Research on Fault Diagnosis Algorithm for the Gearbox and Coupling of EMU Based on Neural Network

The gearbox and coupling are key components for EMU to transmit torque and drive vehicles moving. Its faults will directly impact the EMU operating reliability and even endanger the driving safety. Taking into account four categories of fault pattern for single-axle gearbox and coupling and twenty kinds of state for a train, train-level fault identification algorithm for the gearbox and coupling and component-level fault location algorithm for axle are designed using neural network model in order to realize the fault diagnosis for the gearbox and coupling of EMU. Forty-five eigenvalues are selected as the input data through correlation analysis including motor voltage, motor current, rotational speed, temperature, vibration signal and accumulated running time. Furthermore, single fault tests and comprehensive fault tests are conducted to verify the generalization ability of the algorithm using rolling vibration test rig. The results demonstrate that the proposed fault diagnosis algorithm can effectively identity the corresponding fault.