研究目的
Investigating the optimum diagonal loading level for adaptive matched filtering in φ-OTDR systems to reduce background noise in distributed acoustic sensing applications.
研究成果
The experimental results confirm that the optimum diagonal loading value is negative, as analytically predicted for beamforming applications, and it can be approximated by the negative of the largest eigenvalue of the data covariance matrix for small filter sizes. This improves noise reduction in φ-OTDR systems for distributed acoustic sensing, with potential applications in structural health monitoring. Future work will explore more complex vibration scenarios.
研究不足
The study is limited to single sinusoidal vibrations and relatively small filter sizes (up to N=50). It does not address multiple vibrations or more complex scenarios, and the approximation for optimal DL may not hold for larger filter sizes or different noise conditions.
1：Experimental Design and Method Selection:
The study adapts a theoretical proof from spatial domain beamforming to temporal domain filtering for φ-OTDR systems, using adaptive matched filtering with diagonal loading to maximize signal-to-noise ratio.
2：Sample Selection and Data Sources:
Real φ-OTDR data was gathered in a laboratory environment using a fiber optic cable with synthetic vibrations. Over 20,000 frames were recorded.
3：List of Experimental Equipment and Materials:
A narrow linewidth CW laser (<1 kHz, 22 mW), acoustic-optic modulator (AOM), balanced photodetector (BPD), RF down-conversion stage, analog-to-digital (A/D) converter, single board computer (SBC), sensing fiber under test (FUT) consisting of two fiber reels (
4：46 km and 2 km long), phase-shifter, function generator, and MATLAB for post-processing. Experimental Procedures and Operational Workflow:
The setup interrogated the fiber with a pulse repetition frequency of
5：5 kHz. Data was collected, digitized, and processed. Steps included selecting training samples with vibration inclusion, estimating the covariance matrix, computing weight vectors for varying δ levels, observing outputs, and comparing with eigenvalues. Data Analysis Methods:
Data was analyzed using MATLAB, with eigendecomposition of the covariance matrix to find the largest eigenvalue and compute optimal δ values. SNR was calculated based on filter outputs.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
