Aeroengine Gas Path Performance Analysis Based On Distributed Filtering Methods

The gas path performance analysis of aircraft engine plays an important role in engine health management.This paper takes a turbofan engine as the research object,and focuses on the study of gas path performance analysis based on distributed Kalman filter.The nonlinear filtering algorithm under distributed architecture is proposed based on the centralized filter structure and linear filter method.Firstly,the component level model of the turbofan engine and the state space model are presented,and the principle of the engine gas path performance analysis method is briefly explained.Then,federated filtering algorithm is introduced to build the distributed linear Kalman filter.EKF and UKF are presented to design the local filters,and two distributed nonlinear Kalman filters are proposed.Simulations shows that the nonlinear filtering algorithm in distributed architecture has higher accuracy.Aiming at the packet dropout of data network in the distributed architecture,this paper proposes the distributed modified extended Kalman filter.The data packet dropout model in the network transmission process is established,and the state receiving matrix is introduced to obtain an improved measurement equation.Then,the distributed modified extended Kalman filter algorithm is derived,and the estimation equivalence of the distributed and centralized modified extended Kalman filter is analyzed.The simulation shows that the distributed modified extended Kalman filter has higher accuracy when the data packet dropout probability is large,and no additional computing time is added.To deal with the network delay in the distributed architecture,the buffer is applied into the distributed filtering algorithm to propose the distributed self-tuning buffer extended Kalman filter.The network data delay model is established,and the buffer state variable is introduced to design the distributed buffer extended Kalman filter algorithm.The relationship between buffer length and filter stability is discussed,and the minimum buffer length for filter stabilization is given.A self-tuning mechanism of the buffer is designed based on data loss probability and the trace of local posterior variance matrix,and the distributed buffer extended Kalman filter is combined with the buffer self-tuning mechanism.In the simulation,distributed self-tuning buffer extended Kalman filter can get high accuracy and appropriate buffer length to reduce computing time.The better accuracy of gas path performance estimation is obtained.The sensor fault isolation methods are designed for sensor faults based on the filtering under distributed architecture.The nonlinear calculation of distributed modified extended Kalman filter is adjusted to the main filter,and the local filters only performs linear operation.A sensor fault analysis method with consistent status is given,and two sensor fault isolation strategies are designed,including the strategy of isolating faulty local filter and the method to isolate sensors.In the simulation,the sensor isolation has less influence on the precision than the local filter isolation,The computing time of the nonlinear computational separation algorithm is analyzed.The computing time of the local filter is greatly reduced,and the computing time of the main filter is increased.The overall computing time is reduced and it is beneficial to the real-time application of the distributed algorithm.