基于高阶统计量自适应滤波的故障特征的提取  

A Method for Extracting Mechanical Faults Features Based On Higher-Order Statistics

作　　者：尚万峰[1] 关惠玲[1] 李志军[1] 

机构地区：[1]郑州大学机械工程学院,河南郑州450002

出　　处：《信号处理》2005年第5期544-547,459,共5页Journal of Signal Processing

摘　　要：本文论述了基于三阶累积量的RLS自适应算法(CDRLS)和基于累积量韵LMS自适应算法(CDEFWLMS)。通过对机车滚动轴承的保持架断裂典型故障信号的分析得出高阶统计量自适应算法具有良好的降噪性,CDEFWLMS算法和CDRLS算法比普通的LMS算法和RLS算法收敛速度快,稳定性好;同时得出CDEFWLMS算法比CDRLS算法更优,经处理后的特征频率更突出明显,实际中有较高的正确诊断率。研究表明:高阶统计量自适应滤波提取信号特征,可以容易的将正常的轴承信号和保持架断裂故障信号分离,从而进一步验证其在实际故障诊断、检测中的有着良好的应用特性。Both adaptive CDRLS filter based third-order cumulants and adaptive CDEFWLMS filter based third-order cumulants apply to fault of diagnosis of rolling bearing. Analyse these typical faults of the locomotive freight car roller bearing such as bearing cage broken, by which the result is that adaptive fliter based third-order cumulants has nice qualities, and CDEFWLMS filter is more effective than CDRLS filter in practice. The results of the research show that normal gear signals,broken bear signals can be easily separated by using higher-order statistics as the signal featuros,which validates adaptive filter based third-order cumulants in the field of fault diagnosis and monitoring.

关 键 词：自适应算法 RLS算法 LMS算法 故障诊断 高阶统计量 自适应滤波 故障特征 LMS自适应算法 提取 故障信号分离 

分 类 号：TN911.7[电子电信—通信与信息系统] TN713[电子电信—信息与通信工程]