Search

International audience; For mechanical system structural health monitoring, a new residual generation method is proposed in this paper, inspired by a recent result on subspace system identification. It improves statistical properties of the existing subspace residual, which has been naturally derived from the standard subspace system identification method. Replacing the monitored system state-space model by the Kalman filter one-step ahead predictor is the key element of the improvement in statistical properties, as originally proposed by Verhaegen and Hansson in the design of a new subspace system identification method.

The anti-lock braking system (ABS) is an essential part in ensuring safe driving in vehicles. The Security of onboard safety systems is very important. In order to monitor the functions of ABS and avoid any malfunction, a model-based methodology with respect to structural analysis is employed in this paper to achieve an efficient fault detection and identification (FDI) system design. The analysis involves five essential steps of SA applied to ABS, which includes critical faults analysis, fault modelling, fault detectability analysis and fault isolability analysis, Minimal Structural Over-determined (MSO) sets selection, and MSO-based residual design. In terms of the four faults in the ABS, they are evaluated to be detectable through performing a structural representation and making the Dulmage-Mendelsohn decomposition with respect to the fault modelling, and then they are proved to be isolable based on the fault isolability matrix via SA. After that, four corresponding residuals are generated directly by a series of suggested equation combinations resulting from four MSO sets. The results generated by numerical simulations show that the proposed FDI system can detect and isolate all the injected faults, which is consistent with the theoretical analysis by SA, and also eventually validated by experimental testing on the vehicle (EcoCAR2) ABS.

Aiming at improving the reliability and durability of Polymer Electrolyte Membrane Fuel Cell (PEMFC) systems and promote the commercialization of fuel cell technologies, this thesis work is dedicated to the fault diagnosis study for PEMFC systems. In consideration of the sophistication in establishing the accurate and diagnosis-friendly model for PEMFC systems, data-driven fault diagnosis is the main focus in this thesis.As a main branch of data-driven fault diagnosis, the methods based on pattern classification techniques are firstly studied. Taking individual fuel cell voltages as original diagnosis variables, fault detection and isolation (FDI) is achieved through a two-step procedure. The first step is for feature extraction, while the second one is for classification. According to this framework, several representative methodologies in each step are investigated and compared from the perspectives of diagnosis accuracy and computational cost.Specific to the defects on novel class detection and online adaptation capability of conventional classification based diagnosis methods, a novel diagnosis strategy is proposed for PEMFC system diagnosis. A new classifier named Sphere-shaped Multi-class Support Vector Machine (SSM-SVM) and modified diagnostic rules are utilized to realize the novel fault recognition. While an incremental learning method is extended to achieve the online adaptation.Apart from the pattern classification based diagnosis approach, a so-called partial model-based data-driven approach is introduced to handle PEMFC diagnosis in dynamic processes. With the aid of a subspace identification method (SIM), the model-based residual generation is designed directly from the normal and dynamic operating data. Then, fault detection and isolation are further realized by evaluating the generated residuals.The proposed diagnosis strategies have been verified using the experimental data which cover a set of representative faults and different PEMFC stacks. The preliminary online implementation results with an embedded system are also supplied., La thèse est consacrée à l'étude de diagnostic de pannes pour les systèmes pile à combustible de type PEMFC. Ce sujet de recherche est dans le cadre du project ANR DIAPASON2 (Jan. 2011- Sep. 2014).Le but est d’améliorer la fiabilité et la durabilité des systèmes pile à combustible. Afin d'assurer un diagnostic précis et applicable en temps réel, les approches explorées dans cette thèse sont celles du diagnostic guidé par les données.Plusieurs essais sont effectués sur les différentes piles PEMFC. Les données sont obtenues dans les expériences impliquant non seulement l'état de fonctionnement normal, mais aussi les états de défaut. Parmi les méthodes dédiées au diagnostic guidé par les données, les techniques basées sur la reconnaissance de forme sont les plus utilisées. Dans ce travail, les variables considérées sont les tensions des cellules et la démarche adoptée se compose de deux étapes. La première étape consiste à extraire les caractéristiques, et la deuxième étape permet d'établir une classification. Les résultats établis dans le cadre de la thèse peuvent être regroupés en trois contributions principales.La première contribution est constituée d'une étude comparative. Plus précisément, plusieurs méthodes sont explorées puis comparées en vue de déterminer une stratégie précise et offrant un coût de calcul optimal.La deuxième contribution concerne le diagnostic online sans connaissance complète des défauts au préalable. Il s'agit d'une technique adaptative qui permet d'appréhender l'apparition de nouveaux types de défauts. Cette technique est fondée sur la méthodologie SSM-SVM et les règles de détection et de localisation ont été améliorées pour répondre au problème du diagnostic en temps réel.La troisième contribution est obtenue à partir méthodologie fondée sur l'utilisation partielle de modèles dynamiques. Le principe de détection et localisation de défauts est fondé sur des techniques d'identification et sur la génération de résidus directement à partir des données d'exploitation.Toutes les stratégies proposées dans le cadre de la thèse ont été testées à travers des données expérimentales et validées sur un système embarqué.

There exist exactly 1122 pairwise non-isomorphic 2-(56, 12, 3) designs being the residual designs of the known symmetric (71, 15, 3) designs. Six pairwise non-isomorphic strongly regular graphs with parameters (35, 16, 6, 8) were constructed as class graphs of the obtained residual designs. Orders of the full automorphism groups of these graphs are 12, 32, 96, 192, 288 and 40320.

We have constructed 377 non-isomorphic 2-(17,4,3) designs as derived and 649 non-isomorphic 2-(52,13,4) designs as residual designs of symmetric (69,17,4) designs admitting an action of Z6.

∗ This work has been partially supported by the Bulgarian NSF under Contract No. I-506/1995., In this note we construct five new symmetric 2-(61,16,4) designs invariant under the dihedral group of order 10. As a by-product we obtain 25 new residual 2-(45,12,4) designs. The automorphism groups of all new designs are computed.

An orderly algorithm combined with clique searching is used to show that there are—up to isomorphism, in all cases—325,062 resolvable 2-(16,4,2) designs with 339,592 resolutions, 19,072,802 2-(13,6,5) designs, and 15,111,019 2-(14,7,6) designs. Properties of the classified designs are further discussed.

Let H be a Hadamard ( 4 n - 1 , 2 n - 1 , n - 1 ) -design. Suppose that the prime p divides n , but that p 2 does not divide n . A result of Klemm implies that every residual design of H has p -rank at least n . Also, every derived design of H has p -rank at least n if p ? 2 . We show that when H is a skew Hadamard design, the p -ranks of the residual and derived designs are at least n even if p 2 divides n or p = 2 . We construct infinitely many examples where the p -rank is exactly n .