TY - JOUR
T1 - Intrusion location technology of Sagnac distributed fiber optical sensing system based on deep learning
AU - Wu, Jinyi
AU - Zhuo, Rusheng
AU - Wan, Shengpeng
AU - Xiong, Xinzhong
AU - Xu, Xinliang
AU - Liu, Bin
AU - Liu, Juan
AU - Shi, Jiulin
AU - Sun, Jizhou
AU - He, Xingdao
AU - Wu, Qiang
PY - 2021/6/15
Y1 - 2021/6/15
N2 - For distributed fiber optical sensing based on Sagnac effect, the intrusion is usually located by notch frequency. However, the notch spectrum is the comprehensive result of the intrusion, so when multiple disturbances simultaneously intrude from different positions of the sensing fiber, it is impossible to establish a mathematical expression between the intrusion position and the notch frequency, this leads to the problem of multi-point intrusion localization. Therefore, in this paper, deep learning technology is used to locate multiple disturbing points in Sagnac distributed optical fiber sensing system, and the related specific technologies of deep learning appling to sagnac distributed optical fiber sensing are studied. First, according to the characteristics of the system, a network structure based on the regression probability distribution is proposed, second, a loss function is constructed. The results show that the trained model can realize the positioning of multiple and single intrusion points.
AB - For distributed fiber optical sensing based on Sagnac effect, the intrusion is usually located by notch frequency. However, the notch spectrum is the comprehensive result of the intrusion, so when multiple disturbances simultaneously intrude from different positions of the sensing fiber, it is impossible to establish a mathematical expression between the intrusion position and the notch frequency, this leads to the problem of multi-point intrusion localization. Therefore, in this paper, deep learning technology is used to locate multiple disturbing points in Sagnac distributed optical fiber sensing system, and the related specific technologies of deep learning appling to sagnac distributed optical fiber sensing are studied. First, according to the characteristics of the system, a network structure based on the regression probability distribution is proposed, second, a loss function is constructed. The results show that the trained model can realize the positioning of multiple and single intrusion points.
KW - Fiber optical sensor
KW - Sagnac interferometers
KW - position measurement
KW - deep learning
KW - Optical fiber sensors
KW - Optical interferometry
KW - Optical fiber cables
KW - Optical fiber networks
KW - Sensors
KW - Optical scattering
UR - http://www.scopus.com/inward/record.url?scp=85103782591&partnerID=8YFLogxK
U2 - 10.1109/jsen.2021.3070721
DO - 10.1109/jsen.2021.3070721
M3 - Article
VL - 21
SP - 13327
EP - 13334
JO - IEEE Sensors Journal
JF - IEEE Sensors Journal
SN - 1530-437X
IS - 12
ER -