A Real-Time Distributed Deep Learning Approach for Intelligent Event Recognition in Long Distance Pipeline Monitoring with DOFS

Abstract

Intelligent event recognition along the fiber is still a challenging problem in long distance pipeline monitoring with distributed optical fiber sensors (DOFS), because the complicated burying environments are changing from time to time, and the interference sources are unpredictable at different fiber locations. The fixed hand crafted feature extraction is always time-consuming and laborious, and the update of algorithm always lags behind the environmental change, which restricts its practical scale applications. Thus in this paper, we propose a real-time distributed deep learning model by using the efficient 1-D convolutional neural network (1-D CNN), to learn the distinguishable features of different disturbances and identify them automatically by training the raw event data (signal), which also can be updated easily. The experimental results from the real field data for the safety monitoring of oil pipelines demonstrate the effectiveness of the proposed method, which performs better than the 2-D CNN in terms of both recognition metrics and speed.