The high-speed train speed has a higher requirement for stability and operating safety with its operating speed increasing. The main focus of this paper is on the fault characteristics analysis of a high-speed train transmission system. The governing equation and boundary conditions of the transmission system are derived using the finite element method, in which a Timoshenko beam element is introduced to represent the wheelset shaft, and a rigid mass element is utilized to represent the gears, bearings, and wheels. To investigate the vibration response mechanism of a high-speed train transmission system, the critical speed of the system and its modal response are given. According to the types of high-speed train component fault features, the local fault features of gear, bearing, and wheelset are given. Healthy and faulty systems with localized faults in gear, bearing, and wheelset are studied. The characteristics of amplitude and frequency are verified numerically and the transmission characteristics of signal and relation are explained when there is local fault defect in the system. The results show that under different fault types, the time domain response of the system has obvious periodic impulse response faults, and the fault frequency band characteristic distribution is different in the frequency domain response. The simulation results are consistent with the experimental results.

Axle box bearings for locomotives play a significant role in guaranteeing the service performance of the locomotives. This paper has studied how to extract the weak feature of compound fault under strong interference excitations and proposed a kernel correlated kurtosis of square envelope spectrum. Firstly, simulated compound fault signals were applied to analyze important issue of its demodulation; Three signal processing tools are compared to evaluate their detection performance. Secondly, frequency domain correlated kernel was used to quantitatively compute the amplitude value of signal squared envelope spectrum, settle the failure that conventional time-frequency kurtosis fail identify signals of a strong pulse interference fault; namely, with frequency-domain correlated kernel values, Kurtogram is generated. Which can identify the optimum frequency range and decomposed the single fault feature adaptively. Finally, the proposed method was proved effective and applicable in contrast with experimental results.