Prediction of fiber Rayleigh scattering responses based on deep learning

Distributed acoustic sensing (DAS) is a fiber sensing technology based on Rayleigh scattering, which transforms optical fiber into a series of sensing units. It has become an indispensable part in the field of seismic monitoring, vehicle tracking, and pipeline monitoring. Fiber Rayleigh scattering responses lay at the core of DAS. However, there are few in-depth studies on the purpose of acquiring fiber Rayleigh scattering responses. In this paper, we establish a deep learning framework based on the bidirectional gated recurrent unit, which is the first time to predict the fiber Rayleigh scattering responses, to the best of our knowledge. The deep learning framework is trained with a numerical simulation dataset only, but it can process experimental data successfully. Moreover, since the responses could have a wider effective bandwidth than the experimental probing pulses, a finer spatial resolution could be obtained after demodulation. This work indicates that the deep learning framework can capture the characteristics of the fiber Rayleigh scattering responses effectively, which paves the way for intelligent DAS.