Approches multi-modèles de la génération de résidus pour le diagnostic de systèmes

Abstract : The work presented in this thesis deals with the improvement of the capacities of fault detection and isolation modules dedicated to physical systems described by multiple linearised models identified around different operating points. The global process is based on the use of the fuzzy logic as a state selector synthesis tool. This state selector allows, when a system switches from an operating point to another, to perform a progressive accommodation of the elements of a residual generator instead of considering an abrupt change of these elements.The first part of this thesis consists in a study of the fuzzy tools that have been used. The aim of this study is to highlight the pertinence of using fuzzy logic as part of the proposed work, and to aliow the definition of known and reproductive operating conditions such as precision and complexity of generated fuzzy models. A structure is defined for the proposed state selector, structure based on fuzzy models. The output of the state selector consists in a localisation criterion, and the definition of the inputs of this selector is discussed due to its dependence towards many parameters. The second part of this thesis deals with the problem of minimising the modelling errors that occurs when the system performs a transition from an operating mode to another. The proposed work is based on a system on which some measurements can be made and then used as inputs of the fuzzy state selector. The synthesis methodology of a fuzzy transition model, based on the so generated localisation criterion, is developed and validated by a theoretical application to an hydraulic system. This work is then validated by its practical application to this system, and its contribution is shown by a comparison to some results issued from other existing methods. The third and last part of this thesis details, in a frrst time, a method that allows to perform the synthesis of an information vector relevant enough to be used as input vector of the fuzzy state seilector in order that the localisation criterion obtained by this mean can be easily used on a large range of systems. This method is based on an on line accommodation of the synthesis parameters of Kalman filters, a weighting being performed between the influence of the observations and the influence of the models in accordance with the system state. In a second time, the capacity to apply synthesised localisation criterions to a diagnostic module dedicated to systems described by multi-model methods is studied and justified. The proposed accommodation method is then application a pilot plant consisting in a winding machine, on which different kind of faults are generated. On this non linear system, eight different operating points have been defined. The results issued from this application highlights the pertinence of this work for improving fault detection and isolation capacities on systems described by multi-model methods.
Document type :
Theses
File URL :
http://docnum.univ-lorraine.fr/prive/SCD_T_1999_0010_GROSS.pdf
Complete list of metadatas

https://hal.univ-lorraine.fr/tel-01747050
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 10:50:10 AM
Last modification on : Monday, April 16, 2018 - 10:40:34 AM

Identifiers

  • HAL Id : tel-01747050, version 1

Collections

Citation

Laurent Gross. Approches multi-modèles de la génération de résidus pour le diagnostic de systèmes. Autre. Université Henri Poincaré - Nancy 1, 1999. Français. ⟨NNT : 1999NAN10010⟩. ⟨tel-01747050⟩

Share

Metrics

Record views

2