Skip to Main content Skip to Navigation

Intégration de l'expertise humaine en modélisation et identification floues de systèmes

Abstract : This thesis deals with the modeling and identification of systems using the fuzzy sets theory. The focus is on the modeling of a system when not enough data is available about its behavior and in which a human being participates either as a sensor or as an expert. The first section is devoted to fuzzy modeling based on expert knowledge from experts, who know the functioning of the system to be modeled, or from the designer. In the first case, the fuzzification process has be studied in order to take into account the linguistic information. Making use of linguistic modifiers we propose to the experts two notions, precision and derivation, with the intention that they express their knowledge. In the first part of this work, we mainly use the aspect of precision for the construction of a fuzzy model. In the last part, the notion of derivation is used. Finally, we present a study demonstrating the influence of the fuzzy sets shape on functions approximation. Since we are mainly interested in fuzzy models with crisp outputs, the second part of this work is focused on defuzzification. After studying tlte theoretical basis of defuzzification, we specify the different objectives of the process with a classification according to different methods presented in the literature. Depending on the defuzzification method, this consists of a conversion from the fuzzy domain to a numerical one, a conversion with preferences, an optimized conversion according to a criterion, or a conversion under constraints. Following the classification, we propose three defuzzification methods to exploit the coded information of the output fuzzy set which correspond to the expert knowledge of the model output variable. The last section addresses the identification of the rules of a fuzzy model, particularly when the observations of the behavior's system are uncertain and imprecise. This method is used for the identification of fuzzy rules in the comfort evaluation of automobile seats. With this industrial application, we demonstrate the role of human operators, either as sensors or as experts, in the evaluation process. Finally, we have applied our method to the identification of the relationship between discomforts and seat characteristics.
Document type :
File URL :
Complete list of metadata
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 10:57:21 AM
Last modification on : Thursday, February 11, 2021 - 1:18:38 PM


  • HAL Id : tel-01747326, version 1



Rubén Ruelas. Intégration de l'expertise humaine en modélisation et identification floues de systèmes. Autre. Université Henri Poincaré - Nancy 1, 1997. Français. ⟨NNT : 1997NAN10281⟩. ⟨tel-01747326⟩



Record views