基于小波熵的滚动轴承早期微弱故障信息提取

Early Weak Fault Information Extraction of Rolling Bearing Based on CWT-entropy

滚动轴承产生早期微弱故障时,故障信息会被淹没在强背景噪声和其他振动源信号中,还会受到低频效应影响,这使得传统的频谱分析很难找到这些被淹没的故障信息。针对这一问题,本文提出一种基于小波熵的故障信息提取技术,首先对测试到的振动信号进行连续小波变换(CWT),获得时间-尺度谱,再计算时频矩阵中每一个尺度下的信号的熵,最后选取熵最小的尺度进行频谱分析。运用该方法对设置了外环故障、内环故障、滚珠故障的三种滚动轴承的振动信号进行了分析,并与传统的傅里叶变换(FFT)和包络解调分析方法进行了对比。分析结果表明,基于小波熵的分析方法能更有效地提取出振动信号中的故障频率信息。最后把该方法应用到某型涡轴发动机的主轴承故障诊断中,成功提取出了故障频率信息,实现了对滚动轴承进行早期故障检测。

When the early failure was generated in rolling bearings, the fault information was submerged in strong background noises and other vibration signals, and also affected by the low frequency effect. It is difficult to find these fault information by traditional spectral analysis methods. This paper presents a new method of extracting weak fault information called wavelet entropy for this problem. Firstly, the measured vibration signalis processed with Continuous Wavelet Transform(CWT) to obtain the time-scalematrix. Secondly, the entropy of the signal under each scalewas computed. And then, the scale with the minimal entropy value was selected for spectrum analysis. This method was applied to analyzing the vibration signals of bearings under three kinds of typical fault(with damage on outer race,inner race and ball respectively), and it is found that the wavelet entropy is more effective in extracting the periodic impulses features produced by localized bearing damage than FFT and the traditional envelope demodulation analysis.