2013

Diego Marcos Gonzalez – Obstacle detection using multiple ultrasonic sensors

About senseFly Ltd., Ecublens

Supervisor: Mr. Adam Klaptocz

Description
The goal of this project is to integrate distance sensors into a small flying robot. The project will include several phases:
– state of the art in lightweight, low-power distance sensing techniques
– analysis of available sensors, including range, angle, sensitivity, frequency, etc.
– design, production and testing of electrical circuits for interfacing of the sensors
– design of code for the excitation and signal processing of the sensors
– integration of sensors within a flying platform

 

SenseFly drones generate an amazing 3D model of the Matterhorn

  

Emmanuel Froustey – Filtrage des courbures des surfaces et détection automatique de lignes caracteristiques

    About Dassault Aviation, Saint-Cloud, France

Supervisor: Dr. Boris Clemençon

Résumé
Dassault Aviation est un groupe industriel qui assure la conception et le développement, la production, la vente et le soutien d’avions militaires (Rafale, Mirage 2000, Atlantique, Alpha Jet), et d’avions d’affaires Falcon. Avec près de 7 500 avions civils et militaires livrés depuis près de 60 ans dans 75 pays ayant totalisé quelques 20 millions d’heures de vol, Dassault Aviation bénéficie d’un savoir-faire et d’une expérience reconnus dans ces domaines d’activité.

2 Stage
2.1 Sujet
Sur une surface paramétrée telle que les NURBS utilisées dans les modèles de CAO, les lignes de crête présentent un intérêt tout particulier pour le maillage surfacique. Ce stage a eu pour but de développer des méthodes numériques pour extraire automatiquement les lignes de crête des modèles de CAO industriels complexes, et filtrer les données obtenues de ces modèles pour obtenir des résultats exploitables dans les logiciels de maillage surfacique.

2.2 Algorithmes développés
Un premier critère a été défini pour caractériser les lignes de crête des surfaces paramétrées : ce sont les lignes le long desquelles l’écart entre les courbures principales de la surface est maximal. Pour extraire ces lignes, un algorithme de séparation des eaux a été implémenté. En parallèle, deux algorithmes de filtrage des données ont été utilisés : un filtre de Gauss pour lisser les données de courbures, et un filtrage par arbre des composantes pour limiter les lignes de crête obtenues aux lignes les plus significatives.

2.3 Résultats
La figure montre un véhicule de rentrée atmosphérique. Les lignes de crête sont visibles en vert. A noter la bonne détection du bord de l’attaque de l’aile et du winglet, ainsi que les deux lignes qui délimitent le winglet. La ligne sur le fuselage est moins significative, et peut être éliminée en ajustant les paramètres du filtre de l’arbre des composantes.


 

 

Federico H. Martinez Lopez – Quantitative Developer

About Global Financial Products AG, Lausanne

Supervisor: Dr. Samuel Quinodoz

Description
Implementation of a Monte Carlo engine on GPU for the pricing and risk management of financial derivatives.

Elena Queirolo – Fiabilité et optimisation des calculs obtenus par des formulations intégrales en propagations des ondes

   About ONERA, The French Aerospace Lab, Toulouse, France

Supervisor: Mr. Louis Holleaux

Description
The internship took place at ONERA, in the center of Toulouse, in a group
that worked on electromagnetic simulations. The code already implemented dealt
with the calculation of the current generated over a metal object by an incident
electromagnetic field. It used an integral formulation law and Raviart-Thomas finite
elements. The aim of the internship was to construct an a posteriori error estimate
of the reconstruction type. That means that, given an initial mesh, a uniformly
refined mesh was constructed and the solution was calculated on the refined mesh.
The solution was the projected on the initial mesh by an L2 projection on a finite
element space of higher order and the L2-norm of the difference between the solution
and its projection was an efficient and efficace estimator of the error. The test made,
indicated that the estimator worked as predicted. Once the estimator calculated,
we used it as the base for an adaptive algorithm, but the results were not as high
as expected. Even if the estimator succeded in finding the element with the largest
error in the L2 norm, the refinement of just those elements could not reach the
optimal convergence.

Numerical Methods for Trajectory Optimization

Andrea Di Blasio – Computational fluid dynamics and scientific visualization

     About FlowKit CFD Solutions, Lausanne, Switzerland

Supervisor: Dr. Jonas Latt

Description
Visualization of 3D content produced by CFD simulation with the renderer Blender.

FlowKit Ltd. is a start-up company which provides professional numerical
solutions for simulation of fluids and complex physics phenomena. The
flagship product of the company is an open-source CFD software, Palabos,
which covers a large range of physical applications. It is a software tool for
classical CFD, particle based models and complex-physical interactions. It is
based on a modern approach called the Lattice Boltzmann method. Palabos
is freely available, while FlowKit Ltd. provides consulting, training, support
and develops additional pieces of software if needed for a specific customer.

Among all the projects they are actually focused on, there is their aim at
building and providing to their customers a web-service for fluid simulations.
Trough this web-service the customer will be able to define his own problem,
specifying geometries and physical properties governing the phenomena, and
to obtain in reasonable time the numerical results desired. So far the numerical
results obtained by Palabos are visualized trough Paraview. Despite
Paraview is particularly good for visualizing scientific data, it does not provide
visualizations that look realistic in terms of material properties such as
color, light reflection/refraction, etc. Therefore in order to make their offer
more attractive, they proposed me to work during the internship on the integration
of a rendering engine within Palabos. Trough this the customer
can receive together with the numerical results also a render of the simulation
computed. The software used for the 3-D rendering was Blender, a free
and open-source 3-D computer graphics software product used for creating
animated films and visual effects.


Figure 1: A frame from the animation made for the dam break problem

Ivan Slijepcevic – Mechanical Engineering : Study on Extended Finite Element Method

Lien retour vers la page d'accueil   About Alstom Switzerland aG

Supervisor: Dr. Irwanto

Description

ALSTOM is a large multinational company working in power and transport sector. While
transport sector focuses on railway industry, power sector activities include the design,
manufacturing, services and supply of products and systems for the power generation
sector and industrial markets. The company employs over 85000 people worldwide.
My internship position was in the electromechanics group within ALSTOM Turbogen-
erator R&D. My task was to get familiar with the extended

Lidia Stepanova – Mathematical analysis of Multilevel Power Converters

 
Supervisor: Dr. Wim van der Merwe
 
Description:
 
During my internship at ABB Corporate Research Center in Baden, Switzerland I have
been working under supervision of Dr. Wim van der Merwe at the department of Power
Electronics in S7 group of Power Conversion Systems, which focuses on research and development of the power electronics converters, its control, analysis and optimization. My primary activities included theoretical analysis of dynamical systems of the circuits, as well as numerical simulation of mathematical models in MatLab and Python to verify the theory, and the design of special algorithms to quickly stabilize the perturbed systems.

There are two types of converters which were studied: flying-capacitor converter (FCC) and modular multilevel converter (MMC), which are both used in medium to high voltage power electronic systems, but have different circuit topologies and thus different operational characteristics. Particularly, the most resent research is dedicated to MMC converters, which were introduced at industry level in 2010 and have shown significant improvement in performance compare to previous converter configurations. The underlying physics of the converters can be described by a set of highly nonlinear differential equations, which compose a switching system with large frequencies