In order to carry out the fault diagnosis of urban gas pipeline timely and accurately,thus ensure the safe operation of gas pipeline system,aiming at the problems of fault diagnosis in the traditional fault diagnosis methods such as the data preprocessing was complex and it was difficult to solve the end-to-end,a fault diagnosis algorithm of gas pipeline based on the one-dimensional convolution neural network was proposed on the basis of the classical model VGG-16.This method could directly extract and select the features of the original acoustic emission signals without any transformation in advance.By building a one-dimensional convolution model on Keras,the acoustic emission signals of gas pipelines were collected in the laboratory,and various model evaluation methods such as accuracy,confusion matrix and recall rate were used to evaluate the fault diagnosis effect of the model.The results showed that the fault diagnosis model of gas pipeline based on the one-dimensional convolution neural network can effectively solve the problems of complex data preprocessing,difficult feature extraction and low recognition accuracy in the fault diagnosis process of gas pipelines.