Research On Aero-Engine Sensor Fault Diagnosis And Fault Tolerant Control

The operation of aero-engine control system is based on sensor signals, thus, whether the sensorsignals are normal will directly affect the performance of aero-engine control system. Therefore, inorder to improve the reliability of the control system, it’s necessary to establish sensor fault diagnosissystem and tolerant control system.In this thesis, the sensor fault detection, reconstruction and fault tolerant control methods arestudied. First of all, a system of engine health parameter estimation based on Kalman filter isestablished, and the deviation of sensor signals and the degeneration of engine can be estimatedaccurately by this system. Secondly, a sensor fault diagnosis system based on online training leastsquares support vector regression with sliding window parsimonious is established. Fault sensors canbe detected accurately by this diagnosis system, and it can be used in different types of aero-engines.Secondly, in order to overcome the signal reconstruction ability limitation of the diagnosis systembased on least squares support vector regression, a sensor fault diagnosis and signal reconstructionsystem based on BP neural network and Kalman filter is designed. The diagnosis system can beapplied to aero-engine both at rated and non-rated status and it is not limited by the number of faultsensors or the engine operating conditions. At last, a passive control system and an active faulttolerant control system switching controllers are designed based on the established diagnosis systems.In the two control systems, the affect of fault sensors to aero-engine control system is effectivelyquarantined.The designed systems are applied to a centain aero-engine model. Lots of digital simulations arecarried out which illustrate the validity of the design method adopted in this thesis.