脉冲编码融合DnCNN提升BOTDA信噪比研究

SNR Enhancement for BOTDA by DnCNN and Pulse Coding

针对单脉冲调制的布里渊光时域分析仪(BOTDA)实现长距离传感时信噪比低的问题,提出采用脉冲编码融合去噪卷积神经网络(DnCNN)提高BOTDA信噪比的方案。通过分析Golay编码BOTDA信号强度提升原理和布里渊增益谱(BGS)的噪声特性,构建Golay编码融合DnCNN降噪的信噪比提升方案。实验证明,与单脉冲调制方式相比,在相同脉冲峰值功率情况下,融合方法可将系统传感距离由10.8 km提升至100.0 km,并在10.8 km处将信噪比提高18.92 dB,与仅用Golay编码调制相比,融合方法在100.0 km末端将信噪比提高9.17 dB。进一步验证在5 m空间分辨率且变温区均方根误差(RMSE)小于0.2 MHz的条件下,融合方法所需的累加平均次数与单脉冲调制相比由2000次降至100次,测量时间由1056 s缩短至194 s。研究表明,脉冲编码融合DnCNN方法可有效提高BOTDA系统的信噪比,实现传感距离延长和测量速度提升。

Objective To address the problem of low signal-to-noise ratio(SNR)in long-distance sensing using a single-pulse Brillouin optical time-domain analyzer(BOTDA),a fusion method of pulse coding and denoising convolutional neural network(DnCNN)is proposed to improve the BOTDA SNR over long distances.This method can be widely used in long-distance engineering fields,such as long-distance oil and gas pipeline leakages,optical fiber composite overhead ground wire(OPGW)cable safety warnings,and submarine cable monitoring.When traditional single-pulse BOTDA performs long-distance sensing,it is generally necessary to increase the peak pulse power or the cumulative average number of measurements to obtain higher SNR.However,an excessively high peak input power causes a modulation instability effect,resulting in a decrease in the measurement accuracy of the system.An excessive cumulative average number significantly increases the measurement time of the system.Therefore,a fusion method of pulse code and denoising convolutional neural network(DnCNN)is used to improve the SNR of BOTDA.This method can effectively improve the SNR,extend the sensing distance,and accelerate the measurement speed,while maintaining the spatial resolution of the system.Methods First,the signal strength enhancement principle of Golay coding BOTDA and the noise characteristics of the Brillouin gain spectrum(BGS)are analyzed,and the SNR enhancement scheme of the Golay coding fusion DnCNN is constructed.Under similar experimental conditions,the BGS along the fiber is acquired using single-pulse and Golay coding,and the BGS acquired using Golay coding is denoised using a trained DnCNN.Subsequently,the peak pulse power is increased to 110 mW,and single-pulse measurement signals averaged 2000 times are collected and compared with the signals obtained by the fusion method averaged 100 times.The results are compared at 5 m spatial resolution and root mean square error(RMSE)of the temperature change area of less than 0.2 MHz.The block-matching 3D filtering algorithm(BM3D),complete ensemble empirical mode decomposition with adaptive noise combined with wavelet threshold(CEEMDAN-WT),and DnCNN are used to reduce the noise of the data collected using the Golay coding,and the effects and running times of different noise reduction methods are compared.Finally,a gradient-temperature experiment is conducted to verify the effectiveness of the fusion method at different temperatures.Results and Discussions The results of the experiments show that compared with the single-pulse modulation mode,at the same pulse peak power,the fusion method can increase the system sensing distance from 10.8 km to 100 km,and the SNR at 10.8 km is increased by 18.92 dB.Compared with the Golay coding modulation mode,the fusion method increases the SNR by 9.17 dB at the 100 km end(Fig.9).It is further verified that at 5 m spatial resolution and RMSE of temperature change area of less than 0.2 MHz,the cumulative average times required by the fusion method decreases from 2000 times to 100 times,and the measurement time is shortened from 1056 s to 194 s compared with single-pulse modulation(Fig.12).Comparing the time required by BM3D,CEEMDAN-WT,and DnCNN to reduce the noise of the experimental data,DnCNN only took 4.62 s,whereas BM3D and CEEMDAN-WT required approximately 8 h and 27 h,respectively(Fig.13).The Brillouin frequency shift(BFS)and temperature change in the BFS curve obtained by the fusion method maintains a good linear relationship,and the temperature information along the fiber can be accurately restored at different temperatures(Fig.14).Conclusions In this study,a fusion method of pulse coding and denoising convolutional neural network(DnCNN)is proposed.This method can increase the sensing distance of the BOTDA system from 10.8 km to 100 km under a cumulative average of 100 times and a pulse peak power of 18 mW,and the SNR along the fiber is improved.At a measurement accuracy of 100 km sensing distance,5 m spatial resolution,and RMSE of temperature change area of less than 0.2 MHz,the measurement time of the fusion method is shortened from 1056 s to 194 s compared with the single-pulse method.This fusion method can be used to measure the temperature in long-distance oil and gas pipeline leakages,OPGW cable safety warnings,submarine cable monitoring,and other engineering fields.