基于EMD和ARMA模型桥梁振动信号降噪的处理方法  

Noise reduction of bridge vibration signal based on EMD and ARMA model

作　　者：刘建军[1] 黄方林[1] 王学敏[1] 

机构地区：[1]中南大学土木建筑学院,湖南长沙410075

出　　处：《铁道科学与工程学报》2006年第5期55-59,共5页Journal of Railway Science and Engineering

基　　金：铁道部科研基金资助项目(2001G025)

摘　　要：在桥梁振动信号的采集和传输过程中,针对外界环境的影响可能会在信号中形成局部强噪声干扰,从而造成分析结果的失真以及由于桥梁振动信号通常具有较宽的频谱成分,致使传统的滤波降噪方法存在很大的局限性等问题,基于经验模态分解(EMD)和自回归滑动平均(ARMA)模型提出了一种信号降噪方法。首先,利用EMD把有强噪声干扰的信号分解成不同时间尺度的本征模函数(IMF)和残余项;然后,分别对每个IMF无干扰区段建立ARMA模型,利用各个模型对有干扰区段进行滤波,用滤波后的数据代替原来的数据,对于残余项,拟合为多项式;最后,将所有的IMF及拟合后的残余项叠加,即得到降噪后的信号。通过对实测南京长江大桥有对讲机干扰的应变信号进行分析,结果表明了该方法的可行性及有效性。During the process of acquisition and transmission of bridge vibration signals, environmental influence may cause local strong noise in signals, consequently leading to analysis results distortion. The vibration signals of bridges usually have wider frequency spectrum, which leads to the traditional filter to invalidation. In order to solve this problem, a method based on the empirical mode decomposition （EMD） and auto - regressive moving average （ARMA） model was introduced to reduce noise from signals. The first step was to decompose the signal with local strong noise into several intrinsic mode functions （IMF） of different time scale and residue by EMD. Then, the ARMA models were built respectively to each IMF according to the section without noise, filtering the sections with noise using the corresponding models and replacing the filtered data for the original ones. The residue was fitted as a polynomial. Finally, a new signal with the strong noise reduced was reconstructed by superposing all the IMF and the fitted residue. The resuits show the feasibility and validity of the method.

关 键 词：经验模态分解 自回归滑动平均模型 本征模函数 降噪 

分 类 号：U448.121[建筑科学—桥梁与隧道工程]