Rayleigh Φ-OTDR based DIS system design using hybrid features and machine learning algorithms

Abstract

There is an increasing interest among the researcher community, and industries on the design, and development of a combined distributed acoustic sensing, and a pattern recognition system to detect, and classify potentially dangerous intrusion events. In this work, we describe the design, and optimization of a distributed intrusion sensing system using Rayleigh-phase sensitive optical time domain reflectometry (Rayleigh Φ -OTDR) technique, and supervised machine learning algorithms. The proposed system can classify an intrusion along with the position of an intrusion caused along a single mode optical fiber. We have considered three different external intrusion events, such as a person walking, digging by pickaxe, and electrical drilling. After training and testing the data samples of the simulated intrusion events, we have achieved an average intrusion classification rate of 100% with a 10 dBm of input laser source power over a 25 km length of sensing fiber. The relevant simulated experiments are carried out using MATLAB 20.0 platform.