & '( )
Input-output 1 2k k k Input-output
⎪
⎪
⎪
⎪
representation = u̇1 (t) − u̇2 (t) + 2 u1 (t) − 2 u2 (t) − ẏ(t) representation ⎪
⎪ = 0
– 2
– 2
⎪
State-space B B B B B State-space
⎪
⎪
⎨ .
⎪
representation ( )* + representation .
.
The model G k The model ⎪
⎪
⎪
G
ḣ1 (t ) ⎪
⎪
Examples Examples ⎪ % (mp,1 ) *
B ⎪
⎪
⎪
⎪
(np )
hp yp (t ), ẏp (t ), . . . , yp (t ), u1 (t ), u̇1 (t ), . . . , u1
(mp ,r )
(t ), . . . , ur (t ), . . . , ur (t ), t
System System
⎪
⎪
⎪
⎪ & '( ) & '( ) & '( )
properties properties ⎪
⎪ output p input r
U
U
⎪
⎪ input 1
⎩
=0
Linear v Nonlinear
The IO system representation is a ordinary differential equation Linear v Nonlinear
A
A
non-stationary k 1 2 k k non-stationary
! ÿ(t) + ẏ(t) = u̇1 (t) − u̇2 (t) + u (t) − u2 (t) ! p = 1, n1 = 2
2 1
With v without B B B B B With v without
delay delay ! r = 2, m1 = m2 = 1

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
identification and The SS system representation is derived from mass conservation identification and
analysis analysis
20
20
Control, Control,
optimisation and ⎧ optimisation and
validation ⎪
⎪ dV1 (t) validation
The resulting SS representation of the system
Fault diagnosis ⎨ = q1 (t) − q2 (t) Fault diagnosis
! dt ⎧
⎪ dV 2 (t) ẋ1 (t) = u1 (t) − u2 (t)
FC 2,
FC 2,
⎪
⎩ = q2 (t) − q3 (t) = q2 (t) − kh2 (t) ⎪
⎪
Representation Representation ⎪
dt ⎪
⎨ k
representation representation ! ẋ2 (t) = − x2 (t) + u2 (t)
B
– 2
– 2
⎪
⎪
⎪ 1 1
The state equation is obtained by setting h1 = x1 /B and h2 = x2 /B ⎩y(t) = x1 (t) − x2 (t)
The model The model
B B
G
G
Examples Examples
The output transformation,
System System
1 1
U
U
Dynamical v
Instantaneous
y(t) = x1 (t) − x2 (t) Dynamical v
Instantaneous
B B
A
A
delay delay
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
General concepts The model is in the general SS form General concepts
– 1
– 1
identification and
analysis
State equation identification and
analysis
20
20
Control, Control,
⎧ % & In general, the choice of the states is not unique
⎪ẋ1 (t) = f1 %x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t &
validation ⎪ validation
⎪
Fault diagnosis ⎨ẋ2 (t) = f2 x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t
⎪ Fault diagnosis
• We could have chosen the levels as states
⎪ ..
FC 2,
FC 2,
Representation
⎪
⎪
⎪ . Representation • x1 = h1 and x2 = h2
⎩ % &
Input-output
representation
ẋn (t) = fn xn (t), . . . , xn (t), u(t), t Input-output
representation ⎧
– 2
– 2
State-space
representation
State-space
representation
⎨ẋ1 (t) = −Bu1 (t) − Bu2 (t)
⎪
The model
Output transformation The model ! ẋ2 (t) = −kx2 (t) + Bu2 (t)
⎪
G
G
⎩
Examples
⎧ % & Examples y(t) = x1 (t) − x2 (t)
System ⎪ y1 (t) = g1 x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t System
properties
⎪
⎪ % & properties "
U
U
Dynamical v
⎨y2 (t) = g2 x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t
⎪
Dynamical v
Instantaneous
⎪ .. Instantaneous
Linear v Nonlinear ⎪ . Linear v Nonlinear
A
A
Stationary v
⎪
⎪ Stationary v
⎩ % &
non-stationary yp (t) = gp x1 (t), . . . , xn (t), u1 (t), . . . , ur (t), t non-stationary
delay delay
Systems Systems System properties

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
analysis analysis
20
20
Control, Control,
We discuss a set of fundamental properties of time-evolving models
Fault diagnosis Fault diagnosis • Proper v Improper
Classification System properties Classification
• Linear v Non-linear
FC 2,
FC 2,
• With v Without delay
• Dynamical v Instantaneous
– 2
– 2
The model The model
• Stationary v Non-stationary
G
G
Examples Examples • Lumped v Distributed parameters
System System
U
U
Yet another way of classifying dynamical systems/models
A
A
delay delay
Systems Systems Dynamical v Instantaneous
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis analysis
20
20
Control, Control,
Definition
Classification System properties Classification Instantaneity
FC 2,
FC 2,
Input-output Properties Input-output
A system is said to be instantaneous if the value of the output y(t) ∈ Rp
representation representation at time t only depends on the value of the input u(t) ∈ Rr at time t
– 2
– 2
The model G The model A system is said to be dynamical, otherwise
G
Examples Examples
"
System System
U
U
A
A
delay delay
Systems Dynamical v Instantaneous (cont.) Systems Dynamical v Instantaneous (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
analysis Proposition analysis
20
20
Control, Control,
validation
IO representation - SISO validation % &
Fault diagnosis Fault diagnosis h y(t), u(t), t = 0
Classification
A necessary and sufficient condition for a SISO system to be instantaneous Classification
is that the IO relationship is expressed by an equation in the form
FC 2,
FC 2,
Representation Representation If a SISO system is instantaneous, the IO relation is an algebraic equation
representation
% ✭ & representation
This is necessary, but it is not sufficient for a system to be instantaneous
– 2
– 2
✭✭ ✭
, y (n) (t), u(t), ✭ , u (m) (t), t = 0
State-space h y(t), ✭
ẏ(t),
✭.✭ . .✭ u̇(t),
✭✭ . . .✭
✭✭ State-space
% &
The model
! h y(t), u(t), t = 0 The model
Consider a differential equation as IO representation of a SISO system
G
G
Examples Examples
System System
! Then, the system is certainly dynamical
properties ! The order of the derivatives of y and u is zero (n = m = 0) properties
U
U
Linear v Nonlinear " Linear v Nonlinear
A
A
delay delay
Systems Dynamical v Instantaneous systems (cont.) Systems Dynamical v Instantaneous (cont.)
UFC/DC UFC/DC
CK0255|TIP8244
2018.2
Example CK0255|TIP8244
2018.2
Counter-intuition
8
– 1
– 1
identification and Consider the algebraic equation as IO representation of a SISO system identification and Proposition
analysis analysis
20
20
Control, + Control,
optimisation and y(t) = u(t − T ), with T ∈ R optimisation and IO representation - MIMO
A necessary and sufficient condition for a MIMO system with r inputs and
Classification Classification p outputs to be instantaneous is that the IO representation is expressed by
FC 2,
FC 2,
Representation
The output y(t) at time t does not depend on the input u(t) at time t Representation a system of equations in the form
Input-output • It depends on the input value u(t − T ), at a preceding moment Input-output
– 2
– 2
State-space State-space ⎧ % &
representation representation ⎪
⎪h1 y1 (t), u1 (t), u2 (t), . . . , ur (t), t = 0
The model The model ⎪ %
⎨ &
u(t) y(t)=u(t-T) h2 y2 (t), u1 (t), u2 (t), . . . , ur (t), t = 0
G
G
Examples Examples
⎪
⎪
⎪ ···
System System ⎩ % &
properties T properties hp yp (t), u1 (t), u2 (t), . . . , ur (t), t = 0
U
U
Linear v Nonlinear t t Linear v Nonlinear
A
A
"
Such a system is not instantaneous, it is dynamical
delay delay
! Finite time delay system
"
Systems Dynamical v Instantaneous (cont.) Systems Dynamical v Instantaneous

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
– 1
– 1
Modelling,
identification and
Modelling,
identification and Proposition
analysis analysis
20
20
Control, Control, SS representation
validation If a MIMO system is instantaneous, then the following conditions are true validation
Fault diagnosis Fault diagnosis Consider a model of a system expressed in SS form
! The order of the derivatives of yi is ni = 0, for all i = 1, . . . , p
Classification Classification # % &
! The order of the derivatives of ui is mj ,i = 0, for all j = 1, . . . , p and
FC 2,
FC 2,
ẋ(t) = f x(t), u(t), t
i = 1, . . . , r % &
Input-output Input-output y(t) = g x(t), u(t), t
– 2
– 2
representation
The IO relation can be expressed as a system of p algebraic equations representation
A necessary and sufficient condition for a system to be instantaneous is
The model The model
that the SS model is zero-order (i.e., there exists no state vector)
G
G
Examples Examples
System
If any of the IO relations is a differential equation, the system is dynamical System
% &
properties properties ! y(t) = g u(t), t
U
U
A
A
"
delay delay
Systems Dynamical v Instantaneous Systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
Example
8
– 1
– 1
Modelling,
identification and
Consider the two serially arranged resistors Modelling,
identification and
analysis analysis
20
20
Control,
optimisation and
v (t) = (R1 + R2 )i(t) = Ri(t) Control,
optimisation and
Fault diagnosis The system is instantaneous Fault diagnosis
Classification Classification Linear v Nonlinear
FC 2,
FC 2,
Properties
Input-output
representation
The IO representation corresponds to the SS output transformation Input-output
representation
– 2
– 2
The model G R1 The model
G
Examples v (t) = (R1 + R2 )i(t) = Ri(t) Examples
System System
R2 1
properties v(t) ! y(t) = u(t) properties
U
U
Dynamical v
Instantaneous i(t) R Dynamical v
Instantaneous
A
A
With v without
The order of the system is zero (no device to store energy) With v without
delay delay
"
Systems Linear v Nonlinear Systems Linear v Non-linear (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
General concepts
Definition General concepts
– 1
– 1
identification and
analysis
Linearity identification and
analysis Proposition
20
20
Control, Control,
optimisation and
validation
A system is said to be linear if it obeys the superposition principle optimisation and
validation
IO representation - SISO
A system is said nonlinear, otherwise A necessary and sufficient condition for a SISO system to be linear is that
the IO representation is expressed by a linear differential equation
FC 2,
FC 2,
Representation " Representation
representation representation a0 (t)y(t) + a1 (t)ẏ(t) + · · · + an (t)y (n) (t)
– 2
– 2
State-space
Superposition principle State-space
= b0 (t)u(t) + b1 (t)u̇(t) + · · · + bm (t)u (m) (t)

The model The model
Consider some system
G
G
Examples Examples
System
• Let the system response to cause c1 be equal to effect e1 System
properties properties "
• Let the system response to cause c2 be equal to effect e2
U
U
The system response to cause (αc1 + βc2 ) equals effect (αe1 + βe2 ) The coefficients of the IO representation are, in general, time dependent
A
A
• (whatever the constants α and β)
delay delay
Systems Linear v Non-linear (cont.) Systems Linear v non-linear (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
Modelling,
identification and
Linear differential equation Modelling,
identification and
analysis analysis
20
20
Control,
optimisation and
Consider the differential equation Control,
optimisation and
% &
Fault diagnosis h y(t), ẏ(t), . . . , y (n) (t), u(t), u̇(t), . . . , u (m) (t), t = 0 Fault diagnosis
Consider a MIMO system in IO representation
FC 2,
FC 2,
Representation Representation Each function hi , i = 1, . . . , p, must be a linear combination of the i-th com-
The equation is linear if and only if the function h is a linear combination
Input-output Input-output ponent of the output and its ni derivatives, and the input and its derivatives
representation of the output and its derivatives, and of the input and its derivatives representation
– 2
– 2
representation representation ! The condition is necessary and sufficient
(n)
The model Gα0 (t)y(t) + α1 (t)ẏ(t) + · · · + αn (t)y (t) The model
G
Examples Examples
+ β0 (t)u(t) + β1 (t)u̇(t) + · · · + βm (t)u (m) (t) = 0
System System
U
U
Linear v Nonlinear ! A zero-sum weighted sum of inputs, outputs, and derivatives Linear v Nonlinear
A
A
delay delay
Systems Linear v non-linear (cont.) Systems Linear v non-linear (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
Proposition
– 1
– 1
analysis analysis
20
20
Control,
SS representation Control,
! "
validation validation ! A(t) = ai,j (t) ∈ Rn×n
Fault diagnosis
A necessary and sufficient condition for a system to be linear is that state Fault diagnosis
# ! "
equation and output transformation in the SS model are linear equations ẋ(t) = A(t)x(t) + B(t)u(t) ! B(t) = bi,j (t) ∈ Rn×r
Classification Classification !
y(t) = C(t)x(t) + D(t)u(t) ! "
! C(t) = ci,j (t) ∈ Rp×n
FC 2,
FC 2,
! "
Input-output
representation ⎧
Input-output
representation ! D(t) = di,j (t) ∈ Rp×r
– 2
– 2
State-space ⎪
⎪ẋ1 (t) = a1,1 (t)x1 (t) + · · · + a1,n (t)xn (t) + b1,1 u1 (t) + · · · + b1,r (t)ur (t) State-space
representation ⎪
⎪ representation
The model
⎪
⎪
⎪ ··· The model
⎪
⎨ (t) = an,1 (t)x1 (t) + · · · + an,n (t)xn (t) + bn,1 u1 (t) + · · · + bn,r (t)ur (t)
ẋ
⎪ Coefficient matrices A, B, C and D are, in general, time dependent
G
G
Examples ⎪ n Examples
System System
properties ⎪
⎪ properties
"
y1 (t) = c1,1 (t)x1 (t) + · · · + c1,n (t)xn (t) + d1,1 u1 (t) + · · · + d1,r (t)ur (t)
U
U
⎪
⎪
Dynamical v ⎪
⎪ Dynamical v
Instantaneous ⎪
⎪ ··· Instantaneous
Linear v Nonlinear
⎪
⎪ Linear v Nonlinear
⎩
A
A
Stationary v yp (t) = cp,1 (t)x1 (t) + · · · + cp,n (t)xn (t) + dp,1 u1 (t) + · · · + dp,r (t)ur (t) Stationary v
delay delay
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
Example
– 1
– 1
analysis analysis
20
20
Control, Control,
optimisation and Consider a system consisting of two cylindric tanks, both of base B [m2 ] optimisation and
validation validation The SS representation of the system, from mass conservation
Fault diagnosis • Output flow-rate from tank 1 is the input flow-rate to tank 2, q2 (t) Fault diagnosis
⎧ ⎧
⎪ ⎪ dV1 (t)
⎪ ⎪ = q1 (t) − q2 (t)
FC 2,
FC 2,
⎪ ⎨
Representation
First tank Representation ⎪
⎨!
⎪ dt
Input-output Input-output ⎪ dV 2 (t)
representation q1 q2 • Input flow-rate q1 [m3 s−1 ] representation ! ⎪
⎩ = q2 (t) − q3 (t) = q2 (t) − kh2 (t)
– 2
– 2
⎪ 8 dt
⎪
1
⎩! y(t) = x (t) − 1 x (t)
representation representation ⎪
⎪
The model The model
1 2
• h1 is the liquid level [m] B B
G
G
Examples Examples
d
System h1 System
properties h2 Second tank properties
U
U
Dynamical v
q3=k⋅h2 Dynamical v
Instantaneous • Input flow-rate q2 [m3 s−1 ] Instantaneous
A
A
non-stationary
(a) non-stationary
Proper v improper
• h2 is the liquid level [m] Proper v improper
delay delay
Systems Examples (cont.) Systems Linear v non-linear (cont.)

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
General concepts We obtained, General concepts Example
– 1
– 1
identification and
analysis
identification and
analysis
Counter-intuition
20
20
⎧
Control,
optimisation and
⎪
⎪ ẋ1 (t) = u1 (t) − u2 (t) Control,
optimisation and Consider the system described by the IO model
⎪
⎪
validation ⎨ −k validation
Fault diagnosis
! ẋ2 (t) = x2 (t) + u2 (t) Fault diagnosis
⎪ B y(t) = u(t) + 1
⎪
⎪ 1 1
FC 2,
FC 2,
⎪
⎩ẏ(t)
Representation = x1 (t) − x2 (t) Representation The system violates the superposition principle
Input-output
B B Input-output
representation representation • It is thus nonlinear
– 2
– 2
State-space
representation
We have, State-space
representation
The model ⎡ ⎤ The model
0 0 9 : Consider two constant inputs
G
G
Examples 1 −1 Examples
A(t) = ⎣ − k⎦ , B(t) = • u1 (t) = 1
System 0 0 1 System
properties B properties
• u2 (t) = 2
U
U
Dynamical v 9 : Dynamical v
Instantaneous 1 −1 % & Instantaneous
Linear v Nonlinear
C(t) = , D(t) = 0 0 Linear v Nonlinear
B B We can calculate the outputs
A
A
Proper v improper " Proper v improper
! y1 (t) = u1 (t) + 1 = 2
With v without
delay
With v without
delay
! y2 (t) = u2 (t) + 1 = 3
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2 Example
8
General concepts General concepts Counter-intuition
Consider the combined input
– 1
– 1
identification and
analysis
identification and
analysis
Consider the system described by the IO model
20
20
Control, u3 (t) = u1 (t) + u2 (t) = 3 Control, √
optimisation and
validation
optimisation and
validation
ẏ(t) + y(t) = t − 1u(t)
Classification
The resulting output Classification
y3 (t) = u3 (t) + 1 = 4 The system is linear
FC 2,
FC 2,
Consider the IO representation for a linear SISO system
– 2
– 2
representation We thus have, representation
The model y3 (t) = 4 ̸
= y1 (t) + y2 (t) = 5 The model ✭✭
✭(n) (t)
a0 (t)y(t) + a1 (t)ẏ(t) + ✭
· · ·✭
+✭an ✭
(t)y
G
G
Examples Examples
✭
System System ✭✭✭✭ (m)
properties The IO representation is a nonlinear algebraic equation properties = b0 (t)u(t) + b1 (t)u̇(t)
✭✭+✭· · · + bm (t)u (t)
✭✭
U
U
✭✭
Instantaneous • (Blame the +1 on the RHS for nonlinearity) Instantaneous ! a0 (t) = 1
A
A
Stationary v Stationary v ! a1 (t) = 1
non-stationary " non-stationary √
Proper v improper Proper v improper ! b0 (t) = t − 1
delay delay √
System ẏ(t) + y(t) = t − 1u(t) is thus linear
"
Systems Systems Stationary and non-stationary systems

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
Definition
– 1
– 1
analysis analysis
Stationarity
20
20
Control, Control,
validation validation A system is said to be stationary (or time-invariant), if it obeys the trans-
Classification Stationary v non-stationary Classification

lation principle
FC 2,
FC 2,
Representation Representation A system is said to be non-stationary (or time-varying), otherwise
representation representation "
– 2
– 2
The model The model
Translation principle
G
G
Examples Examples
System System Consider some system

U
U
Dynamical v
Instantaneous
Dynamical v
Instantaneous
Let the system response to a cause c1 (t) be equal to an effect e1 (t)
A
A
non-stationary non-stationary System response to cause c2 (t) = c1 (t − T ) equals effect e2 (t) = e(t − T )
delay delay
Systems Stationary and non-stationary systems (cont.) Systems Stationary and non-stationary systems (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 Let the same cause be applied to system S at 2 different times 2018.2
! t=0
8
! t=T
– 1
– 1
analysis analysis
20
20
Control, Control,
validation
c(t) c(t-T) validation
Fault diagnosis Fault diagnosis In nature, no system is stationary
Classification T Classification
Yet, there exists a wide range of variations that can be neglected
FC 2,
FC 2,
Input-output
t t Input-output • Over large time intervals
– 2
– 2
Over such intervals, the systems can be considered as stationary
The model The model
e(t) e(t-T)
G
G
Examples Examples
System T System
U
U
Linear v Nonlinear
t t Linear v Nonlinear
A
A
Proper v improper The resulting effects are analogous Proper v improper
delay • Shifted by time interval T delay
"
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
General concepts
Proposition General concepts
– 1
– 1
identification and
analysis IO representation
identification and
analysis
Consider a linear SISO system
20
20
Control, Control,
validation A necessary and sufficient condition for a system to be stationary validation a0 (t)y(t) + a1 (t)ẏ(t) + · · · + an (t)y (n) (t)
= b0 (t)u(t) + b1 (t)u̇(t) + · · · + bm (t)u (m) (t)
! The IO representation must not explicitly depend on time
FC 2,
FC 2,
Input-output
representation
" Input-output
representation
The model becomes a linear differential equation
– 2
– 2
The model Consider the SISO system The model ! a0 y(t) + a1 ẏ(t) + · · · + an y (n) (t)
G
G
Examples % &
h y(t), ẏ(t), . . . , y (n) (t), u(t), u̇(t), . . . , u (m) (t), t = 0
Examples
= b0 u(t) + b1 u̇(t) + · · · + bm u (m) (t)
System System
U
U
Instantaneous Then, the stationary model becomes Instantaneous
Linear v Nonlinear Linear v Nonlinear The coefficients are constant
A
A
% &
Stationary v
non-stationary
! h y(t), ẏ(t), . . . , y (n) (t), u(t), u̇(t), . . . , u (m) (t) = 0 Stationary v
non-stationary
delay delay
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
Theorem
8
SS representation
– 1
– 1
analysis
A necessary and sufficient condition for a system to be stationary analysis
Consider a linear system
20
20
Control, Control,
optimisation and optimisation and #
ẋ(t) = A(t)x(t) + B(t)u(t)
Fault diagnosis ! The SS representation must not explicitly depend on time Fault diagnosis
y(t) = C(t)x(t) + D(t)u(t)
• (Both state equation and output transformation)
FC 2,
FC 2,
Input-output
representation
" Input-output
representation The model becomes
– 2
– 2
State-space State-space #
Consider the system ẋ(t) = Ax(t) + Bu(t)
The model The model
!
y(t) = Cx(t) + Du(t)
G
G
Examples # % & Examples
System
ẋ(t) = f x(t), u(t), t System
% &
properties y(t) = g x(t), u(t), t properties
U
U
Dynamical v
Instantaneous
Dynamical v
Instantaneous
The (elements of the) coefficient matrices A, B, C and D are constant
Linear v Nonlinear Then, the stationary model becomes Linear v Nonlinear
A
A
non-stationary # % & non-stationary
Proper v improper ẋ(t) = f x(t), u(t) Proper v improper
With v without
! % & With v without
delay y(t) = g x(t), u(t) delay
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
8
The same input is applied with one (1) time-unit delay
– 1
– 1
identification and
analysis Example identification and
analysis
20
20
Control, Control,
optimisation and Consider the instantaneous and linear system optimisation and u(t) u(t-1)
Fault diagnosis
y(t) = tu(t)
Fault diagnosis
1 1
FC 2,
FC 2,
Representation The system is non-stationary Representation
Input-output Input-output 0 1 t 0 1 2 t
(a) (c)
– 2
– 2
State-space
representation We can show this by using the translation principle
State-space
representation
y(t-1)
2
The model The model
y(t)
Consider the input
G
G
Examples # Examples
1, t ∈ [0, 1] 1 1
System System
properties u(t) = properties
0, elsewhere
U
U
Instantaneous Instantaneous 0 1 t 0 1 2 t
A
A
Stationary v If the same input is applied with a delay, the output is not simply shifted Stationary v
(b) (d)
With v without With v without "
delay delay
Systems Systems Proper and improper systems
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
identification and
analysis
identification and
analysis
Definition
20
20
Control, Control, Appropriateness
Fault diagnosis Fault diagnosis A system is said to be proper, if it obeys the causality principle
Classification Proper v improper Classification
The system is said to be improper, otherwise
FC 2,
FC 2,
"
– 2
– 2
The model G The model Causality principle
G
Examples Examples
The effect does not precede the generating cause
System System
U
U
In nature, all systems are (obviously?) proper
A
A
Stationary v Stationary v • Only the model can be improper
delay delay
Systems Proper v improper (cont.) Systems Proper and improper systems (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2 The result can be extended to MIMO systems
8
8
General concepts General concepts ⎧ + (m1,1 )
(n ) (m ,r ) ,
⎪h1 y1 (t ), ẏ1 (t ), . . . , y1 1 (t ), u1 (t ), u̇1 (t ), . . . , u1
⎪ (t ), . . . , ur (t ), . . . , ur 1 (t ), t
– 1
– 1
Modelling, Modelling, ⎪
⎪
⎪
⎪ & '( ) & '( ) & '( )
⎪
analysis Proposition analysis ⎪
20
20
⎪
⎪
Control, Control,
⎪
⎪ =0
⎪
⎪
validation
IO representation - SISO validation
⎪
⎪
⎪
+ (m )
⎪h y (t ), ẏ (t ), . . . , y (n2 ) (t ), u (t ), u̇ (t ), . . . , u 2,1 (t ), . . . , u (t ), . . . , u (m2 ,r ) (t ), t ,
⎪
⎪ 2 2 2 2 1 1 1 r r
Fault diagnosis Fault diagnosis ⎪
⎪ & '( ) & '( ) & '( )
⎪
Classification
A necessary and sufficient condition for a SISO system to be proper Classification
⎪
⎪
⎪
⎨ output 2 input 1 input r
=0
FC 2,
FC 2,
Representation Representation ⎪
⎪
⎪ .
⎪
Input-output ! The order of derivation of the output (n) is equal to or larger than Input-output
⎪
⎪
⎪
⎪
.
.
the order of derivation of the input (m) ⎪
⎪
– 2
– 2
⎪
⎪
⎪
⎪
representation % & representation
⎪h + y (t ), ẏ (t ), . . . , y (np ) (t ), u (t ), u̇ (t ), . . . , u (mp,1 ) (t ), . . . , u (t ), . . . , u (mp ,r ) (t ), t ,
⎪
The model h y(t), ẏ(t), . . . , y (n) (t), u(t), u̇(t), . . . , u (m) (t), t = 0, with n ≥ m The model
⎪
⎪
⎪
⎪
⎪
p
&
p p
'(
p
) &
1 1
'(
1
) &
r
'(
r
)
⎪
G
G
⎪
Examples Examples ⎪
⎪
⎪ output p input 1 input r
⎩
System System
=0
U
U
Dynamical v A system where n > m is said to be strictly proper Dynamical v
None of the equations must include derivatives of the input variables whose
" order is larger than the derivation order of corresponding output variables
A
A
ni ≥ max mi,j , for all i = 1, . . . , p
With v without With v without j =1,...,r
delay delay
A system is strictly proper if the inequality is strictly true, for all i = 1, . . . , p

Systems Proper and improper systems (cont.) Systems Proper and improper systems (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
Proposition
– 1
– 1
analysis SS representation analysis
20
20
Control, Control,
validation Consider a system described by a SS model validation
#
Classification
%
ẋ(t) = f x(t), u(t), t
&
Classification The SS model of a linear, stationary and strictly proper system
FC 2,
FC 2,
% & #
Representation y(t) = g x(t), u(t), t Representation
ẋ(t) = Ax(t) + Bu(t)
representation representation !
y(t) = Cx(t)
– 2
– 2
State-space Such a system/model is always proper State-space
G
Examples A strictly proper system has an output transformation independent on u(t) Examples
System # System
properties
% & properties
ẋ(t) = f x(t), u(t), t
U
U
Dynamical v ! % & Dynamical v
Instantaneous
y(t) = g x(t), t Instantaneous
A
A
With v without
" With v without
delay delay
Systems Systems Systems with and without delay

UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
Definition
8
8
General concepts General concepts Finite time delay
– 1
– 1
identification and
analysis
identification and
analysis
A finite delay is a system whose output y(t) at time t is equal to the input
20
20
Control, Control, u(t − T ) at time t − T
"
Classification With v without delay Classification
FC 2,
FC 2,
u(t) y(t)=u(t-T)
representation representation S
– 2
– 2
The model The model elemento
G
G
Examples Examples
! T ∈ (0, +∞) is called the time delay
System System
U
U
Linear v Nonlinear Linear v Nonlinear u(t) y(t)=u(t-T)
A
A
T
delay delay
t t
Systems Systems with and without delay (cont.) Systems Systems with and without delay (cont.)
UFC/DC UFC/DC
CK0255|TIP8244 CK0255|TIP8244
2018.2 2018.2
8
– 1
– 1
analysis analysis
20
20
Control, Control,
validation Consider the algebraic equation describing a finite delay element validation
Proposition
Classification y(t) = u(t − T ), with T ∈ R+ Classification IO and SS representation
FC 2,
FC 2,
Input-output Input-output A necessary and sufficient condition for a system to be without a time delay
• Such a system is not instantaneous
– 2
– 2
• The system is dynamic ! All the signals in the model (IO or SS) must share the same argument
G
Examples Examples
The output at time t depends on the previous values of the input "
System System
U
U
A
A
delay delay
Systems Systems with and without delay (cont.)

UFC/DC
CK0255|TIP8244 Example
2018.2
Consider a system described by the IO model
8
General concepts
4ẏ(t) + 2y(t) = u(t − T )
– 1
Modelling,
identification and
analysis
20
Control,
optimisation and
validation
The system has delay elements
Fault diagnosis • There are signals that are dependent on t
Classification
• There are signals that are dependent on t − T
FC 2,
Representation
Input-output
representation
"
– 2
State-space
representation
The model
Example
G
Examples Consider a system described by the SS model

System #
properties
ẋ (t) = x (t − T ) + u(t)
U
Dynamical v
Instantaneous y(t) = 7x (t)
Linear v Nonlinear
A
Stationary v
non-stationary
Proper v improper The system/model has delay elements
With v without
delay • There are signals that are dependent on t
• There are signals that are dependent on t − T
"

01 Systems

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

01 Systems

Uploaded by

Copyright:

Available Formats

Systems Systems Systems

Systems Systems (cont.) Systems

Classification Classification General concepts

Systems Modelling Systems Modelling (cont.)

Classification Case A) We have a knowledge on the type/number of component devices Classification R1

Systems Identification (cont.) Systems Analysis

Systems Control Systems Control (cont.)

Classification We need to explicitly formulate what we mean by ‘desired behaviour’ Classification

Systems Validation Systems Validation (cont.)

Classification Fault diagnosis Classification

Systems Fault diagnosis (cont.) Systems

System A doctor, once identified the pathology, prescribes a therapy System

! Discrete time-evolving systems

System By representation System • The time variable takes discrete values

Systems Systems Time-evolving systems (cont.)

Classification Time-evolving systems Classification

Systems Time-evolving systems(cont.) Systems

System The diﬀerence equation is a relation between discrete-time functions System

Systems Discrete-event systems (cont.) Systems Discrete-event systems (cont.)

Systems Hybrid systems Systems Hybrid systems (cont.)

The cabin can be represented using a time-evolving model

Systems Systems Representation

Classification Classification Input-output representation

Systems Input-output representation Systems Input-output representation (cont.)

System System • The changes depend on the specific system (car)

Systems Input-output representation (cont.) Systems Input-output representation (cont.)

System • h1 is the liquid level [m] System Output, d = h1 − h2

Classification State-space representation Classification

Systems State-space description(cont.) Systems State-space representation (cont.)

• Over the entire interval [to , t]

Systems State-space representation (cont.) Systems State-space representation (cont.)

Systems State-space representation (cont.) Systems

System • Output flow-rate q3 [m3 s−1 ] System

The entire (two-tank) system has order 2

Systems Systems Input-output model

Classification Input-output model Classification

System System h is a multi-parametric function that depends on the system

Systems Systems State-space model

Systems State-space model (cont.) Systems State-space model (cont.)

Systems State-space model (cont.) Systems State-space model (cont.)

Classification Examples Classification d

Systems Examples (cont.) Systems Examples (cont.)

Systems Examples (cont.) Systems Examples (cont.)

Systems Systems System properties

Systems Dynamical v Instantaneous (cont.) Systems Dynamical v Instantaneous (cont.)

! Finite time delay system

Systems Dynamical v Instantaneous (cont.) Systems Dynamical v Instantaneous

Classification Classification Linear v Nonlinear

Systems Linear v Nonlinear Systems Linear v Non-linear (cont.)

= b0 (t)u(t) + b1 (t)u̇(t) + · · · + bm (t)u (m) (t)

Systems Linear v non-linear (cont.) Systems Linear v non-linear (cont.)

Systems Examples (cont.) Systems Linear v non-linear (cont.)

Systems Systems Stationary and non-stationary systems

Classification Stationary v non-stationary Classification

System System Consider some system

A system is strictly proper if the inequality is strictly true, for all i = 1, . . . , p

Systems Systems Systems with and without delay

Systems Systems with and without delay (cont.)

Examples Consider a system described by the SS model

You might also like