| The high-speed train is a complex nonlinear system.In the process of operation,the limit cycle bifurcation of the vehicle system will change due to the influence of conicity of wheel profiles,track excitation and the change of vehicle’s suspension parameters so on,which will affect the operation states of the vehicle(i.e.stability,small amplitude hunting instability and large amplitude hunting instability).The hunting instability of vehicles(small amplitude hunting instability or large amplitude hunting instability)will not only worsen the running performance of vehicles,but also aggravate the wear of wheel and rail,damage vehicles and tracks,and endanger the operation safety of vehicles.Therefore,it is of great significance to study influence factors and detection methods of hunting motion of vehicles for improving the running safety of the high-speed trains.The existing on-line monitoring standards of vehicle hunting instability can only detect the large amplitude hunting instability states of vehicles,but the small amplitude hunting instability states of vehicles can not be effectively monitored.At the same time,the monitoring amount is relatively single,the accuracy is not high,and there are some problems such as false alarm and alarm delay,which pose a potential threat to the operation safety of vehicles.At present,the research on online monitoring methods of small amplitude hunting instability states of vehicles is mostly based on machine learning,deep learning and neural network,and the algorithm is complex,and the theoretical research on monitoring of small amplitude hunting instability states of vehicles is less.Based on this,this paper takes a domestic high-speed train as the research object,and studies influence factors and detection methods of the hunting motion of the high-speed trains.The main research contents and conclusions are as follows:1.According to the dynamic parameters of the domestic high-speed train,the dynamic model of half car and bogie and the dynamic model of the high-speed train are established by MATLAB software and SIMPACK software respectively;The dynamic simulation models is used to reconstruct the three states of high-speed train in operation,namely,stability,small amplitude hunting instability and large amplitude hunting instability,using the measured wheel profiles and the anti-yaw dampers with different damping characteristics.2.Based on the established dynamic model of the high-speed train,the influences of different factors on Hopf bifurcation point,nonlinear critical speed and bifurcation type of vehicle system are analyzed,and then the influence factors of hunting motion of high-speed train are indirectly studied.It is found that the primary suspension parameters,the damping coefficient of the anti-yaw damper and the secondary lateral damper,the creep coefficient and the equivalent conicity have great influence on the Hopf bifurcation point of the high-speed train;the positioning stiffness of the rotary arm positioning nodal points,the damping coefficient and the joints stiffness of the anti-yaw damper,the damping coefficient of secondary lateral damper and the friction coefficient have great influence on the nonlinear critical speed of the high-speed train;the longitudinal stiffness of the rotary arm positioning nodal points,the damping coefficient of the anti-yaw damper,the friction coefficient and the wheel diameter difference have great influence on the bifurcation type of the high-speed train system within a cycle of wheel tread reprofiling.3.29 acceleration sensors are installed at different positions on the carbody,bogie frame and axlebox of the high-speed train;mutual information and dominant frequency analysis are carried out on the lateral acceleration data of the same type of acceleration sensors at different positions,and the position and number of acceleration sensors are optimized.At the same time,considering the compatibility with the existing monitoring system,11 acceleration sensors are left;Then the lateral acceleration data obtained from the experiment are classified,and the mutual information and dominant frequency of the simulation data and the experimental data are carried out.It is found that the simulation data and the experimental data are highly similar in time domain,which indicates that the simulation data can be used in the research.4.By analyzing the periodicity of the simulation data and the experimental data of lateral acceleration,it is found that the periodicity of lateral acceleration data of the large amplitude hunting instability state is the best,the small amplitude hunting instability state is the second,and the stability state is the worst.By using the periodicity difference,two kinds of hunting instability detection methods are proposed to calculate the self-correlation coefficient and the standard deviation of periods of signals.It can effectively detect the small hunting instability states and large hunting instability states of vehicles,which makes up for the deficiency of the existing online monitoring standards only detecting the large hunting instability states of vehicles to a certain extent;The simulation data samples and the experimental data samples are mixed,and the features of time domain and frequency domain of the data are calculated and input into the Support Vector Machine(SVM)classification model,the classification accuracy is 96.67%.The result shows that the simulation data can be used as a supplement in the machine learning model training to make up for the lack of experimental data. |
