A Kalman Filter-based Approach To Sensor Fault Diagnosis Of Aero-Engines

Aero-engine sensor failure may result in failure of the control system. To improve the reliabilityof the control system, a fault monitoring system must be established. In this dissertation, the researchon sensor fault diagnosis of the aero-engine based on the Kalman Filter has been conducted.The state variable model for the engine has been built first using the least squares fitting method.Then the establishment of the large deviation model for the engine based on the SVM has been madeand the problem of the transformation between the discrete domain model and the continuous domainmodel is solved.Then, according to the theory of Kalman filtering for the fault diagnosis system, a bank ofKalman filters with constant gains is designed. The simulation results show that the proposedapproach can meet the requirement that the output parameters can be well tracked.The design of fault diagnosis system for both soft and hard sensor faults is made. On this basis,the isolation and accommodation of fault signal can be achieved. The simulation tests verify that thissystem can achieve the goal that the fault signal can be rapidly detected and timely isolated. Based onthe single fault diagnosis system, a method for fault diagnosis with two-sensor-failure has beenproposed and proved that it has satisfactory simulation results. Some of the methods to reduce the rateof misdiagnosis and the missed diagnosis rate have been summarized.