基于CEEMDAN-SOBI对桥梁监测挠度的分离研究

Study on separation of bridge monitoring deflection based on CEEMDAN-SOBI

为实现桥梁挠度监测信号各种效应值的分离,提出1种基于自适应噪声的完备集合经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,CEEMDAN)结合二阶盲源辨识(Second Order Blind Identification,SOBI)的单通道盲源分离算法。首先利用自适应噪声的完备集合经验模态分解将单通道的桥梁挠度信号分解为一系列线性平稳的本征模函数,计算各子序列的排列熵(Permutation Entropy,PE)并将排列熵值相近的序列相加组成新的序列;然后采用K-L散度的判别法剔除虚假的分量,将真实的分量组成盲源分离模型的输入信号;最后采用二阶盲源辨识对输入信号进行盲源分离,得到桥梁监测挠度的各效应值。结果表明:该方法能有效分离挠度监测信号中的各种效应值。

In order to realize the separation of various effect values for the monitoring signals of bridge deflection,a singlechannel blind source separation algorithm based on the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)combined with Second Order Blind Identification(SOBI)was proposed.Firstly,the single-channel bridge deflection signal was decomposed into a series of linear stable Intrinsic Mode Functions(IMF)by using the CEEMDAN,then the permutation entropy(PE)of each subsequence was calculated,and the sequences with the similar permutation entropy values were added to form a new sequence.Secondly,the false components were eliminated by the discriminant method of K-L divergence,and the real components were composed to the input signal of the blind source separation model.Finally,the second-order blind source identification was used to carry out the blind source separation on the input signals,and the individual effect value of the bridge monitoring deflection was obtained.The results showed that the method could effectively separate various effect values of the deflection monitoring signals.