Liens vers les séminaires de 2021, 2020, 2019, 2018, 2017, 2016, 2015, 2014, 2013, 2012, 2011, 2010*Voir aussi la liste des « Réunions de Travail et d’Echange » du Département MIS.*

*Ci-dessous la liste des derniers séminaires :*

*01/12/2016, 15h00 : Roland Toth (Eindhoven University of Technology)*

ECL, H9, salle Acquitaine

* Developments in LPV state-space model identification. *

### 29/11/2016, 14h00 : Carlos Correia (Laboratoire d’Astrophysique de Marseille)

INSA St-Exupéry, salle M2B

* Adaptive optics in the Extremely Large Telescope’s era: new requirements, new concepts and new challenges. *

Modern astronomy relies on telescopes to overcome the barriers of the infinitely faint, distant and uncharted universe. The extraordinary progress foreseen with next generation ground-based facilities with 5x greater angular resolution and 600x the sensitivity over today’s telescopes will revolutionize the landscape in observational visible and near-infrared astronomy. Such advances rely on disruptive adaptive-optics concepts to counter the nefarious effects of earth’s turbulence and on pioneering new paradigms in data-reduction algorithms to reveal the finest details. Taking the case of HARMONI, the 1st-light integral-field-unit for the European Extremely Large Telescope, I will cover the main science cases, the current design and the R&D development program that we have at LAM-Marseille.

### 23/11/2016, 12h30 : Debarghya Ghosh (doctorant Ampère)

ECL, H9 (PhD Club)

* Optimal experiment design for local LPV identification * (en anglais).

In local LPV identification, a number of local LTI models of a Linear Parameter Varying (LPV) system are identified by performing identification experiments for fixed values of the scheduling variable. These local LTI models are then interpolated in order to yield a LPV model of the LPV system. In this talk, we are looking at ways to optimally choose the values of the scheduling variables at which the local identification experiments have to be performed as well as the excitation signals that have to be applied to the system during these local identification experiments. The objective of this optimal experiment design problem is to obtain a sufficiently accurate LPV model with the least injected power.

### 27/10/2016, 10h15 : Sjoerd Boersma (doctorant Delft)

ECL, H9 (PhD Club)

* Ferme éolioenne * (en anglais).

A wind farm is a group of wind turbines placed together. Advantages are, i.a., reduction of installation and maintenance costs. However, placing turbines together (typical spacing is around 7 rotor diameter from each other) will result in an interaction between the turbines via wakes of upwind turbines.

The most important wake characteristics are:

• Wind velocity deficit due to the turbine’s energy extraction.

• Increased turbulence intensity due to, i.a., the turbine blades rotation.

• Wake recovery, which is the phenomena where, i.a., the wind velocity recovers to the free stream velocity due to mixing.

• Wake expansion occurs with distance from the turbine and can be explained using the law of mass conservation and the assumption of flow incompressibility. It can be shown that a decrease in velocity means a proportional increase in the wake’s cross-sectional area.

• Properties change in time and position.

Wake behaviour is nonlinear and stochastic hence difficult to capture in a model and still under investigation. The wake of an upwind turbine can influence the performance of a downwind turbine. For example, power production and loading are important performance indicators which can be influenced negatively by a wake.

In current wind farm control, single turbine control strategies are pursued while not taking the losses due to wake interactions into account. The idea of wind farm control is to find desired control variables while taking wake interactions into account. The biggest challenge can be found in finding and using a suitable wake model.

During my thesis we have developed a dynamic wind farm model and applied it in an ensemble Kalman filter and MPC approach. In the next phase, these applications will be tested on high-fidelity models. Other questions are: 1) can uncertainty be included in the model and 2) can the dynamical wind farm model be used to design local dynamical controllers ensuring global performance.

### 27/10/2016, 14h00 : Richard Kern (doctorant TU München)

INSA St-Ex, MB1 (PhD Club)

Diaporama présenté

* Physical Modelling of a Long Pneumatic Transmission Line: Models of Successively Decreasing Complexity and their Experimental Validation *

There exists a significant number of models which describe the dynamics of pneumatic transmission lines. The models are based on different assumptions and, thereby, vary in the physical phenomena they incorporate. These assumptions made are not always stated clearly and the models are rarely validated with measurement data. The aim of this presentation is to present multiple distributed parameter models that, starting from a physical system description, successively decrease in complexity and finally result in a rather simple system representation. In this process, all assumptions are explicitly stated and justified from an engineering point of view. Data, both from simulation studies as well as from a pneumatic test bench, serve as a quantitative validation of these assumptions. Based on a detailed discussion of the different models, this contribution aims at facilitating the choice of an appropriate model for a given task where the effect of long pneumatic transmission lines cannot be neglected and a trade-off between accuracy and complexity is required.

### 15/09/2016, 12h30 : Jean-François Tregouet (Ampère)

INSA, Bât. Saint-Exupéry, salle MB1 (1er étage)

* A periodically time-varying approach for input-delay problems: Application to network controlled systems. *

This talk deals with robust stability and stabilization of linear discrete-time systems subject to uncertainties and network constraints. In network control systems, the control loop is closed over a network, which induces additional dynamics to the original control loop such as delays, sampling, quantization among many others. This talk focuses on networked induced delays due to unreliable network for which packet losses may occur.

An equivalent periodic-like representation of the resulting system is proposed. This allows first, to revisit existing results in this framework and second, to take model uncertainties into account by analyzing the closed-loop model by means of a recent method based on robust control for discrete-time time-varying systems. Stability analysis and dynamic state-feedback stabilization are characterized via new conditions, whose conservatism is extensively discussed. Effectiveness of the proposed methodology is illustrated by numerical examples.

### 09/09/2016 : Sridharakumar Narasimhan (Indian Institute of Technology de Madras)

ECL, H9 (Bourbonnais), 12h30 à 13h30.

* Monitoring and operation of water distribution networks. *

Urban water distribution networks (WDNs) are large scale, complex systems with limited instrumentation. Due to aging and poor maintenance, significant loss of water can occur through leaks in a WDN. The nexus between water and energy reveals that energy consumption requires significant amounts of water while transporting water for end use is a highly energy intensive operation. Hence, it is important to minimize energy consumption while meeting consumer demands at required pressures.

In the first part, we will present our approach to monitoring and leak detection in networks. We present an off-line technique for leak detection in WDNs using repeated water balance and minimal use of additional flow measurements. A multi-stage graph partitioning is used to determine the location of flow measurements, with the objective of minimizing the measurement cost. A multi-objective integer linear programming (ILP) approach is used to solve the graph partitioning problem. We also show that a modified spectral bisection method can be used to derive an approximate solution for large scale networks. The proposed methods are tested on large scale benchmark networks, and the results indicate that on an average flows in less than 3% of the pipes need to be measured to identify the leaky pipe or joint.

In the second part, we will discuss optimal operation of water networks. The system we consider consists of pumps delivering water to different reservoirs in a network, with each reservoir catering to time varying demand. Pumps and ON/OFF valves are used as manipulated variables to control the flow and pressure. The decision variables are the number of pumps to be turned on and the state of the valves in the network over a given horizon and the objective is to minimize energy consumption while meeting the time varying demand. Given the nonlinear nature of the pump operating curve and the hydraulics, this results in a Mixed Integer NonLinear Program (MINLP). We propose to solve by decomposing it into series of sub-problems that can be solved efficiently. Application of these ideas to distribution networks reveals potential significant savings in energy or improvement in supply.

### 08/09/2016 : Hubert Razik, Guy Clerc (Ampère)

INSA (LdV, 3ème étage), 14h00

* Méthodes de diagnostic, health monitoring *

### 02/06/2016, 13h00 : Yvan Duroc (Ampère)

INSA, Bât. Saint-Exupéry, salle MB1 (1er étage)

* Modélisation et conception de composants et systèmes radiofréquences - Exploitation des non-linéarités générées par les puces RFID UHF *

La présentation débutera par un aperçu rapide des thèmes de recherche développés autour des composants et systèmes RF, dans les domaines de l’Ultra Large Bande, la radio cognitive, et plus particulièrement la RFID et ses évolutions. La technologie RFID, connue principalement pour des applications comme la traçabilité, la logistique ou le contrôle d’accès, est en train d’évoluer très significativement. Avec notamment l’émergence de la 5G et surtout de l’internet des objets, la RFID se voit en effet doter de nouvelles fonctionnalités avec de nouveaux types de tags dits augmentés ; tags qui intègrent de nouvelles capacités de capture et de traitement d’informations.

Après avoir rappelé brièvement les grands principes de la RFID, un nouveau concept basé sur l’exploitation des signaux harmoniques générés par les puces RFID UHF sera présenté ; les applications et preuves de concept réalisées seront aussi introduites.

### 19/05/2016, 14h00 : Laurent Piétrac (Ampère)

INSA, salle Corto Maltese, RdC bât. Jules Verne, département Génie Industriel, INSA de Lyon

* Une introduction à la théorie du contrôle par supervision *

Cette présentation a été donnée dans le cadre des conférences thématiques de la priorité M3 "Systèmes et Energies sûrs"

### 13/05/2016 : Sérgio Waitman (Doctorant Ampère)

ECL, H9, 12h30 (PhD Club)

* Incremental stability of piecewise affine (PWA) systems *

The success of the weighted Hoo norm approch to robust analysis of LTI systems has led to several attempts to find a suitable extension for nonlinear systems. Among the propositions, the L2-gain emerged as a natural candidate. However, L2-gain stability is not enough to guarantee certain desired specifications such as unique steady state and disturbance rejection. The use of the incremental L2-gain has been proposed to cope with these deficiencies, and allows to guarantee both qualitative and quantitative properties for nonlinear systems.

In this presentation, we are interested in the incremental stability of piecewise affine (PWA) systems. We present conditions to compute an upper bound to the incremental L2-gain based on dissipativity analysis. These conditions are expressed as linear matrix inequalities (LMI) allowing the construction of a continuous piecewise quadratic storage function. The approach is less conservative than existing techniques in the literature, which are based on the search for a quadratic storage function. It is also shown that these conditions imply incremental asymptotic stability of the system. The result is illustrated with numerical examples.

### 28/04/2016, 13h00 : Laurent Bako (Ampère)

* A nonsmooth optimization approach to robust estimation: from exact recoverability analysis to derivation of error bounds. *

In this talk, we will discuss the problem of identifying a linear model from measurements which are corrupted by both dense and sparse noise sequences.

While the dense noise is generally assumed to be of moderate amplitude and zero-mean, the sparse noise shows up only intermittently in time but when it does, it can take on arbirarily large values. This is a fundamental problem in many estimation-related applications such as fault detection, state estimation in lossy networks, hybrid system identification,

robust estimation, etc. The problem is computationally hard to handle because it exhibits some intrinsic combinatorial features. Therefore, obtaining an effective solution necessitates relaxations that are both solvable at a reasonable cost and effective in the sense that they can return the true parameter vector. The talk will discuss a nonsmooth convex optimization approach (a sum-of-norms estimator) and provides an analysis of its behavior. In particular, it is shown that exact recovery of the true parameter vector is possible under appropriate conditions on the data (when only sparse noise is considered). When sparse and dense noises are simultaneously active, (computable) parametric error bounds are provided.

### 31/03/2016, 14h00 : Claire Lesieur (Ampère)

* Diagnostic and pronostic tools to correct genetically-driven diseases *

Proteins, nanometer size natural object playing many critical roles in living organisms will be discussed. After a brief introduction on the construction and the function of proteins, their robustness, adaptability and fragility to perturbations will be presented in more details. The proteins are modeled as complex systems, graph theory and tiling theory are used to investigate the mechanisms of robustness, adaptability and fragility.

The issue is to understand how a local perturbation propagates globally in the system and leads to functional faults, namely diseases.

The challenges are to find the appropriate local and global measures that underlie faults in order to establish diagnostic tools

### 11/02/2016, 14h30 : Laurent Seppecher (ICJ)

ECL,Bât. E6 (2ème étage)

*Ce séminaire est proposé en collaboration avec l’Institut Camille Jordan. Il aura lieu en salle de conférence du département de mathématiques, bâtiment E6.*

* Méthodes perturbatives et méthodes interférométriques pour l’inversion de problèmes d’imagerie mal posés *

Nous verrons que dans le cadre de certains problèmes d’imagerie médicale mal posés, l’ajout de perturbations mécaniques contrôlées permet d’augmenter considérablement la résolution et la stabilité de la reconstruction. Après modélisation des interactions mécanique/électromagnétisme dans les tissus biologiques, nous verrons qu’il est possible de construire une donnée interne au domaine avec une très bonne résolution. À partir de cette donnée, des problèmes de couplage d’EDP elliptiques permettent de construire le paramètre recherché. Nous porterons spécialement notre attention sur les hypothèses de régularité minimale nécessaires pour assurer la convergence et la stabilité.

Je parlerais également d’une nouvelle méthode d’inversion de problèmes linéaires en présence d’incertitudes sur l’opérateur direct. Cette méthode dite d’inversion interférométrique a des propriétés de stabilité remarquables et s’applique très bien à tout problème de détection et d’identification de sources dans un contexte ondulatoire.