盲解卷积的机械振动信号分离技术  

Mechanical Vibration Signal Separation Using Blind De-Convolution Technology

作　　者：刘婷婷[1,2] 任兴民[1] 杨永锋[1] 郭峰[1] 

机构地区：[1]西北工业大学振动工程研究所,西安710072 [2]第二炮兵指挥学院工程保障研究室,武汉430012

出　　处：《振动．测试与诊断》2009年第4期419-423,共5页Journal of Vibration,Measurement & Diagnosis

基　　金：中国博士后科学基金资助项目(编号:200902605)

摘　　要：针对机械振动信号往往是多个信号卷积混合的结果,阐述了卷积混合的模型和原理。利用扩展的H-J网络结构,给出了在线实时的盲解卷积迭代算法,并通过仿真试验验证了算法的有效性和准确性。该法与传统的傅里叶变换频谱分析相比,能获得更多的振源振动信息,可更准确地进行机械故障诊断。The application of blind source separation (BSS) to the mechanical vibration signal processing provides a new technique for mechanical fault diagnosis. Mechanical vibration signals in practice can be viewed as sums of differently convolved source. For this characteristic,based on the convolution model and basic theories of blind de-convolution (BD),a BD method on-line real time processing technique using the extended H-J network framework was proposed. The effectiveness and accuracy of the BD algorithm was verified by the numerical simulation data. Applied to the actually measuring data,the method provides more information about the vibration and diagnoses faults more accurately than the conventional Fourier transformation techniques.

关 键 词：盲源分离 盲解卷积 机械振动信号 故障诊断 

分 类 号：TH133[机械工程—机械制造及自动化] TN911[电子电信—通信与信息系统]