The objective of this PhD thesis is the monitoring of industrial electric systems. In particular, the focus is put on rare events detection with important failures on the system (as electric arcs) from a distributed acoustic sensors network. This PhD deals with the optimization of the following steps: signal digitalization, signal processing and data transmission. Those subjects lie in the context of compressive sensing and signal processing of compressed data. The objective is to drastically reduce both the amount of acquired data by a sensor node and the amount of transmited data in order to evaluate the feasibility of a real-time monitoring of the system equipped with a 100-acoustic-sensors network. Three phases are identified. The first one is an exhaustive state of the art on compressive sensing techniques. In the second phase, the student will adapt classical methods of detection/classification/localization of acoustic sources to the compressive sensing framework. In that phase, performances of the developed method will be assessed with simulations and then through real data. Finally, the third phase will consist in embedding the developed algorithms on a hardware platform (Labview and/or Simulink) in order to validate the approach in real conditions.

This position is open until it is filled.

Département: Département Systèmes et Intégration de Solutions (LETI)
Laboratory: Laboratoire Fonctionnalisation et Autonomie des Objets
Start Date: 01-10-2015
ECA Code: SL-DRT-15-0645
Contact: vincent.heiries<στο>cea.fr