宽带耦合机械振源的卷积盲分离研究  

Study on Convolutive Blind Separation of Broadband Coupling Mechanical Vibration Sources

作　　者：江涌[1,2] 李军[1] 章林柯[1] 何琳[2] 

机构地区：[1]防化研究院装备评价与计量研究室,北京102205 [2]海军工程大学振动与噪声研究所,武汉430033

出　　处：《武汉理工大学学报（交通科学与工程版）》2011年第3期496-499,504,共5页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：国家自然科学基金项目(批准号:50775218);"十一五"国防预研项目(批准号:51310030201)资助

摘　　要：利用卷积盲分离算法,对宽带耦合机械振源信号进行了盲分离研究.在实际环境中,由于相邻设备和环境干扰的影响,传感器往往只能获得混合信号.且源信号在频谱上可能相互重叠,此时,传统的信号滤波、净化方法就无能为力.文中利用盲信号处理技术对源和传递路径先验知识要求较少的特点,仅从混合信号出发,利用信号的二阶统计特性,较好地恢复出了频谱上相互重叠的机械振源信号.电机和海水泵的仿真试验验证了该方法的有效性.With the convolutive blind separation algorithm,the broadband coupling mechanical vibration sources were separated in this paper.In practical conditions,the mixed signals can only be measured by the sensors due to the influence of neighboring equipments.And,source signals may be overlapping in spectrum,so the traditional method of signal filtering is ineffective to separate.Employing the blind signal processing technology,this paper proposed a separate method.With its second order statistic characteristic and little requirements of pre-knowledge of sources and transfer paths,the mechanical vibration source signals of spectrum overlapping were successfully recovered only from the mixed signals.Finally,the feasibility of this method was testified by the simulation experiment of electromotor and seawater pump.

关 键 词：卷积盲分离 频谱重叠 源识别 耦合 

分 类 号：TN912.3[电子电信—通信与信息系统]