Fault diagnosis by time-domain beamforming integrated with kurtosis method

Railway engineers have been troubled by imperfections on train wheels since the wheel/rail system was used in modern society. During the railway system operating, impact noise can be generated due to discontinuities of the surface structure of the wheels, but it is often very difficult to determine the position of imperfection in real-time. In this paper, a fault detection approach is proposed to identify and localize the wheel with wheel-flat with the interference of normal wheel and background noise during the rotating movement. The impulsive signal generated by wheel-flat is totally merged in rolling noise and background noise. Beamforming integrated with kurtosis technique is used to search for the temporal-spectrum feature and as an indicator of impulsive signals. The location of the wheel-flat on the wheels can be identified and localized by observing the value of kurtosis on beamforming sound map.