Abstract:Based on the noise vibration multichannel realtime acquisition analysis system, the noise sources of the highspeed train surface were tested and analyzed using the planar microphones and highintensity microphones, when the running speed of the train was over 300 km?h-1. The two line sources at the bottom and middle of the train and the point sources for pantographs were equated to the sound source of highspeed train. The noise reference levels of the three sources were determined by the least square method and the theoretical calculation of train through noise test data, based on which, the model of highspeed railway noise prediction was established under multisource mode. At the moment, the highspeed train was running on an embankment line at a speed of 300 km?h-1. The validity of noise prediction was checked against experimental results, with good agreement. The results provide theoretical basis for establishing the systemic model of highspeed railway noise prediction.