研究目的
To propose a novel method based on artificial neural network for SNR enhancement and measurement acceleration in distributed fiber-optic sensors based on Brillouin scattering.
研究成果
The proposed ANN method significantly improves SNR by 16-22 dB, reducing measurement time and enhancing accuracy in BFS estimation. Combining with moving average filter reduces resource usage and computational complexity, making it faster and more efficient than state-of-the-art methods. Future work could focus on optimizing the ANN further for broader applications.
研究不足
The ANN requires a large amount of memory and has higher computational complexity than ensemble averaging. It is data-dependent, which may limit performance in some scenarios. The implementation is specific to FPGA and may not be directly applicable to other platforms without modifications.
1：Experimental Design and Method Selection:
The paper uses a five-layer feedforward artificial neural network (ANN) combined with a moving average filter for noise reduction and SNR enhancement in Brillouin optical time-domain analysis (BOTDA) sensors. The ANN is designed with specific topology (41-21-21-21-41 neurons) and implemented on FPGA for hardware acceleration.
2：Sample Selection and Data Sources:
A dataset of 100,000 noisy Brillouin gain spectrums (BGS) is used, with 75% for training, 15% for validation, and 10% for testing. The BGS are generated with varying SNRs (0, 5, 10, 15 dB).
3：List of Experimental Equipment and Materials:
FPGA board (ZC706 evaluation board with Xilinx Zynq XC7Z045 SoC), optical fibers, and related sensing equipment for BOTDA.
4：Experimental Procedures and Operational Workflow:
The process involves storing ADC output in memory, applying moving average filter to estimate BFS region, and then processing the reduced data through ANN for noise reduction and SNR enhancement. Quantization to 16-bit fixed-point format is used for hardware implementation.
5：Data Analysis Methods:
Performance is evaluated based on SNR improvement, average absolute error in BFS estimation, processing time, and resource utilization on FPGA. Comparisons are made with existing methods like cross-correlation and curve fitting.
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
