• Financement : ANR
  • Thèse soutenue le :
  • 31 oct. 2025

This thesis falls within the scope of uncertain knowledge representation and the modeling of wastewater networks. The first part addresses the problem of possibilistic conditioning, a fundamental mechanism for updating prior beliefs in response to new observations. The aim is to propose a possibilistic counterpart to Fagin-Halpern conditioning (FH-conditioning), originally defined in Dempster-Shafer theory, and to extend it to the setting of weighted possibilistic knowledge bases, where each logical formula is associated with a degree of certainty. A new revision operator is proposed, fully aligned with the semantic foundations of FH-conditioning and designed to be computationally efficient. This contribution thus enables a syntactic revision process that faithfully reflects the semantic structure of the belief base. The approach is then extended to handle inputs in the form of sets of uncertain observations. The proposed extension is examined from semantic, syntactic, and computational perspectives within the framework of possibility theory, enabling the integration of partial or imprecise information into the belief revision process. The second part of the thesis focuses on the graph-based representation of wastewater networks. In Geographic Information Systems (GIS), network components are often stored in separate geometric databases, which makes it difficult to model the physical connectivity between elements. This thesis addresses connectivity issues commonly found in wastewater networks, such as isolated nodes and unconnected pipe ends. To solve this problem, a graph-based approach is proposed where each component (manhole, pump, structure, etc.) is represented as a node, and pipes are modeled as edges. The algorithm developed ensures the correction of connectivity errors, resulting in a graph representation that more accurately reflects the real-world topology of the network. The approach has been validated using five real-world datasets, confirming its ability to effectively reconstruct the topology of wastewater networks.

Thesis committee

  • Salem Benferhat – PES, Faculty of Sciences Jean Perrin, University of Artois – Supervisor

  • Carole Delenne – HDR, Aix-Marseille University – Supervisor

  • Ahlame Begdouri – PES, Faculty of Sciences and Techniques of Fez – Supervisor

  • Madalina Croitoru – PES, Faculty of Sciences, University of Montpellier – Reviewer

  • Abdelkrim Haqiq – PES, Faculty of Sciences and Techniques, University of Settat – Reviewer

  • Azzeddine Zahi – PES, Faculty of Sciences and Techniques, University of Fez – Reviewer

  • Franco-Alberto Cardillo – Researcher, National Research Council of Italy – Examiner

  • Naneé Chahinian – HDR, Institut de Recherche pour le Développement – Examiner