Abstract: The existing mine ventilator rolling bearing fault diagnosis method only extracts time-frequency component characteristics and adopts shallow network structure, thus causing low fault diagnosis accuracy. In order to solve this problem, a fault diagnosis method of rolling bearing of mine ventilator based on multi-domain characteristic fusion and deep belief network (DBN) is proposed. Firstly, the method performs wavelet packet noise reduction on the bearing vibration signal, and extracts time domain characteristics, frequency domain characteristics and IMF energy characteristics from the bearing vibration signal after noise reduction to obtain a relatively comprehensive set of high-dimensional characteristic set. Secondly, the characteristic selection method based on intra-class and inter-class standard deviation is used to eliminate the characteristics that are not effective for classification and characteristics with no obvious effect so as to filter out high-efficiency characteristics. Finally, kernel principal component analysis (KPCA) is used to reduce and fuse the high-dimensional screening characteristics, eliminate the redundancy between characteristics, and input the fused characteristics into DBN to complete the fault diagnosis. The experimental results show that compared with the diagnosis method based on single characteristic and shallow network, the average accuracy of mine ventilator rolling bearing fault diagnosis method based on multi-domain characteristic fusion and DBN has average accuracy and less average diagnosis time showing good stability and generalization ability for different damage fault data.