摘要
针对滚动轴承故障振动信号的非平稳特征和传统包络分析法的缺陷,提出了一种基于经验模态分解包络谱的滚动轴承故障诊断方法。该方法首先采用经验模态分解将原始信号分解为若干个平稳的固有模态函数之和,然后求出包含主要故障信息的若干个固有模态函数分量的包络谱,再定义包络谱中故障特征频率处的幅值比为特征幅值比,最后以特征幅值比作为故障特征向量,输入神经网络,以神经网络的输出来判断滚动轴承的工作状态和故障类型。对滚动轴承内国、外国故障振动信号的分析结果表明,基于经验模态分解包络谱的故障诊断方法能有效地提取滚动轴承的故障特征。
According to the non-stationary characteristics of the roller bearing fault vibration signals and the limitation of the traditional envelope analysis, a fault diagnosis approach based on EMD (empirical mode decomposition) and envelope spectrum was proposed. First, the original signals were decomposed into a number of IMFs (intrinsic mode functions): then, the ratios of amplitudes in the characteristic frequencies were defined as the characteristic amplitude ratios after the envelope spectra of some IMFs including the main fault information were obtained; finally, the characteristic amplitude ratios were served as the fault characteristic vectors to be put into the neural network and the work conditions and fault patterns were identified by the output of the neural network. The analysis results from roller bearing signals with inner-race or out-race fault show that the fault diagnosis approach based on EMD and envelope spectrum can extract roller bearing fault characteristics effectively.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2004年第16期1469-1471,共3页
China Mechanical Engineering
基金
国家自然科学基金(50275050)
高等学校博士学科点专项科研基金(20020532024)
关键词
滚动轴承
经验模态分解(EMD)
包络谱
特征幅值比
神经网络
故障诊断
roller bearing
EMD(empirical mode decomposition)
envelope spectrum
characteristic amplitude ratio
neural network
fault diagnosis