基于自适应噪声完备经验模态分解的Φ-OTDR信号去噪算法

De-Noising Algorithm of Φ-OTDR Signal Based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise

摘要:

针对相位敏感光时域反射仪(Φ-OTDR)信号信噪比较低的问题,提出了一种基于自适应噪声完备经验模态分解(CEEMDAN)的小波信息熵阈值去噪算法。该算法配合CEEMDAN分解,通过小波信息熵阈值去噪,提取扰动位置的高频信息,从而提高系统的信噪比。首先,采用Savitzky-Golay滤波算法对不同光脉冲间的信号进行预处理;其次,采用CEEMDAN将滑动差分后单个脉冲内信号分解为不同的固有模态函数(IMF),并利用连续均方误差准则确定含噪较多的高频IMF分量;然后,采用小波信息熵阈值去噪算法对以上高频IMF分量进行阈值去噪处理;最后,将其与低频IMF分量以及残差分量进行重构。采用自行研制的相干Φ-OTDR系统实测数据进行了验证,结果表明,文中算法与两种传统去噪算法相比,信噪比提升了3dB,这对于系统的实际应用具有重要意义。

Abstract:

Aiming at the problem of low SNR of the phase-sensitive optical time domain reflectometer(Φ-OTDR), proposed is a wavelet information entropy denoising algorithm based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN). Combining with the CEEMDAN, the high frequency information of disturbance position was extracted by wavelet entropy threshold de-noising, and the SNR of the system was improved. First, Savitzky-Golay filter algorithm was used to smooth the original Rayleigh scattering backscattering signals between different optical pulses collected by Φ-OTDR. Secondly, CEEMDAN was used to decompose the single pulse signal into intrinsic mode function (IMF) after sliding difference, and the CMSE criterion was used to identify the high frequency band IMF with high noise. Then, the wavelet entropy adaptive threshold denoising algorithm was used to denoise the above high frequency band IMF with wavelet threshold. Finally, it was reconstructed with low frequency band IMF and residual component. The measured data of the self-developed coherent Φ-OTDR system was verified. Results show that SNR is improved by 3dB compared with two traditional denoising algorithms, which is of great significance for practical applications.