Simultaneously identify multi-location disturbance events based on SR-CNN and 2D-TCN in ϕ-OTDR system

Abstract

To tackle with the issue that different types of disturbance events will occur at different locations simultaneously in the phase-sensitive optical time-domain reflectometer (ϕ-OTDR) system, a multi-location disturbance event identification method based on Static Regional Convolutional Network (SR-CNN) and 2D Temporal Convolution Network (2D-TCN) is proposed. Taking use of SR-CNN, the 25 km-long sensing fiber is divided into 10 regions and the target region where disturbance events occur can be preliminarily identified. The 2D-TCN is introduced to identify the types of disturbance events as well as their precise locations. Experimental results show that, the different types of disturbance events simultaneously occur at 5 locations can be effectively identified with an average identification accuracy of 93.76 %. Even when the 5 locations are in different target regions, the identification accuracy can still exceed 92.78 %, with an identification time of only 0.77 s. The high identification accuracy, short identification time, and multi-location identification make this method of great value in the application of ϕ-OTDR system.