旋转机械振动的频域多维时序预报

THE MULTI-DIMENSIONAL PREDICTION MODEL IN FREQUENCY DOMAIN OF ROTATING MACHINERY VIBRATION BY TIME SERIES METHOD

旋转机械是化工、发电大型企业的关键设备，为减少因定期维修和突发故障造成的经济损失，需在实时状态监测的条件下进行状态预报．目前一般通过对机组振动的时域信号预报机组的运行状态，但时域信号与机组运行状态之间的本质特征关系并不直观和确切，同时在旋转机械的转动速度有波动时该方法并不有效．本文在考虑机组转速有波动的前提下，对振动频谱进行规范化处理，并根据相应的的数学力学分析。把振动频谱作为多维时间序列进行机械状态预报．

Currently, the on-line running conditions of rotating machinery is predicted mainly through the vibrational signal of machinery in time domain. In this paper,the vibrational spectrum of rotating machinery is standardized to eliminate the fluctuation of rotating speed of machinery. The vibration of rotating machinery is classified into two categories according to the characteristics of its frequency. By simplifications of mathematics and mechanics, the corresponding spectrum is assumed as a multi- dimensional time series to predict the running condition of rotating machinery.