早期故障信号基于调制随机共振的核主元检测  

Early Stage Fault Detection Based on Modulated Stochastic Resonance and Kernel Principal Components

作　　者：董华玉[1] 李金寿[1] 翟链[1] 

机构地区：[1]镇江船艇学院装备保障系,江苏镇江212003

出　　处：《昆明理工大学学报（自然科学版）》2014年第4期55-62,共8页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：总装备部通用装备保障部课题

摘　　要：针对淹没在强噪声中具有高频多周期成分的早期故障信号,提出了并联调制随机共振与核主元分析系统.将信号的特征频率通过调制共振系统进行增强,并将功率谱组成多元统计向量,通过非线性映射,在高维空间进行特征提取,利用T2统计量和F分布的控制限进行检测.利用仿真的轴承转子振动信号与滚动轴承冲击信号进行了验证,实验表明系统能够及时地对早期故障进行预警.Aiming at the detection of early stage fault which have multiple high frequency composition,a system based on parallel stochastic resonance and kernel principal components is proposed. The characteristic frequencies are strengthened by modulated stochastic resonance.The power spectrums of several frequencies are formed into a multivariant statistical vector.By nonlinear mapping,the statistical vector’s feature is extracted in higher dimension space.The T2 -statistic and control limit of F-distribution are utilized to detect early stage faulty signal.The simulated bearing rotor vibration signal and ball bearing impulse signal are used to verify the proposed method.The experiment results show that the system can alert the early stage fault promptly.

关 键 词：早期故障检测 核主元分析 调制随机共振 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]