Enabling Variable High Spatial Resolution Retrieval From a Long Pulse BOTDA Sensor

Spatial resolution (SR) is one of the most important parameters Brillouin optical time-domain analysis (BOTDA) sensors, which determines minimum length that a perturbation event can be distinguished. In field Internet Things (IoT), there an urgent need for sensors with large-scale high-precision sensing capability scenarios, such as intelligent monitoring production lines and urban infrastructure. Conventionally, SR normally restricted to longer than 1 m due $\sim 10$ -ns acoustic lifetime limitation in silica fibers. For long-distance smart systems, generally on order several meters or even worse. However, it does not meet needs many applications. Therefore, achieve submeter magnitude. this work, first time best our knowledge, we propose convolutional neural network (CNN) process data conventional BOTDA achieves unprecedented performance improvement allows directly retrieve from system use long pump pulses. By using simulated gain spectrums (BGSs) CNN input corresponding high frequency shift (BFS) output target, trained able obtain higher theoretical value determined by pulse width. experiment, accurately retrieves 0.5-m hotspots measured BGS pulses 20 50 ns, acquired BFS great agreement 45/40 ns differential pulse-width pair (DPP) measurement results. Compared DPP technique, proposed demonstrates twofold uncertainty only half time. addition, changing training sets, tunable retrieval based without any requirement hardware modifications. It worth mentioning method also applicable larger widths SR. The post-processing approach paves way enable novel brings substantial over state-of-the-art techniques terms complexity, time, reliability, etc.