Identification des systèmes par modèles flous linguistiques : prise en compte des aspects numériques et symboliques

Abstract : Building models from numeric and symbolic information relating to system behavior is the subject of this thesis. In the first chapter, the approach used in the domain of identification which is mainly based upon numeric information is compared with the expert system approach which is based on symbolic information. Then, we propose a strategy which takes into account both types of information by using a linguistic fuzzy model. Parametric estimation of these kinds of models from numeric information leads to two possible methods, classical or fuzzy. The second, which is chosen for our development, has the advantage of highlighting two particular criteria for the choice of the model and the parametric estimation method. The second chapter presents the analysis of linguistic fuzzy models and parametric estimation methods, with respect to the criteria defined in chapter 1. We demonstrate that only one model verifies all criteria, the single-input single-output gradual rules model. We also emphasize that no parametric estimation method satisfies the specified conditions. In the third chapter, the gradual rules model is extended to the multi-inputs single-output case. A symbolic form of this model is then proposed in order to define a parametric estimation strategy which fulfills our criteria and is based on the resolution of fuzzy relational equations. A methodology of identification which takes into account numeric and symbolic information is proposed. This is applied to the modelling of a static non-linear system which showcases the main characteristics of the method.
Document type :
Theses
File URL :
http://docnum.univ-lorraine.fr/prive/SCD_T_1997_0278_RONDEAU.pdf
Complete list of metadatas

https://hal.univ-lorraine.fr/tel-01747329
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 10:57:24 AM
Last modification on : Friday, March 30, 2018 - 1:28:32 AM

Identifiers

  • HAL Id : tel-01747329, version 1

Collections

Citation

Laurent Rondeau. Identification des systèmes par modèles flous linguistiques : prise en compte des aspects numériques et symboliques. Autre. Université Henri Poincaré - Nancy 1, 1997. Français. ⟨NNT : 1997NAN10278⟩. ⟨tel-01747329⟩

Share

Metrics

Record views

5