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

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.