The safety detection parameters of the transmission mechanism, which are mainly temperature, vibration, current and voltage, and the safety detection parameters are obtained through the measurement of optical sensors. Aiming at the problem of local fitting of the neural network, it is proposed to apply the principal element analysis method to linearly combine the inputs of the neural network to realize the optimization of the neural network, and the improved neural network is applied to the fault diagnosis of the transmission mechanism of the coal machinery to form a fault diagnosis model. The safety detection and fault diagnosis model of coal mining machinery drive mechanism is constructed using optical sensor and neural network. Comprehensively test the experimental platform, dataset, and static performance indexes to explore the practical application effect of the system in this paper. In the high-speed zone spur gear train fault detection, the recognition rate of the improved neural network (principal element analysis-neural network) is 97.28%, which is much higher than that of the traditional support vector machine and supervised BPDBN network model, and the same is true in the low-speed zone fault testing experiments. Safety detection parameters of the four static characteristics of the index value, are in the system within the allowable range, indicating that the coal mining machine transmission mechanism safety detection and fault diagnosis system of can meet user requirements, can be applied to the actual detection, a better protection of the mine coal mining transmission mechanism safety.