Abstract: Gear box, as an important transmission mechanism of helicopter, its operation reliability plays an important role in ensuring the safety of the helicopter system. A lot of expert experience is required to identify the fault classification in the traditional signal processing method, this traditional identification method brings great inconvenience and complexity to fault diagnosis. Based on this above deficiency, a fault diagnosis method of helicopter gearbox based on Short-time Fourier Transform and deep convolutional neural network is proposed. Firstly, the collected vibration signal of the helicopter gearbox is used to extract the time-frequency characteristics of the vibration signal by utilizing the short-time Fourier transform. Afterwards, the forward propagation and back propagation in the deep convolutional neural network are used to train the time-frequency maps of different faults, in order to establish the relationships between different faults and fault features. Then the constructed model is employed to perform the fault diagnosis of the gearbox. The experimental results show that the proposed method can accurately identify different fault of the gearbox with an accuracy rate of over 99%.