The convolutional neural network was introduced into the field of fan fault detection for the first time, a new method based on one dimensional convolution neural network (CNNs) and SoftMax classifier was proposed, which was applied to the fault diagnosis of gearbox planetary gear under different operating conditions. The structure of the network was a double layer structure, the improved convolutional neural network was used for feature extraction, and the SoftMax classifier was used to classify the health status of the signal. Compared with the traditional intelligent algorithm, this method had the advantages of fewer training samples, direct training of network with raw data, high computational efficiency, and it can meet the needs of realtime diagnosis. The data of multi operating conditions are fused and verified by experiments. The experimental results showed that the method can effectively diagnose the gear faults in planetary gear box under different working conditions.