Research On Fault Estimation Method Of High-Speed Trains Based On Kalman Filtering

High-speed railway has the characteristics of high speed,high efficiency and low energy consumption,and has played an increasingly significant role in today’s transportation system.The safe and stable operation of high-speed train is the foundation of the development of high-speed railway,and faults are important factors that affect the safe and stable operation of high-speed train.As the last step of fault diagnosis and the basis of fault tolerant control,fault estimation plays an important role in the safety and stability of high-speed trains.Therefore,it is necessary to study the fault estimation algorithm of high-speed trains.The main work of this thesis is as follows:First,the main force principles of high-speed trains are analyzed.Based on the multiparticle model,the longitudinal dynamics model of high-speed train is established considering the influence of coupler force,velocity and displacement delay between adjacent cars.Lemmas and basic algorithms required in the following chapters are given to provide an important foundation for the following chapters.Second,considering the effect of the wind gust disturbance,the concurrent actuator and sensor faults are viewed as the augmented variables of trains,and a nonlinear descriptor system is established.A robust nonlinear Kalman filter based on disturbance observer is proposed,the disturbance observing item is utilized to estimate the modelable wind gust disturbance,and a robust upper bound is introduced to decrease the influence of system linearization errors on the filtering accuracy.Furthermore,the whale optimization algorithm is improved using the differential evolution algorithm and adaptive parameters for optimizing the covariance matrix of the noise so as to decrease the influence of the measurement deviation of the noise on the filtering accuracy.A simulation example is given to verify the effectiveness of the proposed method.Then,considering speed delays and displacement delays of the train,the concurrent actuator and sensor faults are viewed as the augmented variables of the train,and the augmented time-delay nonlinear system is established.A robust extended Kalman filter is proposed and a robust upper bound is introduced to decrease the influence of linearization errors of nonlinear term and nonlinear time-delay term on the filtering accuracy.On this basis,the wind gust disturbance and external disturbances are regarded as bounded process uncertainties of the system,and multiple sub-optimal fading factors are introduced to adjust the prediction covariance of the robust extended Kalman filter,so as to decrease the influence of uncertainties on the estimation accuracy.Furthermore,the proposed improved whale optimization algorithm is used to optimize the covariance matrix of the noise to decrease the influence of the measurement deviation of the noise on the filtering accuracy.A simulation example is given to verify the effectiveness of the proposed method.Finally,considering the measurement delays and measurement uncertainties,the concurrent actuator and sensor faults are viewed as the augmented variables of the train,and the uncertain augmented nonlinear system with measurement delays is established.Based on matrix transformation,the filtering of the system with measurement time delays is transformed into the filtering of system without time delays.A robust unscented Kalman filter is proposed and a robust upper bound is introduced to decrease the influence of measurement uncertainties on the filtering accuracy.Then,considering the influence of unknown timevarying or drastic changing noise,an adaptive algorithm based on random weighting and moving windowing is utilized to design an unbiased noise estimator,so as to ensure that the robust unscented Kalman filter can still achieve unbiased filtering estimation.Furthermore,the proposed improved whale optimization algorithm is utilized to optimize the stochastic process uncertainties of the system,so as to achieve the concurrent fault estimation of highspeed trains with unknown noise,measurement delays,stochastic process uncertainties and bounded measurement uncertainties.A simulation example is given to verify the effectiveness of the proposed method.