基于CEEMD和小波包阈值的组合降噪及泄流结构的模态识别方法  

A de-noising and modal identification combined method based on CEEMD and wavelet packet threshold for flood discharge structures

作　　者：胡剑超[1] 练继建[1] 马斌[1] 董霄峰[1] 

机构地区：[1]天津大学水利工程仿真与安全国家重点实验室,天津300072

出　　处：《振动与冲击》2017年第17期1-9,共9页Journal of Vibration and Shock

基　　金：国家重点基础研究发展计划(973计划)子课题"极端条件下高碾压混凝土坝动力响应机制"(2013CB035905-1)

摘　　要：水流荷载激励下,泄流结构模态识别的实测振动信号常受噪声干扰。以往EEMD算法有添加白噪声造成非标准IMF导致的模态分裂问题及IMF有噪声残余不能完整重构信号的完备性问题。而完备总体经验模态分解(CEEMD)法,通过在信号分解的每一层面添加特定高斯白噪声,利用分解后第一阶分量加总平均得到唯一余量计算固有模态函数,克服了EEMD的缺点;同时提出CEEMD与小波包阈值结合的组合降噪方法,运用到向家坝水弹性模型实测振动信号降噪中,验证了该组合方法降噪的有效性;为了提高带噪振动响应模态识别的精度,基于数据相关技术,利用Markov参数构造相关矩阵R,用该相关矩阵重构Hankel矩阵后SVD分解得到系统最小实现,即数据相关特征系统实现法(ERA/DC)。将滤波降噪结合模态识别的整套方法,应用到锦屏一级拱坝的泄流实测振动响应中,得到了较好的应用效果。Vibration signals measured for modal identification of a flood discharge structure under water load excitation are often interfered by noise. The EEMD algorithm has problems of modes splitting caused by nonstandard IMFs produced due to adding white noise and IMFs with residual noise not being able to completely reconstruct original signals.Here,the complete ensemble empirical mode decomposition（ CEEMD） was proposed. Using it,the drawbacks of EEMD were overcome through adding a specific white noise to each stage of signal decomposition and using the unique residual obtained by adding the average to the first order component gained after decomposition to calculate IMFs. Furthermore,a de-noising method with CEEMD and the wavelet packet threshold combined was proposed to be applied in de-noising vibration signals measured from Xiangjia dam＇s hydro-elastic model,the effectiveness of this method was verified. In order to improve the precision of modal identification with the noised vibration responses,based on the data correlation technology,Markov parameters were used to construct a correlation matrix R. The R matrix was employed to reconstruct a Hankel matrix. After its SVD,the minimum system realization was obtained. This algorithm was called the eigen-system realization algorithm with data correlation（ ERA/DC）. Finally, the proposed de-noising and modal identification combined method was applied in the measured vibration responses of Jinping arch dam to get a better application effect.

关 键 词：泄流结构 滤波降噪 完备总体经验模态分解 小波包 模态识别 相关特征系统实现法 

分 类 号：TV65[水利工程—水利水电工程]