Aiming at the problem of complex gear operating environment, difficult fault feature extraction and low accuracy, a deep learning fault diagnosis method based on the combination of convolutional neural network (CNN) and Transformer, which is based on three sensors, three channels, and multivariate data processing fusion, is proposed. The method firstly processes the raw data notational by Fast Fourier Transform (FFT) and Short-Time Fourier Transform (STFT), merges the features after feature extraction of these two data processing methods respectively, and then utilizes the CNN-Transformer neural network model for fault diagnosis of gears. The two dimensions of data processing and deep learning model fusion solve the problem of complementary local and global information processing in fault diagnosis, and realize the accurate classification of gear faults.