融合KFCM及频谱质心的瞬时频率估计算法研究

Research on instantaneous frequency estimation algorithm based on fused KFCM and spectral centroid

基于时频分析和峰值搜索的瞬时频率估计方法对噪声敏感,结果误差较大。为降低噪声引起的瞬时频率估计误差,提出一种融合核模糊C均值聚类(KFCM)及频谱质心的旋转机械瞬时频率估计算法。依据不同信号分量能量的差异,采用核模糊C聚类的方法,对时频矩阵中的元素点聚类分析,去除时频面上能量噪声点和孤立点,在此基础上利用频谱质心算法,提取瞬时频率,进而估计转速。采用仿真实验对该方法进行验证,误差为0.5%,证明了该算法的有效性。用该方法对旋转机械加速过程中的振动信号进行分析,取得了明显的转速估计效果,误差为1.87%。仿真算例和应用实例表明该方法能实现对旋转机械振动信号的瞬时频率估计。

The traditional instantaneous frequency estimation method based on time-frequency analysis and peak search algorithm is sensitive to noise,which leads inaccurate estimation under strong noise environment.To solve this problem,a rotating machine instantaneous frequency estimation algorithm based on kernel-based fuzzy C-means clustering(KFCM)algorithm and spectral centroid algorithm is proposed.According to the difference of energy of different signal components,this method first discriminates the time-frequency distribution points from the sources of the reference axis and others by adopting the clustering algorithm.Secondly,the discrete time and frequency points related to the instantaneous rotation frequency of the reference axis is located,on the basis of which the instantaneous rotation frequency is estimated using spectral centroid algorithm.The simulations verify that the method can realize the instantaneous rotation frequency estimation from the vibration signals of complex rotating machinery precisely,and the error of result is 0.5%.This method is applied to analyze the vibration signal of the rotating machinery in the station of run up,and the remarkable speed estimation results are obtained.And the error of result is 1.87%.Simulation and experimental results verify that the proposed method can realize the instantaneous frequency estimation of the vibration signal of the rotating machine.