Kalman-Filtering Based Algorithm for Sensor's Channel Fault Detection and Isolation

The paper aims to the problem of an aircraft engine's automatic control systems (ACS GTE) reliability improvement by using of an algorithmic redundancy. The purpose of the study is the development of the validation algorithms of input measured parameters for the linear adaptive on-board engine model (LABEM) built into the ACS. LABEM is designed for a work in conjunction with ACS GTE in a real environment and satisfy the requirements for compactness, speed and accuracy of engine parameters' identification in statics and dynamics in a wide range of operating modes, flight and engine conditions. The technical and theoretical difficulties of practical implementation of LABEM are associated with the high dimensionality of an engine state space, that are significantly higher than the dimension of the vector of parameters measured on board. The study is devoted to the critical problem of sensor fault identification with subsequent replacement of the measured value with the modeling information. The main relationships for one-dimensional Kalman filter based on the developed predictive model of the metering pin are presented. The fault detection and isolation algorithms for metering pin sensors' channels using the Kalman filter were designed. The algorithms are based on the calculation of the fault signature as weighted sum of the squares of residuals, which is compared with the selected threshold value. The practice results of engines' stand tests and MATLAB-simulation showed the high reliability and quality of ACS GTE based on proposed algorithms.