Skip to Main content Skip to Navigation

Contribution à la surveillance des systèmes de production à l'aide des réseaux de neurones dynamiques : Application à la e-maintenance

Abstract : The industrial monitoring methods are divided into two categories: monitoring methods based on the existence of the equipment formal model, and those which not use any equipment formal model. Generally, there are many uncertainties in the formal model and for complex industrial equipment, it is very difficult to obtain a correct mathematical model. This thesis presents an application of the artificial neural networks to the industrial monitoring. We propose a new architecture of Radial Basis Function Networks which exploits the dynamic properties of the locally recurrent architectures for taking into account the input data temporal aspect. Indeed, the consideration of the dynamic aspect requires rather particular neural networks architectures with special training algorithms which are often very complicated. In this sense, we propose an improved version of the k-means algorithm which allows to determine easily the neural network parameters. The validation tests show that at the convergence of the learning algorithm, the neural network is situated in the zone called « good generalization zone ». The neural network was then decomposed into elementary functions easily interpretable in industrial automation languages. The applicative part of this thesis shows that a real-time monitoring treatment is possible thanks to the automation architectures. The neural network loaded in a PLC is completely configurable at distance by the TCP/IP communication protocol. An Internet connection allows then a distant expert to follow the evolution of its equipment, and also to validate the artificial neural network learning.
Document type :
Complete list of metadata

Cited literature [147 references]  Display  Hide  Download
Contributor : Ryad Zemouri <>
Submitted on : Monday, May 3, 2004 - 11:52:49 AM
Last modification on : Thursday, November 12, 2020 - 9:42:06 AM
Long-term archiving on: : Friday, April 2, 2010 - 8:07:30 PM


  • HAL Id : tel-00006003, version 1


Ryad Zemouri. Contribution à la surveillance des systèmes de production à l'aide des réseaux de neurones dynamiques : Application à la e-maintenance. Automatique / Robotique. Université de Franche-Comté, 2003. Français. ⟨tel-00006003⟩



Record views


Files downloads