Aeroengine Fault Diagnosis Based On Federal Kalman Filter

The diagnosis systems of aircraft engine plays a significant part in realizing condition-based maintenance,ensuring flight safety and reducing operation cost.Aero-engine diagnosis methods are studied in this dissertation.The research of a turboshaft engine fault diagnosis is conducted based on federated Kalman filter.It covers gas path component diagnosis and sensor diagnosis and isolation.In order to solve the problems caused by the variation of measurement noise,the federated fuzzy adaptive Kalman filter is proposed.Firstly,engine state variable model is built based on the component-level nonlinear model of the turboshaft engine,simulations results have verified the precision of the state variable model.Secondly,federated Kalman filter is applied in the design of gas path component diagnosis system and sensor diagnosis system.Compared with the conventional Kalman filter,the federated Kalman filter has higher estimation accuracy and fault tolerance.The advantages of the federated Kalman filter are verified by means of theoretical analysis and simulation experiments.Sensor fault diagnosis system which employs federated Kalman filter is proposed based on conventional Kalman filter.The sensor fault diagnosis system is implemented through adding the sensor fault diagnosis section between local filter and main filter to estimate the fault sensor.Simulation results show that the sensor fault diagnosis system can effectively estimate and isolate fault sensor and has a satisfactory performance estimation on gas path component health parameters when the sensor breaks down.Finally,federated fuzzy adaptive Kalman filter is proposed to deal with time-varying statistics of measurement noise in different working conditions.When the statistics property of the measurement noise has changed,this algorithm modifies the measurement noise covariance matrix through the fuzzy inference system to make it approach the real measurement covariance matrix gradually.The experiments show that this algorithm could weaken the effect of the changed measurement noise to the engine diagnosis systems.