Abstract: It first introduces the principle of vibration signal analysis and its application, then designs the vibration signal acquisition and processing module of combine engine, finally realizes the combine fault detection system based on convolution neural network, and carries out the engine fault diagnosis simulation with MatLab. The experimental results show that the fault detection model of convolutional neural network can realize the fault detection of combine engine rotor, bearing and gearbox, and the recognition rate is more than 95%, which proves the reliability and feasibility of the system.