A fault diagnosis method based on convolutional neural network (CNN) was proposed for the weak fault of the engine casing under the rolling bearing fault excitation. The one-dimensional original signal was converted into image signal by using three preprocessing methods: matrix graph method

kurtosis graph method and wavelet scale spectrum. Then the convolutional neural network was used to identify the fault. Through comparative analysis

the fault identification rate of rolling bearing was 95.82%

which was higher than other vibration signal pretreatment methods. At the same time

the fault recognition rate of convolutional neural network was about 7% higher than that of traditional support vector machine (SVM) because it can use deep network structure to extract the fault characteristics of rolling bearing adaptively. The results show that the proposed method is feasible and effective

and has a good generalization ability and robustness.