Page 14 - Study At Centrale Supélec 2021

Mechanical &

Aerospace Engineering

method» to find an optimum between convergence speed (of the optimization
ENGINEERING CHALLENGE TERM algorithm) and the quality of the minimization of execution time and energy
footprint (of the targeted HPC application).
CORE COURSE : OPTIMISATION (2,5 ECTS)
(MANDATORY) ACADEMIC TERM
This course will explore various fundamental notions of both continuous and
discrete optimisation. The following topics will be addressed and implemented :
formulation of optimisation problems, existence conditions for global and local ELECTIVE SERIES 2.4 (2,5 ECTS)
minimizers, convexity, duality, Lagrange multipliers, first-order methods, linear
programming, integer linear programming, branch and bound approaches, • ARTIFICIAL INTELLIGENCE
preliminary stochastic optimisation concepts. What do web-based information retrieval, personal assistant development,
autonomous driving or automatic planning have in common ? These are all
CONCENTRATION (7,5 ECTS) complex real-world problems that artificial intelligence (AI) aims at solving
by addressing them in rigorous methods. In this course, you will study the
• HIGH PERFORMANCE SIMULATION FOR FOOTPRINT fundamental principles that guide these applications and implement some of
these systems. Specific topics include automated learning, research, gaming,
REDUCTION Markov decision processes, constraint satisfaction, graphic models and logic.
The main objective of the course is to provide you with the framework to
PARIS SACLAY and Algorithms, and Optimisation Methods philosophical aspects of AI will also be discussed.
Concentration course : Parallel Computing Methods
address new AI problems that you may encounter in the future. The ethical and
ELECTIVE SERIES 2.5 (2,5 ECTS)
Simulation today is at the heart of many design and optimisation approaches,
to reduce the impact of the products created: reducing the carbon footprint,
• MAINTENANCE AND INDUSTRY 4.0
the sound footprint, etc. Such problems are often complex systems, whose
simulation requires specific skills in high performance and large-scale
simulations.
involved in the implementation of a predictive maintenance approach.
In this course, students will learn to develop models and simulations whichever This course provides a solid culture on the concepts, methods and tools
The objective of this course is to give future decision-makers the necessary
the size of the problem, without sacrificing the accuracy of computations. culture to design, model and recommend predictive maintenance strategies.
Emphasis is placed on data-driven approaches and probabilistic or statistical
Concentration projects (SELECT ONE) models that apply to any industrial system. This background should allow
effective interaction with engineers « business » very close to applications and
Project (1) : Shape optimisation and drag reduction in « data scientist » in charge of data processing.
SEMESTER BY COURSEWORK
aeronautics
In partnership with ONERA ELECTIVE SERIES 2.6 (2,5 ECTS) (SELECT ONE)
Air traffic is steadily increasing each year, to the point that without improvements
in aircraft performance in terms of energy consumption, the share of air • MULTIPHYSICS COUPLING SIMULATION WITH FINITE
transport in greenhouse gas emissions may become unsustainable in the ELEMENT METHODS
future. The aim of this class is to give theoretical and applied insights on multiphysics
Aircraft consumption can be decreased by either increasing the engine coupling simulations such as: thermomechanical, piezoelectric, vibro-acoustic,
efficiency or by improving the aerodynamic design of the aircraft, e.g. reducing
the aircraft weight. Computational tools have been widely used in aeronautics magneto-mechanic.
for a long time to help design and optimise systems. For example, the shape
of the wing can be improved to reduce its drag or lift, or its inner structure • NUCLEAR ENGINEERING
can be lightened. This course will present the operating principles of nuclear reactors and
describe in detail all the stages of the fuel cycle. Students will be able to
Project (2) : Low cost optimisation of acoustic wave appreciate the advantages and drawbacks of this low carbon source of
propagation code performance energy, from a technical perspective.
In partnership with INTEL
Regardless of the type of application running on parallel machines in an HPC
(high performance computing) environment, and regardless of their level of INTENSIVE SEMINAR
efficiency (in terms of performance or energy footprint), we can easily see
that the impact of input parameters is generally not negligible. It is therefore
necessary to choose optimization methods that do not consume too many (SELECT ONE)
experiments, and then to optimize their use to reduce the footprint of the
targeted HPC code, without this pre-study consuming too many computing HUMANITIES AND SOCIAL SCIENCES (2 ECTS)
resources ! This amounts to «looking for the least expensive optimization Students are offered a choice of Humanities and Social Sciences courses in
14