Search

International audience; This article presents a new micro and nanoforce sensor used to perform a mechanical characterisation of human oocytes. This device is based on the use of low-sti_ness magnetic springs. The oocytes to be characterised are placed on a force-sensitive platform. A manipulator equipped with a standard micropipette is used to mechanically compress the oocyte. Some complete \force-compression length" curves associated with mechanical load-unload cycles are given. These curves show the linear, the non-linear and also the plastic mechanical behaviour of the oocytes. These characterisations must be considered as a preliminary result which illustrates that the mechanical variability and the mechanical evolution of human oocytes during their maturation process can be observed with a force sensor based on magnetic springs.

International audience; This paper presents a visual method for 1D in-plane displacement measurement which combines a resolution of a few nanometers with an unambiguous excursion range of 168 µm. Furthermore, position retrieval is only based on elementary phase computations and thus might become compatible with high-rate processing by implementing the processing algorithm on high speed computing architectures like a DSP or a FPGA device. The method is based on a twin scale Vernier micro-pattern fixed on the moving target of interest. The two periodic grids have slightly different periods in order to encode the period order within the phase difference observed between the two sub-patterns. As a result, an unambiguous range of 168 µm is obtained from grid periods of 8 µm and 8.4 µm. The resolution is evaluated to be of 11.7 nm despite remaining mechanical disturbances. Differential measurements demonstrated indeed a measurement accuracy better than 5 nm.

International audience; Particle filters are of great concern in a large variety of engineering fields such as robotics, statistics or automatics. Recently, it has developed among Prognostics and Health Management (PHM) applications for diagnostics and prognostics. According to some authors, it has ever become a state-of-the-art technique for prognostics. Nowadays, around 50 papers dealing with prognostics based on particle filters can be found in the literature. However, no comprehensive review has been proposed on the subject until now. This paper aims at analyzing the way particle filters are used in that context. The development of the tool in the prognostics׳ field is discussed before entering the details of its practical use and implementation. Current issues are identified, analyzed and some solutions or work trails are proposed. All this aims at highlighting future perspectives as well as helping new users to start with particle filters in the goal of prognostics.

An efficient predictive maintenance is based on the reliability of the monitoring data. In some cases, themonitoring activity cannot be ensured with individual or wired sensors. Wireless sensor networks (WSN) arethen an alternative. Considering the wireless communication, data loss becomes highly probable. Therefore,we study certain aspects of WSN reliability. We propose a distributed algorithm for network resiliency and datasurvival while optimizing energy consumption. This fault tolerant algorithm reduces the risks of data loss andensures the continuity of data transfer. We also simulated different network topologies in order to evaluate theirimpact on data completeness at the sink level. Thereafter, we propose an approach to evaluate the system’sstate of health using the random forests algorithm. In an offline phase, the random forest algorithm selects theparameters holding more information about the system’s health state. These parameters are used to constructthe decision trees that make the forest. By injecting the random aspect in the training set, the algorithm (thetrees) will have different starting points. In an online phase, the algorithm evaluates the current health stateusing the sensor data. Each tree will provide a decision, and the final class is the result of the majority voteof all trees. When sensors start to break down, the data describing a health indicator becomes incompleteor unavailable. Considering that the trees have different starting points, the absence of some data will notnecessarily result in the interruption of the prediction process., Une maintenance prédictive efficace se base essentiellement sur la fiabilité des données de surveillance.Dans certains cas, la surveillance des systèmes industriels ne peut pas être assurée à l’aide de capteurs individuels ou filaires. Les Réseaux de Capteurs Sans Fil (RCSF) sont alors une alternative. Vu la nature de communication dans ces réseaux, la perte de données est très probable. Nous proposons un algorithme distribué pour la survie des données dans le réseau. Cet algorithme réduit le risque d’une perte totale des paquets de données et assure la continuité du fonctionnement du réseau. Nous avons aussi simulé de différentes topologies du réseau pour évaluer leur impact sur la complétude des données au niveau du noeud puits. Par la suite, nous avons proposé une démarche d’évaluation de l’état de santé de systèmes physiques basée sur l’algorithme des forêts aléatoires. Cette démarche repose sur deux phases : une phase hors ligne et une phase en ligne. Dans la phase hors ligne, l’algorithme des forêts aléatoires sélectionne les paramètres qui contiennent le plus d’information sur l’état du système. Ces paramètres sont utilisés pour construire les arbres décisionnels qui constituent la forêt. Dans la phase en ligne, l’algorithme évalue l’état actuel du système en utilisant les données capteurs pour parcourir les arbres construits. Chaque arbre dans la forêt fournit une décision, et la classe finale est le résultat d’un vote majoritaire sur l’ensemble de la forêt. Quand les capteurs commencent à tomber en panne, les données décrivant un indicateur de santé deviennent incomplètes ou perdues. En injectant de l’aléatoire dans la base d’apprentissage, l’algorithme aura des points de départ différents, et par la suite les arbres aussi. Ainsi, l’absence des mesures d’un indicateur de santé ne conduit pas nécessairement à l’interruption du processus de prédiction de l’état de santé.

International audience; In prognostic's field, the lack of knowledge on the behavior of equipments can impede the development of classical dependability analysis, or the building of effective physic-based models. Following that, artificial neural networks (ANNs) appear to be well suited since they can learn from data gathered from equipments. In this paper, an approach combining a Recurrent Radial Basis Function network (RRBF) and an AutoRegressive with eXogenous inputs model (ARX) is proposed in order to perform the prediction step of prognostics: the ARX attempts to correct the error of predictions of the RRBF. Moreover, since performances of an ANN can be closely related to initial parameterization of the network, a criterion is defined to quantify the reproducibility of predictions and thereby a priori estimate the usefulness of neural network structure. The whole aims at improving the prediction step of prognostics, which is critical with respects to real applicative conditions.

A fuzzy approach for modelling and analysing the recovery activities in discrete event systems is presented. Those essential components of the management of discrete event systems require special reasoning and methods to manage uncertain knowledge. For those purposes, we introduce a tool derived from the fuzzy Petri nets. This tool, inspired from the fault tree, generalizes the defects analysis by a temporal fuzzy approach. The correction requires a recovery which will be developed on a fuzzy rules basis. The recovery, modelled by a dedicated tool, preserves the fuzzy temporal aspect due to a real time information exchange mechanism provided by the monitoring system.

International audience; Disassembly of manufactured products induces both disassembly costs and revenues from the parts saved by the process. At the planning stage a good trade-off has to be found between the costs of disassembly and the final profit. At the control stage it is important to assure an optimal balance of the line as well as the complete disassembly processing during the rest of the working time. A real time control method based on modeling of disassembly by the precedence graph and on a stochastic algorithm is presented in this article.

International audience; The purpose of this paper is to define a framework for the implantation of a distributed, adaptable and open prognostic system able to take into account, on one hand, the dynamic of the monitored equipment and, on the other hand, the evolution of performance criteria. In this way, the prognostic process is (re)defined : at the component level and at the global level of the system (whole equipment). In the distributed model proposed, the interest of neural networks as "prognostic tools" is pointed out. The work is in coherence with actual industrial maintenance developments like "e-maintenance systems" or "web-services applications".

Our thesis work is concerned with the development of experience based approachesfor criticalcomponent prognostics and Remaining Useful Life (RUL) estimation. This choice allows us to avoidthe problematic issue of setting a failure threshold.Our work was based on Case Based Reasoning (CBR) to track the health status of a new componentand predict its RUL. An Instance Based Learning (IBL) approach was first developed offering twoexperience formalizations. The first is a supervised method that takes into account the status of thecomponent and produces health indicators. The second is an unsupervised method that fuses thesensory data into degradation trajectories.The approach was then evolved by integrating knowledge. Knowledge is extracted from the sensorydata and is of two types: temporal that completes the modeling of instances and frequential that,along with the similarity measure refine the retrieval phase. The latter is based on two similaritymeasures: a weighted one between fixed parallel windows and a weighted similarity with temporalprojection through sliding windows which allow actual health status identification.Another data-driven technique was tested. This one is developed from features extracted from theexperiences that can be either mono or multi-dimensional. These features are modeled by a SupportVector Regression (SVR) algorithm. The developed approaches were assessed on two types ofcritical components: turbofans and ”Li-ion” batteries. The obtained results are interesting but theydepend on the type of the treated data., Nos travaux de thèse s’intéressent au pronostic de défaillance de composant critique et à l’estimation de la durée de vie résiduelle avant défaillance (RUL). Nous avons développé des méthodes basées sur l’expérience. Cette orientation nous permet de nous affranchir de la définition d’un seuil de défaillance, point problématique lors de l’estimation du RUL.Nous avons pris appui sur le paradigme de Raisonnement à Partir de Cas (RàPC) pour assurer le suivi d’un nouveau composant critique et prédire son RUL. Une approche basée sur les instances (IBL) à été développée en proposant plusieurs formalisations de l’expérience: une supervisée tenant compte de l’état du composant sous forme d’indicateur de santé et une non-supervisée agrégeantles données capteurs en une série temporelle mono-dimensionnelle formant une trajectoire de dégradation.Nous avons ensuite fait évoluer cette approche en intégrant de la connaissance à ces instances.La connaissance est extraite à partir de données capteurs et est de deux types : temporelle qui complète la modélisation des instances et fréquentielle qui, associée à la mesure de similarité permet d’affiner la phase de remémoration. Cette dernière prend appui sur deux types de mesures : une pondérée entre fenêtres parallèles et fixes et une pondérée avec projection temporelle. Les fenêtres sont glissantes ce qui permet d’identifier et de localiser l’état actuel de la dégradation de nouveaux composants.Une autre approche orientée donnée a été testée. Celle-ci est se base sur des caractéristiques extraites des expériences, qui sont mono-dimensionnelles dans le premier cas et multidimensionnelles autrement. Ces caractéristiques seront modélisées par un algorithme de régression à vecteurs de support (SVR). Ces approches ont été évaluées sur deux types de composants : les turboréacteurs et les batteries « LI-ION ». Les résultats obtenus sont intéressants mais dépendent du type de données traitées.

My postdoctoral research deals with the proposition of data-based failure detection and prognostics approaches. Developments aim at transforming raw data (gathered online on the monitored system) into health indicators that enable estimating the Remaining Useful Life (RUL) and the confidence interval associated to it. Three sets of contributions can be distinguished.–Characterize degradation phenomena. I mainly proposed a new features extraction and selection approach that allows building monotonic and predictable features. This proposition enhances detection of drifts and improve long term predictions of the ageing behavior.–Improve genericity of prognostics. From this point of view, my intent was to develop neural networks and neuro-fuzzy systems that aim at facilitating the building of prognostics algorithms by systematizing the generation of models while reducing the learning time required and avoiding arbitrary choices.–Reliable data-based prognostics. I developed news neuronal algorithms that enable, on one side, to quantify and master the error of prediction, and on the other side, to take into account the non-deterministic aspect of health states.The research project is articulated as follows (short term and mid-term perspectives). First, I would like to develop protocols and metrics for verification and validation of PHM approaches. I am also willing to go through a better modeling of multidimensional systems (physics, time, space). Finally, my intent is to extend PHM to what could be called the “predictive medicine”., L’essentiel des travaux de recherche post-doctorale porte sur le développement d’approches orientées données de détection et de pronostic de défaillances. Les propositions faites visent à transformer un ensemble de données brutes recueillies sur l’équipement surveillé, en un indicateur de temps résiduel avant défaillance (RUL) auquel est associée une confiance. Trois foyers de contributions peuvent être distingués.–Caractériser les phénomènes de dégradation. Je me suis attaché à proposer une nouvelle approche d’extraction et de sélection de descripteurs permettant in fine d’obtenir des indicateurs de santé monotones facilitant la détection d’une dérive et les prédictions à long terme.–Améliorer la généricité du pronostic. De ce point de vue, mon travail a consisté à proposer des méthodes (neuronales et neuro-floues) permettant de systématiser la génération des modèles de pronostic, et de réduire le temps d’apprentissage nécessaire.–Fiabiliser le pronostic orienté données. J’ai sur cet aspect développé de nouveaux algorithmes permettant d’une part, de quantifier et maîtriser l’erreur de prédiction, et d’autre part, de tenir compte de la nature non−déterministe des états de santé.Le projet de recherche à court et moyen terme s’articule autour de trois axes. Le premier a trait à la définition de protocoles et de métriques de vérification et de validation des approches de PHM. Le second porte sur le développement d’outils pour les systèmes multidimentsionnels (physique, temps, espace). Le dernier vise à étendre la thématique à un cadre applicatif pour lequel elle n’a pas été initialement imaginée : le PHM médical.

Constructing prognostics models rely upon understanding the degradation process of the monitoredcritical components to correctly estimate the remaining useful life (RUL). Traditionally, a degradationprocess is represented in the form of physical or experts models. Such models require extensiveexperimentation and verification that are not always feasible in practice. Another approach that buildsup knowledge about the system degradation over time from component sensor data is known as datadriven. Data driven models require that sufficient historical data have been collected.In this work, a two phases data driven method for RUL prediction is presented. In the offline phase, theproposed method builds on finding variables that contain information about the degradation behaviorusing unsupervised variable selection method. Different health indicators (HI) are constructed fromthe selected variables, which represent the degradation as a function of time, and saved in the offlinedatabase as reference models. In the online phase, the method estimates the degradation state usingdiscrete Bayesian filter. The method finally finds the most similar offline health indicator, to the onlineone, using k-nearest neighbors (k-NN) classifier and Gaussian process regression (GPR) to use it asa RUL estimator. The method is verified using PRONOSTIA bearing as well as battery and turbofanengine degradation data acquired from NASA data repository. The results show the effectiveness ofthe method in predicting the RUL., La construction de modèles de pronostic nécessite la compréhension du processus de dégradation des composants critiques surveillés afin d’estimer correctement leurs durées de fonctionnement avant défaillance. Un processus de dégradation peut être modélisé en utilisant des modèles de connaissance issus des lois de la physique. Cependant, cette approche nécessite des compétencespluridisciplinaires et des moyens expérimentaux importants pour la validation des modéles générés, ce qui n’est pas toujours facile à mettre en place en pratique. Une des alternatives consiste à apprendre le modèle de dégradation à partir de données issues de capteurs installés sur le système.On parle alors d’approche guidée par des données.Dans cette thèse, nous proposons une approche de pronostic guidée par des données. Elle vise à estimer à tout instant l’état de santé du composant physique et prédire sa durée de fonctionnement avant défaillance. Cette approche repose sur deux phases, une phase hors ligne et une phase en ligne. Dans la phase hors ligne, on cherche à sélectionner, parmi l’ensemble des signaux fournis par les capteurs, ceux qui contiennent le plus d’information sur la dégradation. Cela est réalisé en utilisant un algorithme de sélection non supervisé développé dans la thèse. Ensuite les signaux sélectionnés sont utilisés pour construire différents indicateurs de santé représentant les différents historiques de données (un historique par composant). Dans la phase en ligne, l’approche développée permet d’estimer l’état de santé du composant test en faisant appel au filtre Bayésien discret.Elle permet également de calculer la durée de fonctionnement avant défaillance du composant en utilisant le classifieur k-plus proches voisins (k-NN) et le processus de Gauss pour la régression.La durée de fonctionnement avant défaillance est alors obtenue en comparant l’indicateur de santé courant aux indicateurs de santé appris hors ligne.L’approche développée a été vérifiée sur des données expérimentales issues de la plateforme PRONOSTIA sur les roulements ainsi que sur des données fournies par le Prognostic Center of Excellence de la NASA sur les batteries et les turboréacteurs.

International audience; Disassembly process decomposes a product into parts or subassemblies. Due to the uncertainties that occur during this process, designing and balancing a disassembly line is a very challenging problem. To deal with the disassembly line balancing problem a new method which relies on the equal piles approach is proposed.

International audience; This paper presents the design of a mesoscale robot for laser phonosurgery. The proposed design is situated between conventional mechanism and MEMS technology.A combination of compliant structures and innovative micromotors enables to achieve two decoupled tilting angle, a high range (up to 45°) and a precise positioning of a laser beam. The design methodology and the optimization of the compliant structure are detailed. Preliminary results and tests are described which have induced promising performances of the mesoscale robot for laser steering.

International audience; In this work we show that implementation of closed loop control to silicon nanotweezers improves the sensitivity of the tool for mechanical characterizations of biological molecules. Micromachined tweezers have already been used for the characterizations of mechanical properties of DNA molecules as well as for the sensing of enzymatic reactions on DNA bundle. However the resolution of the experiments does not allow the sensing on single molecules. Hereafter we show theoretically and experimentally that, reducing the resonance frequency of the system by the implementation of a state feedback, the sensitivity to stiffness variation is enhanced. Such improvement leads to better resolution for detection of enzymatic reactions on DNA.

International audience; Non-contact microrobotics is a promising way to avoid adhesion caused by the well-known scale effects. Nowadays, several non-contact micro-robots exist. Most of them are controlled by magnetic or dielectrophoresis phenomena. To complete this, we propose a method based on electrostatic force induced by chemical functionalisation of substrates. In this study, we show a model of this force supported by experimental results. We reached long range forces measuring an interaction force of several microNewtons and an interaction distance of tens micrometers. This paper shows the relevance of using chemical electrostatic forces for microrobotics applications.

International audience; This work achieved the control of micromachined tweezers for the enhancement of the sensing of DNA molecules and related enzymatic reactions. The mechanical stiffness of the silicon nanotweezers is decreased by feedback design and the sensitivity of the system is drastically improved.

International audience; This paper presents a linear method for kinematic and dynamic modeling of parallel kinematic manipulators. This method is simple, compact and clear. One can write all the equations from the beginning till the end with pen and paper. It is thus well suited to mechanical understanding and computer implementation. We can apply it to many parallel robots. This method relies on a body-oriented representation of observable rectilinear kinematic structures (kinematic elements) which form the robot legs.

International audience; This work presents a stable vision based haptic feedback for micromanipulation using both an asynchronous Address Event Representation (AER) silicon retina and a conventional frame-based camera. At this scale, most of the grippers used to manipulate objects lack of force sensing. High frequency vision detection thus provides a sound solution to get information about the position of the object and the tool to provide virtual haptic guides. Artificial retinas present high update rates, which enables to address one of the major challenge of haptic feedback teleoperation systems, namely stability. However static objects are not detected. The haptic feedback is thus based on an asynchronous silicon retina to provide a high update rate of moving objects and a framebased camera to retrieve the position of the target object. This approach is validated by pick-and-place of microspheres (diameter: around 50 micrometers) using a piezoelectric microgripper. The displacement of the tool, as well as the opening and closing of the gripper are controlled by the haptic device. Haptic virtual guides are transmitted to users to assist them in the different steps of the pick-and-place task: a virtual stiffness ensures the correct alignment of the tool with respect to the object, a repulsive haptic force enables users to monitor the gripping step, and operators are assisted while picking and placing the object.

International audience; Prognostic aims at estimating the remaining useful life (RUL) of a degrading equipment, i.e at predicting the life time at which a component or a system will be unable to perform a desired function. This task is achieved through essential steps of data acquisition, feature extraction and selection, and prognostic modeling. This paper emphasizes on the selection phase and aims at showing that it should be performed according to the predictability of features: as there is no interest in retaining features that are hard to be predicted. Thereby, predictability is de ned and a feature selection procedure based on this concept is proposed. The e ectiveness of the approach is judged by applying it on a real-world case: through comparison is made in order to show that the better predictable features lead to better RUL estimation.

International audience; Although efforts in this field are significant around the world, real prognostics systems are still scarce in industry. Indeed, it is hard to provide efficient approaches that are able to handle with the inherent uncertainty of prognostics and nobody is able to a priori ensure that an accurate prognostic model can be built. As for an example of remaining problems, consider datadriven prognostics approaches: how to ensure that a model will be able to face with inputs variation with respect to those ones that have been learned, how to ensure that a learned-model will face with unknown data, how to ensure convergence of algorithms, etc. In other words, robustness, reliability and applicability of a prognostic approach are still open areas. Following that, the aim of this paper is to address these challenges by proposing a new neural network (structure and algorithm) that enhances reliability of RUL estimates while improving applicability of the approach. Robustness, reliability and applicability aspects are first discussed and defined according to literature. On this basis, a new connexionist system is proposed for prognostics: the Improved-Extreme Learning machine (Imp-ELM). This neural network, based on complex activation functions, enables to reduce the influence of human choices and initial parameterization, while improving accuracy of estimates and speeding the learning phase. The whole proposition is illustrated by performing tests on a real industrial case of cutting tools from a Computer Numerical Control (CNC) machine. This is achieved by predicting tool condition (wear) in terms of remaining cuts successfully made. Thorough comparisons with adaptive neuro fuzzy inference system (ANFIS) and existing ELM algorithm are also given. Results show improved robustness, reliability and applicability performances.

International audience; In the field of microrobotics, actuators based on smart materials are predominant because of very good precision, integration capabilities and high compactness. This paper presents the main characteristics of Magnetic Shape Memory Alloys as new candidates for the design of micromechatronic devices. The thermo-magneto-mechanical energy conversion process is first presented followed by the adequate modeling procedure required to design actuators. Finally, some actuators prototypes realized at the Femto-ST institute are presented, including a push-pull bidirectional actuator. Some results on the control and performances of these devices conclude the paper.

International audience; The health assessment of composite structures from acoustic emission data is generally tackled by the use of clustering techniques. In this paper, the K-means clustering and the newly proposed Partially-Hidden Markov Model (PHMM) are exploited to analyse the data collected during mechanical tests on composite structures. The health assessment considered in this paper is made difficult by working in unconstrained environments. The presence of the noise is illustrated in several examples and is shown to distort strongly the results of clustering. A solution is proposed to filter out the noisy partition provided by the clustering methods. After filtering, the PHMM provides results which appeared closer to the expectations than the K-means. The PHMM offers the possibility to use uncertain and imprecise labels on the possible states, and thus covers supervised and unsupervised learning as special cases which makes it suitable for real applications.

International audience; Platinum (Pt) nanoparticles were successfully electrodeposited in situ on an organic conductive polymer, poly(3,4-ethylenedioxythiophene) (PEDOT), using for the first time sinusoidal voltages of various frequencies in a chloroplatinic acid solution. The organic PEDOT matrix was electrodeposited on Pt electrode chips. The Pt electrode chips consist of a 150 nm Pt layer deposited on 100-oriented standard 3'' silicon wafers. The cyclic voltammograms of the PEDOT-Pt-nanoparticles composite material recorded in 0.5 M H2SO4 aqueous solution demonstrated that Pt nanoparticles are electrochemically active. Values of the roughness of the composite materials, measured by optical non-contact 3D profilometry, ranging from 880 nm to 1.6 m were obtained depending on the time of deposition of the nanoparticles. The PEDOT-Pt-nanoparticles composite deposited by a sinusoidal voltage with a frequency range of 0.1 Hz - 100 kHz, 50 frequencies, has the largest active surface area (5.16 cm2) compared with other composite coatings prepared in this work and those previously reported. Atomic force microscopic (AFM) images revealed the presence of numerous deposited Pt nanoparticles on the organic PEDOT polymer film.

International audience; This paper discusses variable selection for the efficient dynamic control of the Quattro parallel robot through an inverse dynamic model expressed by means of leg orientations. A selection is made within a group of variables where each can imply the state of the robot. Besides, in this work, steering a parallel robot dynamically using its self-projection onto the image plane (where the edges of the lower-legs are exploited in control) is proposed and validated for the first time. In the light of the realistic control simulation, the formative points of better control of the Quattro robot are figured out.

International audience; Within condition based maintenance (CBM), the whole aspect of prognostics is composed of various tasks from multidimensional data to remaining useful life (RUL) of the equipment. Apart from data acquisition phase, data-driven prognostics is achieved in three main steps: features extraction and selection, features prediction, and health-state classification. The main aim of this paper is to propose a way of improving existing data-driven procedure by assessing the predictability of features when selecting them. The underlying idea is that prognostics should take into account the ability of a practitioner (or its models) to perform long term predictions. A predictability measure is thereby defined and applied to temporal predictions during the learning phase, in order to reduce the set of selected features. The proposed methodology is tested on a real data set of bearings to analyze the effectiveness of the scheme. For illustration purpose, an adaptive neuro-fuzzy inference system is used as a prediction model, and classification aspect is met by the well known Fuzzy Cmeans algorithm. Both enable to perform RUL estimation and results appear to be improved by applying the proposed strategy.

International audience; The aim of this paper is to study the force and position tracking problem of parallel robot manipulators. Relying on a recent work showing that computed torque control in Cartesian space is suitable for parallel structures, we propose a parallel force position control scheme of a parallel robot based on the visual servoing of the end effector pose. Simulation results show the efficiency of the proposed approach.

International audience; Condition-based maintenance (CBM) appears to be a key element in modern maintenance practice. Research in diagnosis and prognosis, two important aspects of a CBM program, is growing rapidly and many studies are conducted in research laboratories to develop models, algorithms and technologies for data processing. In this context, we present a new evolving clustering algorithm developed for prognostics perspectives. E2GK (Evidential Evolving Gustafson-Kessel) is an online clustering method in the theoretical framework of belief functions. The algorithm enables an online partitioning of data streams based on two existing and efficient algorithms: Evidantial c-Means (ECM) and Evolving Gustafson-Kessel (EGK). To validate and illustrate the results of E2GK, we use a dataset provided by an original platform called PRONOSTIA dedicated to prognostics applications.

International audience; Full automated micro-assembly is an ongoing challenge for researchers. The use of force control constitutes a suitable approach to achieve automated micro-assembly. It takes into account microscale specificities like pull off forces, measurement resolution... The integration of force sensors in micro-assembly station is discussed. An experimental setup is proposed to achieve automated guiding tasks of 2mm x 50 μm x 50 μm microparts. It is based on two-sensing-finger microgripper of 2 mN force range. Interaction forces are modeled and used to establish suitable strategy for guiding according to stability conditions and response time. Control scheme which combines force and position control is proposed and its integration to dSPACE board for real time control is achieved. Automated guiding task is performed with dynamic rejection of perturbation within 35 ms and a ramp tracking with a contact force of 20 μN between the rail side and the micropart is obtained.

International audience; Force feedback control constitutes a promising solution for achieving fully automated micro-assembly. In this paper, we report the study of force feedback control used during guiding task. An experimental setup with two sensorized fingers is proposed. It enables to grasp a micropart of 2 mm x 50 μm x 50 μm in size and to estimate the lateral contact force. A simulator is developed to take into account the gripping forces evolution during a lateral perturbation. External hybrid force/position control is chosen for performing automated guiding tasks. Experimental results demonstrate the ability of the implemented controller and suitable design of microsystems.

International audience; During micro-assembly processes, planar contacts often appear (between two manipulated components, one manipulated component and its gripping tool or substrate). Adhesion forces being predominant at the microscale, such contacts have important consequences on assembly strategies. One of these adhesion forces is the pull-off force which is the necessary force to break a contact. This article focuses on the measurement of pull-off force of micrometric planar contacts (50x50 m2) and shows that it can be in the range of several hundreds of microNewtons. Consequences of this pull-off force on assembly strategy, especially guiding tasks, are then introduced. A strategy based on a force study is applied on teleoperated assembly sequences and validates the proposed guiding strategy. The study also enables to establish design rules that are specific for the microscale and that can e applied to the field of MOEMS assembly.

International audience; One of the key steps in high-speed control of a parallel robot is to define an efficient dynamic model. It is usually not easy to have such a model for parallel robots, since many of them have complex structures. Here, we propose a vector-based approach, which employs the robot leg orientations, to obtain a simplified inverse dynamic model. At the least, this vector-based methodology is pioneering, when combined with the observation of orientations by a calibrated camera, in the sense of solving the entire control-oriented (hard) modeling problem, both kinematics and dynamics, in an almost algebraic manner through the knowledge of only a nominal set of image features: the edges of the robot legs and their time derivatives. Proposed method is verified on a simulator of the Quattro robot with a computed torque control where the leg orientations are steered.

International audience; Vision has often been considered as not suitable for dynamic control of robots. The experimental results presented in this paper show that it is possible to perform better with a vision based dynamic control than with a modelbased control. These results were obtained using a Cartesian computed torque control fed back, without any joint sensing, by a novel Cartesian pose and velocity estimator. The latter is designed as a virtual visual servoing scheme based on sequential acquisition of sub-images and a constant acceleration motion assumption.

International audience; The paper deals with the problem of imaging at the microscale. The trifocal transfer based novel view synthesis approach is developed and applied to the images from two photon microscopes mounted in a stereoscopic configuration and observing vertically the work scene. The final result is a lateral virtual microscope working up to 6 frames per second with a resolution up to that of the real microscopes. Visual feedback, accurate measurements and control have been performed with, showing it ability to be used for robotic manipulation of MEMS parts. Keywords: Novel view synthesis, trifocal tensor, photon microscope, microassembly, micromanipulation, MEMS.

International audience; In maintenance field, prognostic is recognized as a key feature as the prediction of the remaining useful life of a system which allows avoiding inopportune maintenance spending. Assuming that it can be difficult to provide models for that purpose, artificial neural networks appear to be well suited. In this paper, an approach combining a Recurrent Radial Basis Function network (RRBF) and a proportional integral derivative controller (PID) is proposed in order to improve the accuracy of predictions. The PID controller attempts to correct the error between the real process variable and the neural network predictions.

International audience; This paper addresses the problem of controlling the dynamics of parallel kinematic manipulators from a global point of view, where modeling, sensing and control are considered simultaneously. The methodology is presented through the examples of the Gough-Stewart manipulator and the Quattro robot.

International audience; This paper is focused on a modeling procedure wellsuited for the design of micro-mechatronic systems and especially for micro-actuators. The purpose of this publication is to show that the variational and energetical methods is not only wellsuited to model classical micro-mechatronic devices but that they are also well-suited to include complex dynamical behaviour such as non-linearity and hysteretical behaviour. This procedure is applied to the design of a new actuator using one of the relatively new smart materials, the Magnetic Shape Memory Alloys (MSMAs). It should be stressed that the presented approach can be extented to a great range of other smart materials and that the description can be easily extented up to the control level.

National audience; Les alliages à mémoire de forme magnétiques sont des matériaux actifs dont les caractéristiques combinent celles des alliages à mémoire de forme classiques et celles des matériaux magnétostrictifs. Ces matériaux relativement récents sont actuellement assez peu utilisés pour des applications pratiques telles que les actionneurs et les capteurs en raison des difficultés de conception et de contrôle résultant d'un comportement complexe. Cet article se propose de présenter le mode de conversion d'énergie à l'oeuvre dans ces matériaux ainsi qu'un état de l'art des principaux travaux en termes d'actionneurs. Il se propose ensuite de présenter les réalisations des auteurs ainsi que les modèles de comportement développés pour finir par une présentation des règles de conception et de contrôle mis en évidence.

Comme cela a été vu précédemment, la manipulation d'objets micrométriques nécessitent l'utilisation de techniques particulières adaptées aux forces en présence à cette échelle. Les solutions présentées apportent des réponses satisfaisantes mais leur intégration dans des systèmes de production représente un nouveau challenge. Il s'agit de passer de la capacité à micromanipuler à la capacité à microassembler, c'est-à-dire à manipuler des microcomposants de façon référencée pour être capable de les positionner et de les solidariser ou encore de créer une mobilité entre eux avec un haut niveau de précision.

In the mechatronics field, smart materials use is in continuous evolution because of the coming of new materials always more efficient. Magnetic Shape Memory Alloys (MSMAs) are active materials relatively recently discovered which present strain of 6 to 10 % under a magnetic field action. Because of their fast actuation and high strain, these alloys are good candidates for new actuators design. However, nonlinear behavior and hysteresis have to be taken into account in the design stage and the control development to increase the characteristics of the new systems. In this thesis, we propose a modeling of MSMA based on the thermodynamics of irreversible processes with internal variables. It takes into account the magnetic and mechanical applied fields including temperature effects. Nonlinear dynamic modeling of systems using MSMA is also developed through lagrangian and hamiltonian formalisms. These energetic models help us to propose new actuator structures and new hybrid controllers well suited for the dynamic hysteretic behavior of this kind of actuators. Theoretical results of these studies are validated by prototypes builing and experimental test benches., Dans le domaine de la méctronique, l'utilisation de matériaux actifs est en constante évolution en raison de l'arrivée de nouveaux matériaux toujours plus performants. Les Alliages à Mémoire de Forme Magnétiques (AMFMs) sont des matériaux actifs apparus relativement récemment qui peuvent se déformer de 6 à 10 % sous l'application d'un champ magnétique. Permettant un actionnement rapide avec des déformations importantes, ces alliages sont des bons candidats pour la conception de nouveaux actionneurs. Cependant, ils présentent des caractéristiques non-linéaires avec hystérésis qu'il faut prendre en compte dès la phase de conception et de développement de leur commande afin d'améliorer les performances des nouveaux systèmes. Dans cette thèse, nous proposons une modélisation des AMFMs basée sur la thermodynamique des processus irréversibles à variables internes prenant en compte les effets du champ magnétique, des contraintes mécaniques appliquées ainsi que de la température. Une modélisation dynamique non-linéaire de systèmes mettant en oeuvre des AMFMs est également développée en utilisant les formalismes de Lagrange et Hamilton. Ces modèlisations énergétiques nous permettent de proposer des nouvelles structures d'actionneurs ainsi que des nouvelles lois de commande hybrides bien adaptées au comportement dynamique hystérétique de ce type d'actionneurs. Les résultats théoriques de ces études sont validés par des réalisations de prototypes et des essais expérimentaux.

International audience; The good generalization performance of support vector machines (SVM) has made them a popular tool in artificial intelligence community. In this paper, we prove that SVM multi class algorithms are not equivalent for all classification problems we present a new approach for incremental learning using SVM that create a rejection class which would be interesting for fault diagnosis where fault classes usually evolve with time : It is when some new samples may be rejected by all the current classes. Hence, these samples may correspond to a new fault (a new class) which may appear after the first training step.

International audience; Multi-agent systems (MAS) are a field of study of growing interest in a variety of domains such as robotics or distributed controls. The article focuses on decentralized reinforcement learning (RL) in cooperative MAS, where a team of independent learning robot (IL) try to coordinate their individual behavior to reach a coherent joint behavior. We assume that each robot has no information about its teammates'actions. To date, RL approaches for such ILs did not guarantee convergence to the optimal joint policy in scenarios where the coordination is difficult. We report an investigation of existing algorithms for the learning of coordination in cooperative MAS, and suggest a Q-Learning extension for ILs, called Hysteretic Q-Learning. This algorithm does not require any additional communication between robots. Its advantages are showing off and compared to other methods on various applications : bimatrix games, collaborative ball balancing task and pursuit domain.

The combination of photon microscope and camera is used in the applications of micro-assembly. This essential instrument have five properties : a weak field of view, a low depth of field, a short working distance, a strong dependence to illumination and bulkiness. In that context of micro-assembly the vision system should be distributed and composed of complementary imagers are expensive and bulky. Therefore our approach consists in obtaining the essential views for the control of the station by the image based rendering. It decreases the number of photonic microscopes in the station. For that, we propose to used two techniques, the mosaicing and the trifocal transfer. First the vision system calibration is required, we present the stages of stereoscopic weak calibration for that context of photonic microscopy with improved new algorithms. Image mosaicing is a technique for the reconstruct a wide image of scene from a set of images that scene where each represents a small part of this scene. After new algorithms improving the construction of mosaic we use this technique for the supervision of micromanipulation station. Trifocal transfer is a technique based on stereoscopic weak calibration for the reconstruction of a virtual view from only two real without 3D explicit information. After an explanation on the theory and a new algorithm proposed we used this technique in a the visual servoing application, to control the position of a microgripper. The virtual view allowing control is obtain from the real view coming from a stereoscopic photonic microscope., La combinaison de microscope photonique et de caméra est largement utilisée dans les applications de micro-assemblage. Cet instrument indispensable comporte cinq propriétés : un faible champ de vision, une faible profondeur de champ, une faible distance de travail, une forte dépendance à l'éclairage et un encombrement important. La conséquence directe de ces propriétés dans un contexte de micro-assemblage est l'utilisation de système de vision distribué composé d'imageurs de caractéristiques différentes et complémentaires qui ont un coût et un encombrement important. C'est pour cela que notre approche consiste a reconstruire certaines vues indispensables au contrôle de la station par des techniques de rendu d'images, limitant ainsi le nombre de vidéo microscopes photoniques réels dans la station. Pour cela, nous proposons d'utiliser deux techniques, la construction d'images mosaïques et le transfert trifocal. Avant toute chose le système de vision doit être calibré, nous présentons les étapes du calibrage stéréoscopique faible pour ce contexte de microscopie photonique ainsi que leurs améliorations par de nouveaux algorithmes. La construction d'images mosaïques est une technique qui permet de reconstruire une image complète d'une scène à partir d'un ensemble d'images représentant chacune une petite partie de cette scène. Après avoir présenté de nouveaux algorithmes en vue d'améliorer la construction de mosaïques nous l'utilisons pour la supervision de station de micromanipulation. Le transfert trifocal est une technique qui permet à partir d'un simple calibrage stéréoscopique faible de reconstruire une vue virtuelle à partir de seulement deux vues réelles sans le besoin d'information 3D explicite. Après un éclaircissement sur la théorie et la proposition de nouveaux algorithmes nous l'utilisons dans une boucle d'asservissement visuel de type look-and-move, pour contrôler le déplacement d'une micropince. La vue de côté virtuelle permettant le contrôle est reconstruite à partir des vues réelles provenant d'un stéréo microscope photonique.

National audience; Les stratégies de manipulation fiables et répétables dédiées aux objets dont la taille est inférieure à 100 µm sont un défi majeur pour la microrobotique. Une approche développée au LAB consiste à effectuer l'assemblage de micro-objets dans un milieu liquide afin de réduire les effets d'adhésion, sources de perturbations. Dans ce cadre, nous proposons un nouveau principe de micromanipulation par glace fonctionnant de manière complètement immergée. Le micro-objet est saisi à l'aide d'une microcouche de glace (4µl) créée dans l'eau, et est relâché simplement par fonte de la glace. Le système de préhension permettant la génération locale de glace est composé de deux modules à effets Peltiers contrôlés à l'aide de deux courants électriques. Les premières manipulations sur des objets d'une taille caractéristique de 600 µm ont été réalisées avec succès. En vue d'une optimisation du principe, le système est modélisé en utilisant l'analogie électrique.

National audience; Dans le cadre de la microrobotique, une méthode de synthèse optimale a été développée pour la conception de structures à déformations et actionnement répartis. Les mécanismes compliants sont décrits par l'assemblage de sous-structures élémentaires appelés "blocs" (voir figure), considérant la répartition de matériaux actifs piézoélectriques sous forme distribuée pour l'actionnement. Les paramètres optimaux définissant totalement les structures sont déterminés par un algorithme génétique multicritères, dont la plupart des fonctions objectif sont obtenues par la méthode des éléments finis. Outre des critères purement mécaniques, la liste des critères d'optimisation disponibles est complétée par des critères liés à la commande des systèmes pour placer, par exemple, de manière optimale les blocs actifs d'actionnement, et ultérieurement de mesure, dans la structure (maximisation de la commandabilité de la partie effectrice et maximisation de l'observabilité mécanique externe).

International audience; This paper presents a new method developed for the optimal design of microrobotic compliant mechanisms. It is based on a flexible building block method, called FlexIn, which uses an evolutionary approach, to optimize a truss-like structure made of building blocks. From the first design step, in addition to conventional mechanical criteria, dynamic gramian-based metrics can be considered in the optimization procedure to fit expected frequency responses of the synthetized mechanisms. A planar monolithic compliant coupling structure is obtained by the optimal design method to act as a stroke amplifier for piezoelectric stacked actuators, to operate in both static and dynamic motions, and to passively filter out undesirable vibrations. Finally, performance comparisons between some of the pseudo-optimal FlexIn synthetized compliant mechanisms demonstrate the interests of the proposed optimization method for the design of dynamic operating smart microrobotic structures.

International audience; Designing in an industrial context is a very complex activity. It needs the integration of heterogeneous knowledge or skills, to transform a set of ill-defined requirements into an artefact that satisfies esthetical, functional, technical, economical criteria. Design process optimization with methods like system engineering and core competency building are key issues for managers of product development process, design skills networks, communities of design practices... In the literature, these issues have been dealt with separately. In this paper, we propose a framework of reference that helps us to explicit the content of the core design competency. This framework is a first step beyond an efficient compentency based management of design structures.

International audience; Skills management in industry is one of the most important factors in order to obtain good performance with production means. Especially in the field of maintenance services where the different practical knowledge or skills are their working tools. We address, in this paper, both the assignment and scheduling problems that can be found in a maintenance service. Each task that has to be performed is characterized by the level of skill required. The problem lies with making the decision of which time is the right time for the assignment and scheduling of the correct resource to do the task. For human resources, all skill levels are different, they are considered as unrelated parallel machines. Our aim is to assign new tasks to the adequate resources by giving to the maintenance expert a good and robust possibility.

International audience; An array-based approach is proposed as an effective mean for providing an alternative solution in team building problem. The proposed approach generates a systematic formation of task and team member families by sequencing the rows and the columns of a task-personnel incidence matrix. To obtain the task-personnel incidence matrix, a fuzzy sets approach will be proposed to combine two fundamental incidence matrices : the task-discipline and personnel-discipline incidence matrices. The proposed method is demonstrated by applying to an example in team building problem.

International audience; Manipulating micro-objects whose typical size is under 100 µm becomes an interesting research topic for micro-assembly applications. A comparative analysis of dry and liquid media impacts on surface forces, contact forces and hydrodynamic forces showed that performing manipulation and assembly in liquid surroundings can indeed be more efficient than in dry conditions. We propose a thermal based micromanipulator designed to operate completely submerged in an aqueous medium. The handling principle and the advantages of our proposed submerged freeze microgripper against air working cryogenic grippers are first described. Then, the thermal principle based on Peltier effect, the characteristics of the prototype, and its first micromanipulation tests are reported. In order to manage the heat exchanges in the microgripper, a dynamic thermal model using electrical analogy has been developed for a 3D heat sink of the microgripper. Its validation is presented in the last section.Further works will be focused on the characterization of all parameters and its experimental validation.