中文摘要：

局部均值分解（Local mean decomposition,简称LMD）方法通过滑动平均方法平滑局部均值线段和局部幅值线段得到局部均值函数和包络估计函数,从而实现信号的分解。但滑动平均方法会产生相位差以及平滑步长的选择具有一定的主观性,这样会使分解结果不理想。对LMD方法进行了改进,采用有理样条插值函数（Rationalspline）求取信号的上下包络线,然后通过上下包络线计算信号的局部均值函数和包络估计函数,克服了原LMD方法中采用滑动平均方法带来的缺陷。通过对仿真信号以及滚动轴承故障振动信号的分析,表明改进后的LMD方法优于原LMD方法。

英文摘要：

In the local mean decomposition（LMD） method,in order to obtain the local mean and envelope estimate function,the local means and the local amplitude lines are smoothed by using moving average method,and thus the original signal can be decomposed.However,phase difference would occur when moving average method are applied and the selection of the moving average length would inevitably bring subjective influence to the analysis results,which would influence the accuracy of the decomposition results.In this paper,in order to overcome the disadvantages mentioned above,a suggested improvement to the original LMD algorithm has been made,namely,calculating the upper and lower envelopes of the signal by the rational splines function,from which the local mean function and envelope estimate function can be got.The analysis results from simulation signal and practical roller bearings fault vibration signal demonstrate that the improved LMD method is superior to the original LMD approach.