摘要
在机械设备盲信号处理和故障诊断中,信号的非高斯性至关重要,而峭度是非高斯性的自然度量指标,它反映了机械信号信息的动态变化特征。基于此提出了基于峭度的ICA特征提取和故障诊断方法,首先提取机械设备多通道观测信号的峭度值(或标准峭度,或峭度绝对值,或峭度平方),并依据观测通道顺序将峭度值组成低维ICA特征向量,进而利用最小二乘支持向量机进行机械设备的模式判别和故障诊断。试验表明:该方法的故障识别率基本上达到80%以上,而且相比标准峭度和峭度平方准则,基于峭度绝对值准则方法的模式识别率更高。
Mechanical signals′ non-gaussian property is very important to blind signal processing and fault diagnosis of mechanical equipment,and the kurtosis is a natural measurement for non-gaussian property of random variables.It denotes the dynamic information features of mechanical signals.ICA（independent component analysis） feature extraction and fault diagnosis algorithm based on kurtosis was proposed.This algorithm is composed of three steps：（1） to Abstract multi-channel observation signals′ kurtosis values（standard kurtosis values,or absolute kurtosis values,or square kurtosis values）;（2） to form an ICA eigenvector from kurtosis values according to sequence of multi-channel observations;（3） to recognize different working patterns and diagnose different faults by least squares support vector machines.Its applications in hydraulic gear pumps′ feature extraction and fault diagnosis show that its fault diagnosis rate reaches 80% on the whole.Furthermore,compared with standard kurtosis value and square kurtosis value,absolute kurtosis value is a more appropriate choice.
出处
《机械科学与技术》
CSCD
北大核心
2011年第9期1583-1587,共5页
Mechanical Science and Technology for Aerospace Engineering
基金
总装备部预研重点基金项目资助
关键词
独立成分分析
峭度
最小二乘支持向量机
特征提取
故障诊断
independent component analysis（ICA）
kurtosis
least squares support vector machines
feature Abstraction
fault diagnosis
