摘要
针对经验模态分解(Empirical Mode Decomposition,EMD)方法在爆破振动信号应用中模态混叠的问题,提出了一种改进的变分模态分解(Variational Mode Decomposition,VMD)方法.首先,使用任意给定的模态分解个数与惩罚因子变分模态分解将爆破振动信号分解为K个模态;然后计算分量的幅值谱熵的局部最小值;其次采用混合GA-PSO算法对任意模态分解个数与惩罚因子进行全局搜索来不断优化参数并更新幅值谱熵的局部最小值,最终最小幅值谱熵与平均幅值谱趋于一致,得出全局最小的局部最小值的幅值谱熵相应的模态分解个数与惩罚因子.仿真结果表明:模态分解个数与惩罚因子作为全局最优输入交叉优化能够准确地确定模态分解个数与惩罚因子,与经验模态分解相比,改进的变分模态分解方法具有很强的鲁棒性和抗噪声干扰能力以及分解与去噪重构精度.最后利用参数优化后的VMD进行爆破振动信号的实测,通过相关系数法检验,各分量相关系数均在一个数量级上,解决了模态混叠问题并且无虚假分量,证明参数优化的VMD在爆破振动信号应用上具有很好的适应性.
Aiming at the problem of mode aliasing in the application of empirical mode decomposition(EMD)to blasting vibration signal,an improved variable mode decomposition(VMD)method was proposed.Firstly,the blasting vibration signal was decomposed into K modes by using any given number of modal decomposition and penalty factor variational modal decomposition;then the local minimum value of amplitude spectral entropy of components was calculated;secondly,the global search of the number of arbitrary modal decomposition and penalty factor was carried out by using hybrid GA-PSO algorithm to optimize the parameters and update the local minimum value of amplitude spectral entropy,and the final minimum amplitude spectral entropy was obtained and tended to coincide with average amplitude spectral.The number of modal decomposition and penalty factor were obtained according to the entropy of the global minimum and local minimum.The simulation results show that:because the number of modal decomposition and penalty factor can be accurately determined as the global optimal input cross optimization,compared with the empirical mode decomposition,the improved variational mode decomposition method has strong robustness,anti noise interference ability,decomposition and de-noising reconstruction accuracy.Finally,the VMD optimized by parameters is applied to the measured blasting vibration signals.Through the correlation coefficient method,the correlation coefficients of each component are all in an order of magnitude,which solves the problem of modal aliasing and has no false components.The VMD optimized by parameters has good adaptability in the application of blasting vibration signals.
作者
杨宗林
熊继军
YANG Zong-lin;XIONG Ji-jun(Key Laboratory of Instrumentation Science and Dynamic Measurement(North University of China),Ministry of Education,Taiyuan 030051,China;Science and Technology on Electronic Test and Measurement Laboratory,North University of China,Taiyuan 030051,China)
出处
《中北大学学报（自然科学版）》
CAS
2020年第5期467-473,共7页
Journal of North University of China(Natural Science Edition)
基金
国家杰出青年科学基金资助项目(51425505)。
关键词
幅值谱熵
变分模态分解
混合GA-PSO算法
爆炸振动信号分解
amplitude spectral entropy
variational modal decomposition
hybrid GA-PSO algorithm
explosion vibration signal decomposition
作者简介
杨宗林(1993-),男,硕士生,主要从事测量数据处理与仿真研究;通讯作者:熊继军(1971-),男,教授,博士生导师,主要从事无线无源微纳传感器以及系统集成技术、高温压力传感器等方面研究。
