脉冲噪声环境下机械故障信号的盲分离  

Time-frequency blind source separation for mechanical fault signals under impulse distributed noise condition

作　　者：杨保海[1] 余香梅 舒彤[1] 

机构地区：[1]九江学院电子工程学院,九江332005 [2]江西省数控技术与应用重点实验室,九江332005

出　　处：《电子测量与仪器学报》2016年第3期440-447,共8页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61261046);江西省自然基金(20142BAB207006;20151BAB207013);江西省教育厅科技基金(GJJ14739;GJJ14721);九江学院校级科研项目(2013KJ02;2013KJ01)资助

摘　　要：针对振动传感器在采集故障信号时,在α稳定分布脉冲噪声的干扰下,使得传统机械故障信号时频盲源分离算法性能退化的问题,提出了一种基于分数低阶和S时频变换的盲源分离新方法。该方法先对传感器测试信号进行分数低阶子空间预白化,再计算低阶化信号的S变换时频分布,最后通过联合近似对角化恢复各个部分的故障源信号。通过计算机仿真实例分析表明,该算法能有效抑制脉冲噪声影响,避免了二阶矩或高阶矩无穷大的缺限,盲源分离效果较好,具有良好的鲁棒性。The impulsive noise of α-stable distribution is characterized by the nonexistence of the finite second order or higher statistics. The blind source separation based on time-frequency distribution（ TFD-BSS） method was poor invalid under α-stable distributed noise conditions. An improved fractional lower order statistics time-frequency distribution blind source separation algorithm was proposed in this paper. First,the signals were pre-whitening based on fractional lower order statistics and subspace technique,and then the fractional lower order time-frequency distribution of generalized s-transform was computed. Finally,the source signals were obtained by joint approximate digitalization of Eigen-matrices. The simulation results analysis shows that the proposed method is more robust in α-stable distribution interference environments than that of the conventional second order statistics based algorithm.Moreover,the decision overcomes the shortcoming of the second and higher order moment infinity for BSS.

关 键 词：Α稳定分布 时频分析 S变换 盲源分离 预白化 分数低阶统计量 

分 类 号：TN911.7[电子电信—通信与信息系统]