Related Experiment Video
Updated: Aug 2, 2025
A Silicon-tipped Fiber-optic Sensing Platform with High Resolution and Fast Response
Published on: January 7, 2019
Quaternion Wavelet Transform and a Feedforward Neural Network-Aided Intelligent Distributed Optical Fiber Sensing
Lei Fan1,2, Yongjun Wang1,2, Hongxin Zhang1
1State Key Laboratory of Information Photonics and Optical Communications, School of Electronic Engineering, Beijing University of Posts and Telecommunications (BUPT), Beijing 100876, China.
This study introduces a quaternion wavelet transform (QWT) for denoising and a feedforward neural network (FNN) for data analysis in structural health monitoring. These methods enable accurate temperature extraction from Brillouin optical time domain analysis sensors for infrastructure monitoring.
Area of Science:
- Structural Health Monitoring
- Optical Sensing Technology
- Signal Processing
Background:
- Large-scale infrastructure requires robust structural health monitoring (SHM) systems.
- Brillouin optical time domain analysis (BOTDA) offers distributed sensing capabilities.
- Raw sensor data often contains noise, necessitating advanced processing techniques.
Purpose of the Study:
- To develop and demonstrate an effective data processing pipeline for BOTDA-based SHM networks.
- To improve the accuracy and efficiency of temperature extraction from BOTDA sensor data.
- To establish a foundation for real-time monitoring of large infrastructure using optical fiber sensors.
Main Methods:
- Implementation of a Quaternion Wavelet Transform (QWT) algorithm for image denoising.
- Introduction of a Feedforward Neural Network (FNN) for extracting physical information from denoised data.
- Establishment of a Brillouin optical time domain analysis (BOTDA)-distributed sensor system for experimental validation.
Main Results:
- The QWT denoising algorithm and FNN temperature extraction scheme were successfully demonstrated.
- Temperature error was maintained within ±0.11 °C for frequency intervals below 4 MHz, and ±0.15 °C at 6 MHz.
- Temperature distribution extraction using FNN was achieved in under 17 seconds.
Conclusions:
- The proposed QWT denoising and FNN information extraction techniques are effective for BOTDA sensor data.
- The developed methods show promise for real-time monitoring applications in large-scale infrastructure SHM.
- Further technological advancements and algorithm optimization can enhance the performance of these optical fiber sensing technologies.
