摘要
实测差异分析表明,速度振幅峰值分布受外界因素影响更具规律性,加速度信号主频分布比速度信号更分散。提出一种加速度信号时域积分算法,先基于经验模态分解法对信号进行趋势项剔除、去均值化和高频降噪,再对存在漂移现象的各固有模态函数分量时域积分速度分量进行最小二乘法处理。研究表明:各固有模态函数分量在处理过程中应分别对待;为达到较好降噪效果,阈值修正系数k应不断调整;分段最小二乘法对漂移现象的消除能力优于单段最小二乘法和经验模态分解法。最后,通过本文定义的全局参量和局部参量验证算法的优越性。
The analysis of measured differences shows that the distribution of peak values of vibration velocity is more regular than that of acceleration under the influence of external factors,and main frequency distribution of vibration acceleration signals is more scattered.A time integration algorithm for acceleration signal was put forward： based on empirical mode decomposition,the signal was processed by residual elimination,mean remove and high-frequency de-noising and velocity components in time integration during which exist drift phenomena were revised by least square method.The research shows： IMF components should be treated separately;to achieve better de-noising effect,threshold verified coefficient k should be adjusted;segment least square is superior to least square in single segment and empirical mode decomposition regarding the ability of eliminating drift phenomena.The advantages of the algorithm were verified through new global and local parameters defined in the paper.
出处
《振动与冲击》
EI
CSCD
北大核心
2012年第9期90-93,101,共5页
Journal of Vibration and Shock
关键词
经验模态分解
分段最小二乘法
时域积分
评价参量
empirical mode decomposition（EMD）
segment least square（SLS）
time integration
evaluation parameters
