Skip to Main content Skip to Navigation
Theses

Conception d'un système de diagnostic adaptatif et prédictif basé sur la méthode Fuzzy Pattern Matching pour la surveillance en ligne des systèmes évolutifs

Abstract : An industrial process works under two operating states : normal and abnormal one. A set of measurements about these operating states forms the training set, which is divided into classes. Each class is associated with an operating state. The purpose of the diagnosis by pattern recognition is to assign, at each instant, a new measurement to one of the known classes. Fuzzy Pattern Matching (FPM) is a classification method, which uses the fuzzy sets and the possibility theory to take into account the imprecision and the uncertainty of the data. Generally, the number of training samples is not enough to show all the functioning modes of a system. Moreover, predicting the evolution of the functioning mode from normal to abnormal one is useful to avoid the bad consequences. Additionally, in a non-stationary work environment, a classifier must be retrained every time a new sample is classified, to obtain a new knowledge from it. This is an impractical solution since it requires the storage of all available samples and a considerable computation time. The solution is the development of a classifier that is capable to acquire new knowledge from each new classified sample while preserving the current one, this solution is known as incremental learning.
The objective of this thesis is to develop FPM to be capable to realize an adaptive and predictive diagnosis in real time and in a non stationary work environment.
Document type :
Theses
Complete list of metadatas

Cited literature [25 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00002637
Contributor : Moamar Sayed Mouchaweh <>
Submitted on : Thursday, March 27, 2003 - 4:45:23 PM
Last modification on : Saturday, April 20, 2019 - 1:37:51 AM
Long-term archiving on: : Friday, April 2, 2010 - 6:26:17 PM

Identifiers

  • HAL Id : tel-00002637, version 1

Collections

Citation

Moamar Sayed Mouchaweh. Conception d'un système de diagnostic adaptatif et prédictif basé sur la méthode Fuzzy Pattern Matching pour la surveillance en ligne des systèmes évolutifs. Autre [cs.OH]. Université de Reims - Champagne Ardenne, 2002. Français. ⟨tel-00002637⟩

Share

Metrics

Record views

456

Files downloads

1004