Kalman Filter Based Gas Path Fault Diagnosis For Commercial Turbofan Engine

Gas path fault diagnosis is an important way to improve flight safety and reduce airline operating costs for commercial turbofan engine.The research in this dissertation is focused on the design and development of commercial turbofan engine gas path fault diagnosis technologies based on Kalman Filter(KF)algorithm.Due to the strong nonlinearity of turbofan engine,current engine linear state variable model(SVM)remains high accuracy merely in neighborhood of its modeling point and cannot be satisfied for the whole life cycle which leads to the fact that estimated errors of health parameters increase with flight cycles.An adaptive fitting modeling method is proposed to establish the SVM,so that the Linear Kalman Filter(LKF)can be designed for health parameters estimation.Health parameter coefficient matrices of the SVM are updated every time step based on the previous health parameter estimated values in order to improve the linear model accuracy.Then linear parameter variable model of the commercial turbofan engine at design cruise point is established through the method.By applying similarity transformation,engine performance degradations of 4 flight points are digitally simulated.The LKF selects 8 health parameters except low pressure turbine mass flow capacity and efficiency to be estimated.The results reveal that the proposed modeling method based LKF reduces sums of root mean square errors of health parameters compared to improved fitting modeling method based LKF by averages of 26.4% for operating states where corrected fan speed is under 95% and averages of 45.8% for operating states where corrected fan speed is over 95%.Meantime,the computational time of 60 s simulation duration is less than 20 s.Aiming at the fact that Kalman filter may estimate health paramters inaccurately when available on-board sensors are unevenly distributed and the number of them is less than the number of health parameters,two Neural Network(NN)improved Kalman filter algorithms are proposed: Neural Network-Soft Constrained-Extended Kalman Filter(NN-SC-EKF)and Neural Network-Particle Filter-Kalman Filter(NN-PF-KF).The former optimizes the filter by introducing the estimated result of BP NN as penalty item into objective function of posterior estimation process of KF.The later uses the estimated result of BP NN to correct particles of which the estimated mean and covariance can be computed by PF so that KF can be applied to update all particles.Both algorithms make the estimated results be inclined to the estimated results of BP NN.5 kinds of single component faults and 4 kinds of multiple components faults are then simulated.The results reveal that the two NN improved nonlinear KFs compared to only BP NN and the original 2 nonlinear KFs are greatly improved in term of accuracy of health parameters estimation.The results indicate that the NN-SC-EKF algorithm reduces sums of root mean square errors of all 10 health parameters by average of 27.5% and 31.4% compared to NN and EKF respectively.And the NN-PF-KF algorithm reduces sums of errors by average of 33.8% and 42.1% compared to NN and UKF respectively.Besides,the NN-PF-KF algorithm reduces sums of errors by average of 15.0% compared to the NN-SC-EKF algorithm due to the right isolation of faults produced by NN for scenarios of which fan component is not degraded.The NN-SC-EKF algorithm reduces sums of errors by average of 8.4% compared to the NN-PF-KF algorithm due to the incorrect results produced by NN for the scenarios of which fan faults are encountered.A gas path fault diagnosis framework on Simulink platform based on adaptive fitting modeling method and NN–SC-KF algorithm is proposed.Function blocks of the framework are designed separately.If abrupt gas path fault is not detected,the original Kalman filter is applied to estimate 8 health parameters(not including two low pressure turbine health parameters).Otherwise,the corresponding BP NN is activated according to the flight condition,and NN-SC-KF is then applied to estimate and upate all 10 health parameters.The designed framework is validated thorough simulation and it can realize health parameters estimation for the whole life cycle and isolation of abrupt gas path faults.