Détection et localisation de défaillances de capteurs dans un réseau de mesures pluviométriques : cas de l'agglomération nancéienne

Abstract : aIn the field of urban sewer network management, rainfall data is intensively used for on-line rainfall monitoring and for off-line study and design of network hydraulic equipment and management strategies. Sensor data validation is thus an issue of great importance. This thesis investigates the off-line failure diagnosis applied to the rain gauge network of the urban area of Nancy. The objective is to provide indicators on the gauge operation to operators from data collected and processed at a 5-min time rate. Analysis of historical data series reveals a significant failure rate and the possibility of multiple failures. The failures mainly encountered concern the clogging of gauges. From a diagnosis point fo view, the general approach tackled relies on the consistency between observations and an analytical model which represents the normal behaviour of the system. This yields the generation of so-called residuals used as ffailure indicators. The required redundancy relations between rain gauges are expressed in the form of multiple variables regression models between each gauge and a group of surrounding sensors. Several failure detection and isolation methods are then evaluated using raw data including failure cases. Different consistency tests between measurements and model equations with respect to adaptive tolerances are studied. For isolation purposes, attention is particularly focused on the formulation of the formulation of the signatures of the failures into the residuals and different inference techniques are evaluated. An architecture dedicated to the detection and isolation of gauge clogging is finally proposed
Document type :
Theses
File URL :
http://docnum.univ-lorraine.fr/prive/INPL_T_1998_GIULIANI_S.pdf
Complete list of metadatas

https://hal.univ-lorraine.fr/tel-01752237
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 1:39:15 PM
Last modification on : Friday, March 30, 2018 - 1:28:52 AM

Identifiers

  • HAL Id : tel-01752237, version 1

Collections

Citation

Stéphane Giuliani. Détection et localisation de défaillances de capteurs dans un réseau de mesures pluviométriques : cas de l'agglomération nancéienne. Autre. Institut National Polytechnique de Lorraine, 1998. Français. ⟨NNT : 1998INPL054N⟩. ⟨tel-01752237⟩

Share

Metrics

Record views

29