Data processing in BOTDR distributed strain measurement based on pattern recognition

As a distributed fiber optic sensor able to monitor strain distribution within dozens of miles, BOTDR is widely applied in structural health monitoring (SHM). However, the difficulty of data processing has been in the way of health monitoring of civil structures and hence an obstacle against BOTDR’s development. In order to realize the rapid processing of distributed strain data, the strain measurement line was divided into a series of waveform units to confirm whether waveform units are normal or not by pattern recognition while taking both mean and standard deviation of characteristic vector into consideration, as well as combine neighboring abnormal units as abnormal area and give evaluation and warning to this area by the correspondence between waveforms and fiber deformation. The application of data processing in distributed monitoring of a certain tunnel showed that the abnormal area can be located fast and accurately and the deformation of optical fiber and structure can be evaluated with this technology.