Abstract
Brillouin optical correlation-domain reflectometry (BOCDR) is a fiber-optic distributed sensing technique with single-end accessibility and high spatial resolution. In BOCDR, the measured Brillouin gain spectrum (BGS) distribution is generally given by a convolution of the intrinsic BGS distribution and the beat-power spectrum. In most conventional implementations, the Brillouin frequency shift (BFS) distribution is directly obtained using the measured BGS distribution. Determining the BFS distribution on the basis of the intrinsic BGS distribution will give potentially higher spatial resolution, which can be achieved by deconvolution of the measured BGS distribution. In this work, we employ a convolutional neural network to perform this deconvolution processing in BOCDR and show its potential for spatial resolution enhancement. A spatial resolution which is 5 times higher than the nominal value is demonstrated.
Highlights
Optical fiber sensors based on Brillouin scattering have attracted immense interest in the past decades because of their distributed strain and temperature measurement capabilities [1,2,3,4,5,6,7], which find promising applications in damage detection and structural health monitoring
This study focuses on Brillouin optical correlationdomain reflectometry (BOCDR) [7], which operates on the basis of correlation control of continuous light waves in a self-heterodyne scheme
We showed that the spatial resolution can be potentially enhanced through this method by at least 5 times compared to the nominal value calculated using the convolved Brillouin gain spectrum (BGS) distribution
Summary
Optical fiber sensors based on Brillouin scattering have attracted immense interest in the past decades because of their distributed strain and temperature measurement capabilities [1,2,3,4,5,6,7], which find promising applications in damage detection and structural health monitoring. Using a machine learning algorithm to improve the signal processing can be advantageous in terms of simplicity and cost, as it requires the neural network to be properly trained, after which the BFS and the corresponding strain or temperature change occurring at a section of the optical fiber can be conveniently and directly obtained from the measured BGS without the need for additional equipment or deconvolution as in conventional systems. We propose a deep learning algorithm based on convolutional neural network (CNN) to directly obtain the intrinsic BFS distribution from the measured BGS distribution in BOCDR using regression. Pooling layers are occasionally periodically inserted between convolutional layers to reduce the spatial size of the feature map, and more importantly, the number of parameters to be adjusted during training. BN is again inserted between the first and second hidden fully connected layers to improve the training speed and performance
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
