Skip to Main content Skip to Navigation

Tolérance aux Défaillances par Capteurs Virtuels : application aux Systèmes de Régulation d'un Turboréacteur

Abstract : Over the years, market pressure has ensured that engine manufacturers invest in technology to provide clean, quiet, affordable, reliable, and efficient power. One of the last improvements is the introduction of virtual sensors that make use of non-like signals (analytical redundancy). This, is expected to improve weight, flight safety and availability. However, this new approach has not been widely investigated yet and needs further attention to remove its limitations for certificated applications.The concept of virtual sensors goes along with fault tolerance control strategies that help in limiting disruptions and maintenance costs. Indeed, a fault-tolerant control (FTC) scheme, allows for a leaner hardware structure without decreasing the safety of the system.We propose in this thesis work, to monitor through a passive FTC architecture, the Variables Geometries subsystems' of the engine: the VSV (Variable Stator Vane) and FMV (Fuel Metering Valve). A strong constrains, is not to change the parameters of the existing controllers. The approach named AVG-FTC (Variable Geometries Aircraft-Fault-Tolerant Control) is based on several cascaded sub-systems that allow to deal with the Linear Parameter Varying (LPV) model of the systems and modelling errors. The proposed FTC scheme uses a neural model of the sensor associated with a Takagi-Sugeno observer and a Neuronal Extended Kalman Filter Neural (NEKF) to account for those dynamics that cannot be explained with the LPV model to produce a real-time estimate of the monitored outputs. In case of sensor abnormality, the proposed virtual sensors can then be used as an arbitrator for sensor monitoring or as a healthy sensor used by the controller. To evaluate the approach, serval closed-loop simulations, on SNECMA jet-engine simulator have been performed. The results for distinct flight scenarios with different sensors faults have shown the capabilities of the approach in terms of stability and robustness.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Tuesday, March 15, 2016 - 5:17:03 PM
Last modification on : Wednesday, September 28, 2022 - 5:53:59 AM
Long-term archiving on: : Thursday, June 16, 2016 - 10:59:46 AM


Version validated by the jury (STAR)


  • HAL Id : tel-01288916, version 1


Yani Souami. Tolérance aux Défaillances par Capteurs Virtuels : application aux Systèmes de Régulation d'un Turboréacteur. Automatique / Robotique. Ecole nationale supérieure d'arts et métiers - ENSAM, 2015. Français. ⟨NNT : 2015ENAM0021⟩. ⟨tel-01288916⟩



Record views


Files downloads