Deep Learning Enhanced Long-Range Fast BOTDA for Vibration Measurement
Hua Zheng, Yaxi Yan, Yuyao Wang, Xingliang Shen, Chao Lü
Abstract
In this paper, we propose and experimentally demonstrate a scheme of deep learning enhanced long-range fast Brillouin optical time-domain analysis (BOTDA). The volumetric data from fast BOTDA is denoised and demodulated by using a deep video denoising network and a deep neural network, respectively. Benefitting from the advanced deep learning algorithms, the sensing range of fast BOTDA is extended to 10 km successfully. In experiment, vibration signal is measured with a sampling rate of 23 Hz, 2 m spatial resolution, and 1.19 MHz accuracy over 10 km single-mode fiber with only 4 averages. Due to the low computational complexity and GPU acceleration, the network takes less than 0.04 s to process 100 × 21800 data, which is much faster than the conventional algorithms. This method provides the potential for real-time vibration measurement in fast BOTDA with long sensing range.