基于改进LMD的大坝变形特征提取分析

The Analysis of the Dam Deformation Feature Extraction Base on an Improved LMD

局域均值分解(LMD)方法采用滑动平均法不断平滑由极值点构成的局部均值线段和局部幅值线段,从而获得连续光滑的局部均值函数和包络估计函数。其中平滑步长具有主观性,而且会影响运算时间及计算效率。进而,采用一种新的LMD方法,确定并采用所有相邻极值点之间距离大小的平均值的1/3作为滑动平均步长,建立改进-LMD分解方法。最后,将多项式插值-LMD、3次样条插值-LMD(CBI-LMD)、分段3次插值-LMD(CHI-LMD)以及传统-LMD多种插值函数分解方法进行对比分析,通过仿真信号与丰满大坝变形数据实验结果分析表明,改进-LMD分解方法的正交指数较小,且完备性较好,能有效提取大坝变形信号特征。

The local mean decomposition（ LMD） uses the moving average method. Firstly,in order to get the local mean amplitude segment and local segment,the extreme point was adopted,the local amplitude continuous smooth function and envelope estimation function were obtained by sliding the local mean amplitude segment and local segment. Where smooth sliding step is subjective,and it will also affect the computation time and computational efficiency. Furthermore,we use a compromise approach. First,confirming the size of all the distance between adjacent extrema,and then determining the size of the third value on the average of all distances,next to moving average value as the step size,and finally creating a new and improving-LMD decomposition method. Finally,the polynomial interpolation-LMD,cubic spline interpolation-LMD（ CBI-LMD）,piecewise cubic interpolation-LMD（ CHI-LMD） as well as a variety of traditional-LMD interpolation function decomposition,these were analyzed in several ways. It Shows that the improved method of orthogonal decomposition-LMD index is smaller and better equipped,this method just through simulation data signal and plump dam deformation analysis of the experimental results,and the deformation characteristics are extracted effectively.