ART-2对滚动轴承振动信号幅值谱的清晰化提取与分类效果研究

滚动轴承振动信号的幅值谱往往在整个频带上都存在着均匀分布的噪声频率干扰，难以用一般的滤波方法去除，这对轴承实际振动频率成分的识别有很大的影响。通过对ART-2神经网络的分析，指出了在信号中的特征信息相对较强，而噪声相对较弱的情况下，利用ART-2可以对信号中的特征信息成分进行增强，并对噪声进行充分抑制。利用这一原理，实现了对轴承振动信号幅值谱的清晰化提取。对13个轴承振动信号幅值谱的处理实例表明，这种方法获得的模式中轴承实际振动频率成分被清晰地提取了出来，而均布噪声频率则被充分抑制。进一步研究了ART-2的警戒系数对幅值谱清晰化处理效果的影响。并对这种影响机理进行了解释。

Study on Clear Extraction of Amplitude Spectrum of Vibrating Signal for A Rolling Bearing by ART-2 and Its Ciassification Effect

In amplitude spectrum of vibrating signal of bearings,there generally exists much disturbance of uniform distributed noise frequencies,which is difficult to be erased by common filter and has a bad influence on identification of bearing vibrating frequency.Through analysis by ART-2 principle,it's pointed out that signal feature information will be strengthened,but noise will be greatly suppressed by ART-2 under a condition that this signal feature information is superior to noise.According to this principle,the clear extraction algorithm for amplitude spectrum of bearing vibrating signal is realized.The example in which 13 bearing amplitude spectrums are processed shows that the bearing vibrating frequencies are clearly extracted and uniform distributed noise frequencies are fully suppressed in the modes formed by ART-2.Furthermore,the influence of ART-2 viglance coefficient on the clear extraction of bearing amplitude spectrum is studied and the explanation for this influence mechanism is given also.