Drilling accidents can cause serious economic losses， wasted time， and even threaten life safety. If the drilling equipment can judge the type of accident in time， the accident processing time can be shortened and the development of the situation can be contained. To solve the above problems， this paper proposes a drilling fault diagnosis model of equipment based on neural network. In order to optimize the correct rate of different neural networks in drilling accidents classification， two neural network models of BP and RBF are constructed respectively in nntool of Matlab. Through the simulation test taking the variation trend of construction parameters in a mining area as input parameters， it is found that the best performance in BP neural network is LM and BR algorithm， and the best performance of RBF neural network is PNN algorithm. All three had an accuracy rate of more than 90 percent. But BP neural network is easy to fall into local optimal with unstable performance. On the contrary， PNN neural network has no such limitation， does not require training， and the design process is simple. So PNN algorithm is more suitable for the establishment of drilling fault diagnosis model.