摘要：滚动轴承是机械设备中必不可少的基础元件，它由内圈、外圈、滚动体、保持架四部分组成。滚动轴承起源很早，历史悠久。随着科学技术的不断发展，轴承制造技术也日新月异，滚动轴承的精确度也越来越高，然而轴承故障是不可避免的，比如轴承长时间工作后会造成疲劳剥落、腐蚀、磨损、擦伤、裂纹等。众所周知，轴承故障带来的危害是巨大的，包括人员伤亡、经济损失等。那么通过什么样的方法可以识别进而诊断轴承故障呢？本文通过获取的滚动轴承的振动信号，这些轴承振动信号来源于美国凯斯西储大学电气工程轴承实验。然后采用由模式滤波法原理编制而成的软件计算和分析这些振动信号，分析结果得到大量声音子的时域图，再利用时域和频谱转换关系得到频谱图，结合时域图和频谱图采用人工智能的方法对这些声音子进行分离并归类，最后分析归类结果中每一类声音子的特征来诊断轴承的故障。

关键词：振动信号  信号分离  特征提取  模式滤波法  故障诊断

Abstract: The rolling bearing is an essential foundation component in the equipment of machinery. It consists of inner ring, Outer Ring, Cage Train and Rolling Element. The rolling bearing origins early and has a long history. Along with the development of technology of science ,the technology of bearing manufacturing also changing, and the accuracy of the rolling bearing is more and more higher. However，the bearing fault is inevitable . For example , the long working time of bearings can cause fatigue peel, corrosion, wear, bruises, crack and so on. As is known to all, the bearing fault can bring a huge harm, Including the injury or the death of  person, economic loss, etc. So what kind of method can be used to identify and then diagnosis the fault of the rolling bearing? This article through the vibration signal of rolling bearings, These bearing vibration signal comes from the electrical engineering bearing experiment of the American case western reserve university. And then use a software to calculate, and analyzed the vibration signal, the software is programmed by the principle of Mode of filtering method . we can get a lot of time domain figure of the voice through the analysis results, and then use the relationship of time domain and frequency conversion to achieve frequency spectrogram. Combined with the time domain figure and frequency spectrogram using artificial intelligence method to separate and classify these voices. The finally, through analysising the feature of every kind of voice of the classified results to diagnosis the bearing fault.

Key words：Vibration signal   Signal separation  Feature extraction Mode of filtering method   Fault diagnosis