Abstract: Sound and vibration are the direct carriers of running data of rotating machinery equipment. According to the principle of voiceprint monitoring, through data collection, pre-processing and extraction of eigen values, a voiceprint sample and model library of running data and voiceprint monitoring system of vibration data of large pumping station units is established by adopting convolution neural network deep learning recognition model so as to realize the real-time monitoring of the unit running data of the large pumping station.