Skip to Main content Skip to Navigation
Theses

Proposition d’une méthode d’apprentissage à base de produits intelligents pour les ateliers partiellement flexibles / dynamiques

Abstract : With the fourth industrial revolution that we are currently experiencing, efficient scheduling of production operations is more than ever necessary for the sustainability of 4.0 industries. While the flexibility of the Flexible Manufacturing System (FMS) provides advantages over uncertain environments, it implies an increasing complexity of the control functions. In this, distributed architectures seem to provide the agility and responsiveness needed. However, they suffer from limitations, mainly due to the myopia phenomenon in decision-making process. Thus, the main difficulty of dynamic control of cyber-physical production systems lies in the balance to be found between a rapid response to disturbances, and the upkeep of a satisfactory overall performance. This thesis proposes an original model that can be applied to a large types of production workshop configurations (flowshop, jobshop, etc.). In addition to the handling of dynamic events, it includes constraints such as multiple product families, interoperability constraints, configuration, etc. The heterarchical architecture thus proposed ensures a control by products exclusively. These are intelligent and autonomous. The main contribution of the thesis work is a hyper-heuristic based mostly on fast scheduling rules used by smart products. To counterbalance their weak overall performance, the concept of decisional context is developed to improve the relevance of selected rules. The proposed hyper-heuristic is based on an original model of encapsulation of rules combined with a simulation tool reducing the phenomenon of myopia. Experiments on four distinct workshop configurations, 64 instances of different dynamic scenarios and more than 30,400 simulations validate the interest of the proposed approach.
Complete list of metadatas

Cited literature [281 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02568076
Contributor : Wassim Bouazza <>
Submitted on : Friday, May 8, 2020 - 1:29:16 PM
Last modification on : Tuesday, May 12, 2020 - 1:11:33 AM

File

Phd BOUAZZA Wassim RG.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-02568076, version 1

Collections

Citation

Wassim Bouazza. Proposition d’une méthode d’apprentissage à base de produits intelligents pour les ateliers partiellement flexibles / dynamiques. Automatique. Université Oran 1 - Ahmed Ben Bella, 2020. Français. ⟨tel-02568076⟩

Share

Metrics

Record views

83

Files downloads

77