Elements of

Complex System Engineering

Antoine B. Rauzy
Department of Mechanical and Production Engineering (IMP)
Norwegian Science and Technology University (NTNU)
and
Chaire Blériot-Fabre
Centrale-Supélec, SAFRAN

Norwegian University of Science and Technology

LECTURE 1. PART 1
GENERAL INTRODUCTION

Notions:
• Systems
• Models of Systems
• Modeling Formalisms

Norwegian University of Science and Technology 2

LECTURE 1. PART 1. SECTION 1.
THE SCIENCE OF MODELS

Norwegian University of Science and Technology 3

 What does “system” mean?

• We call system “a combination of interacting elements that are organized

so to achieve one or several established objectives” (INCOSE(1) 2007).
• This course deals more specifically with intentionally designed technical
and sociotechnical systems, i.e. de systems designed by means of an
engineering process:
– Industrial products ;
– Organizations that design, produce, operate and/or regulate these products;
– Processes by which these organizations design, produce, operate and/or
regulate these products;
• We shall not consider:
– Natural systems (molecules, cells, volcanos, rivers…)
– Human systems not resulting of an engineering process (cities, internet…)

(1) International Council on System Engineering, http://www.incose.org/

Norwegian University of Science and Technology 4

 Models are everywhere

• The systems designed by industry are more and more complex. Not only
these products are more and more complex but also the processes by
which they are designed/produced/operated/decommissioned and
organizations that implement these processes are.

• To face this complexity, the different engineering disciplines (mechanics,

thermic, electric and electronic, software, architecture…) virtualize their
contents to a large extent, i.e. they are designing models. We entered the
era of:

Model-Based Systems Engineering

• Each system comes with dozens of models. More and more of these
models are embedded into systems and used for their operation.

Norwegian University of Science and Technology

 Systems in Silico

Norwegian University of Science and Technology 6

 Science and engineering of models

Models must be taken seriously and considered as first class citizens. This raises a
number of challenges:
• Better understand the nature of models and their roles in industrial processes.
• Develop the “Art of Modeling”(*) in each and every engineering discipline.
• Manage models throughout the life-cycle of systems.
• Design tools and methods to support the integration of engineering
disciplines/processes through the integration of models they produce.
• …
The emerging science of complex systems
is the science of models

(*) In reference to Knuth’s famous series of books about “The Art of Programming”

Norwegian University of Science and Technology

LECTURE 1. PART 1. SECTION 2.
WHAT IS A MODEL?

Norwegian University of Science and Technology 8

 What is a model?

Mathematical Model Cognitive Model

Star

Mockup Graphical Representation

Code

All these “things” are models in some way

Norwegian University of Science and Technology 9

 Models in systems engineering

mind & paper

models

Cognitive Model Mathematical Model

computerized
models

Code Graphical Representation(s)

Models are working tools, not (platonic) ideals the system should comply to.

Norwegian University of Science and Technology 10

 Modeling languages
A computerized model is eventually nothing but a sequence of symbols that obeys
certain rules. To design a model, we need a modeling language (would it be purely
graphical), just as to design a program, we need a programming language.
Purely graphical languages reach quickly their limits when models get complex.
Textual modeling languages are then necessary. Parts of a large model can be
represented graphically, but they should not be confused with the model itself.

domain State {Idle, Active, OutOfService}

automaton BankATM
State state(init = Idle);
event service, fixed, in(card), cancel, done;
transition
service: state==Idle -> state:=OutOfService;
fixed: state==OutOfService -> state:=Idle;
in(card): state==Idle -> state:=Active;
cancel: state==Active -> state:=Idle;
done: state==Active -> state:=Idle;
end

Norwegian University of Science and Technology 11

 Specific purposes, specific models
Fluid mechanics

Multiphysics simulation

Safety analyses
overflow
The content and the level of abstraction
of a model depends on what is to be OK KO
observed, i.e. on the virtual
experiments to be performed on that Insurance
model.

Norwegian University of Science and Technology 12

 All models are false. Some are useful.

George Edward Pelham Box

(1919-2013)

A model is an abstraction and a point of

view on the system at hand.
It is designed with a particular purpose,
i.e. to realize particular virtual
experiments, to learn lessons from
these experiments and to make
decision about the system

Norwegian University of Science and Technology

 Taxonomy of models
Models are designed at different level of abstraction, for different purposes and in
different modeling formalisms.
Models to calculate
performance indicators

Models to communicate
amongst stakeholders

Models to generate artefacts

(via code generation) or
physical components (via
additive manufacturing)
Informal models, even thought they Formal models, that essentially encode
are written in standardized notations, and organize (a given type of)
sometimes called semi-formal and mathematical equations
computerized
Norwegian University of Science and Technology 14
 Discrete Mathematics
The incredible successes of sciences, the fantastic development of technologies rely for a
great part on mathematical analysis, i.e. mathematics of the continuum. However, when
considering a system (a car, an airplane, an engine…) at system level, it turns out that
relevant conceptual tools come rather from discrete mathematics.

To reason about a system, it is convenient to see it as a finite hierarchy of components

and to consider that each component can be in finitely many states. As a consequence,
the number of possible states of the system itself is also finite, although it can be
extremely large. The modeling frameworks we shall study in this course are discrete.

Norwegian University of Science and Technology 15

 Complexity of calculations

The value of a model stands entirely in the lessons one can draw on the
system from the virtual experiments made on the model. These virtual
experiments have a computational cost. There are cases in which this cost is
actually the decisive factor.
The rules of chess are fully known and fully
deterministic. Moreover, a game can end only in three
ways: white wins, black wins or draw.
It is easy to design a model of what is the initial
position, what is a legal move, and what is a final
position.
However, determining whether there is a winning
strategy for white, i.e. a way of playing white that wins
against any defense of black, is computationally so
expensive that no virtual experiment can answer the
question.

Norwegian University of Science and Technology 16

LECTURE 1. PART 1. SECTION 3.
OBJECTIVES AND ORGANIZATION OF THE COURSE

Norwegian University of Science and Technology 17

 Knowledge base of engineers
As future engineers and scientists, we should acquire the scientific knowledge and
engineering know-how making you able:
• To select modeling framework and tools to solve problems at stake, meeting
needs and constraints of your work environment;
• To design models;
• To understand the representativeness and the limits of these models;
• To perform virtual experiments with these models, defining appropriate
experimental protocols;
• To interpret results of these experiments and to check them against the real
world;
• To tune models accordingly;
• To manage models and virtual experiments;
• To integrate your models and virtual experiments with those of other
teams/disciplines/organizations.

Norwegian University of Science and Technology 18

 Organization of the course
We shall try to follow the following scheme for lectures and the tutorial.

Observe • Description of a use case

• Informal analysis of the use case
lectures

Formalize • Presentation of a modeling framework

• Study of the use case in this modeling framework

Model • Modeling of similar use cases in the framework

tutorials

Experiment • Design of virtual experiments

• Report on the result of the virtual experiments

Norwegian University of Science and Technology 19

 Domain Specific Languages

Modeling activities in this course (and the associated tutorials) rely on:
1. Small textual domain specific modeling languages;
2. Ad-hoc assessment tools for models written in these modeling
languages.

Assessment tools have been written in the Python scripting language.

You should download python and install it on your machine.
Please download and install the standard Python 3.5.x environment.
https://www.python.org/downloads

Norwegian University of Science and Technology 20

 Scripting Languages
Programs play a central role in all engineering processes (and more generally in
modern societies). Therefore, systems engineering has much to do with computer
science and software engineering. However, this relationship must be well
understood.
• Computer Science is the science of discrete structures (as opposed to
mathematical analysis). It provides many concepts (automata, languages,
computational complexity…) used in other sciences (including cognitive sciences)
and of course in Systems Engineering.
• Software Engineering is a high technology that aims at designing methods to
develop software. It is only marginally related to Systems Engineering.
• Scripting (with programming languages such as Perl, Python, Ruby) is a particular
type of programming. Scripting is mandatory to be efficient when performing
virtual experiments (e.g. to change formats of data, to chain tools…).

Mastering at least one scripting language is one of expected skills of new generations
of engineers and scientists.

Norwegian University of Science and Technology

 Content of the course
9 lectures + 9 tutorials + one written exam
Introduction
• Lecture 1, part 1: The science of models

Models to communicate
• Lecture 1, part 2 + tutorial 1: Business Process Model and Notation
• Lecture 2 + tutorial 2: System Architecture & Design Structure Matrices
• Lecture 3 + tutorial 3: System Architecture & Requirement Engineering

Models to calculate
• Lecture 4 + tutorial 4: Graphs
• Lecture 5 + tutorial 5: Automata
• Lecture 6 + tutorial 6: Constraint Solving
• Lecture 7 + tutorial 7: Discrete Event Simulation
• Lecture 8 + tutorial 8: Stochastic Simulation
• Lecture 9 + tutorial 9: Limits of Calculability

Norwegian University of Science and Technology 22

Louis Charles Joseph Blériot (1872 -1936) Henri Marie Léonce Fabre (1882 -1984)
is an airplane designer and one of the is a French engineer and pilot. He
pioneer pilot of French aviation. He has invented the seaplane in 1910.
been the first to cross the channel on July He graduated from Supélec.
the 25th onboard of the Blériot XI.
He graduated from Ecole Centrale de Paris

Norwegian University of Science and Technology 23