Skip to Main content Skip to Navigation
Theses

Intégration de connaissances a priori pour l'estimation de paramètres physiques de modèles à temps continu : apport pour le diagnostic de systèmes incertains

Abstract : The main concern of this work is the grey box identification of continuous time mono- and multiva- rkiable models that are non-linear with respect to theparameters.Its contributions are twofold. On the one hand, a priori knowledge is taken into account to estimate both physical parameters and uncertainty intervals. On the other hand, the relevance of this approach to the uncertain system diagnosis is shown. This thesis is divided into four parts. The first part presents the main issues and state of the art related to grey box identification as well as PL synthesis of some of the available methods. The working hypotheses and tools are also given. The second part describes the proposed methodology in dealing with a priori knowledge into the physical parameter estimation process.This approach is based on regularization theory and uses a compound criterion. It enables to overcome global non identifiability and to estimate physically sound parameters. This approach is interpreted in the frame of Bayesian inference. The asymptotic covariance matrix of the estimator including a priori information, is obtained analytically under a white noise hypothesis. The methodology is applied, in the third part, to the physical parameter estimation of the a winding process electromechanical. The last part is dedicated to the diagnosis of systems represented by continuous time uncertain models. A new approach that improves fault detection and isolation performances is proposed. It appears to be more relevant as a grey box model reflecting the physical parameter dependence, is considered rather than a black box model with identical input-output behavior.
Document type :
Theses
File URL :
http://docnum.univ-lorraine.fr/prive/SCD_T_2001_0025_DEFRANOUX.pdf
Complete list of metadata

https://hal.univ-lorraine.fr/tel-01746783
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 10:44:46 AM
Last modification on : Monday, December 14, 2020 - 2:48:10 PM

Identifiers

  • HAL Id : tel-01746783, version 1

Collections

Citation

Céline Defranoux. Intégration de connaissances a priori pour l'estimation de paramètres physiques de modèles à temps continu : apport pour le diagnostic de systèmes incertains. Autre. Université Henri Poincaré - Nancy 1, 2001. Français. ⟨NNT : 2001NAN10025⟩. ⟨tel-01746783⟩

Share

Metrics

Record views

19