On Data-Driven Fault-tolerant Control For Traction System Of High-speed Trains

As an efficient,convenient and environment-friendly means of transportation,highspeed train has become a Chinese business card going abroad.The safe and reliable operation of high-speed train is related to the safety of passengers’ lives and property.Therefore,the research on fault-tolerant control of its train traction system has important theoretical significance and application value.At present,most fault-tolerant control are based mathematical modeling methods.However,high-speed train is a strong nonlinear system,so it is difficult to establish an accurate dynamic model.Considering that the repeated operation of high-speed train can produce a large amount of operation data,a data-driven fault-tolerant traction operation control method is proposed in this thesis to ensure the safe and stable operation of high-speed trains.The main work of this thesis is as follows:Firstly,the single mass point model and the multi mass point model of high-speed train along the iteration axis are established,and then they are transformed into equivalent single input and single output compact format dynamic linearization data model and multi input and single output compact format dynamic linearization data model respectively.Secondly,considering the problem that it is difficult to estimate the additive fault of high-speed train actuator,taking the speed,speed error,position,position error,traction/braking force of the train as the sample characteristics and the degree of additive fault as the label,a random forest model with excellent performance is established by training a large number of sample data.And through the test set,the generalization ability,training time and test time of the fault estimation model in this paper is verified better than the fault degree estimation model based on decision tree and the fault degree estimation model based on support vector machine.Then,based on the compact dynamic linearization data model of single mass point train and the fault degree estimation method based on random forest,a single input and single output iterative learning control add-on to model free adaptive fault-tolerant control algorithm is designed,and the strict convergence proof of the control system is given.The simulation results show that the performance of the fault-tolerant control algorithm designed in this thesis is better than pure feedforward adaptive iterative learning fault-tolerant control and pure feedback model free adaptive fault-tolerant control strategy.Finally,based on the compact dynamic linearization data model of multi mass point train,and considering the fault degree estimation model based on random forest,a multi input and single output iterative learning control add-on to model free adaptive faulttolerant control algorithm is designed,and the strict convergence proof is given.The simulation results show that the speed tracking accuracy of the control scheme proposed in this thesis is getting higher and higher with the number of iterations increases,which is better than the model free adaptive fault-tolerant control.There are 50 figures,11 tables and 71 references in this thesis.