中文摘要：

针对齿轮箱升降速过程中振动信号非平稳的特点，将计算阶次跟踪方法与经验模态分解技术相结合，提出一种研究旋转机械瞬态信号故障诊断的分析方法。首先对齿轮箱启动时测得的振动信号进行时域采样，再对时域信号进行等角度重采样，将其转化为角域准平稳信号，然后对角域里的信号进行经验模态分解得到多个固有模态函数分量，最后对包含轴承故障信息的高频固有模态分量进行包络解调分析。结果显示：阶次跟踪技术能够有效地避免传统频谱方法所无法解决的“频率模糊”现象，将非平稳信号转化为准平稳信号：经验模态分解方法能够提取包含故障信息的固有模态分量，将两种方法相结合是对传统频谱分析法的有力补充，具有很广阔的应用前景。

英文摘要：

In order to process the non-stationary vibration signals such as speed up or speed down signals effectively, a new method combined computed order tracking analysis with empirical mode decomposition technique to analyze the instantaneously signals in the rotate mechanism is explored. Firstly, the vibration signals at start-up in the gearbox are sampled at constant time increments in time-domain and then the data are resampled with software at constant angle increments in angle-domain. Therefore, the time domain instantaneously signals is changed into angle domain stationary ones. Then, many IMF components are obtained to decompose the signals in angle domain with empirical mode decomposition （EMD）. At last, the high frequency IMF component, which contains the fault information of bearing, is analyzed with order envelope spectrum analysis. The outcome shows it can avoid effectively the ＂frequency smear＂ phenomenon with order tracking technology, which cannot be solved with the traditional frequency spectrum method and the non-stationary signals are translated into stationary signals. The IMF component, which contains the fault information, can be extracted with EMD. It is trenchancy supplement for traditional spectrum analysis combined the order envelope spectrum analysis with EMD, and has more amplitude appliance foreground.