Bouzid, Soumia2024-05-292024-05-292024https://dspace.univ-annaba.dz//handle/123456789/3503The sensors measuring process variables are important to the success of any diagnosis technique. This thesis describes a method for detecting and diagnosing sensor problems using a Bottleneck Neural Network (BNN). To detect and isolate sensor problems in drinking water distribution systems (DWDS), the BNN technique is employed as a statistical process control tool. This method has been validated in simulation on a nonlinear system: a quadruple tank and a real-world drinking water distribution system. The goal of this study is to make a contribution to the supervision of a DWDS utilizing the Neural Network diagnosis approach. With the increased demand for water owing to population expansion, the approval of new technologies to ensure water quality at a lesser cost is critical.enThesis