According to the real working status of running reciprocating machines,a grey prediction method of diagnosis based on local wave time-frequency spectrum and SVM was presented. By local wave time-frequency spectrum, power fluctuate of the vibration signals could be reflected on time and frequency domain. Comparing the precision of the two main analysis models in diagnosis prediction on industrial equipment, a step-changing grey prediction model was chosen, in which the local wave decomposition was set as the basic input. Application of the method in fault prediction on a reciprocating compressor system illustrates the precision improvement.

local wave method, support vector machine(SVM), fault diagnosis, trend prediction, reciprocating machinery

针对往复机械系统工况的动态特性,提出了一种基于非平稳振动信号局域波分析和支持向量机（SVM）的故障预测方法。对于往复机械的振动监测信号，利用局域波法获得其中所包含的特征信息，以此作为预测模型的数据源;采用

SVM作为预测手段，将局域波时频谱中所包含的局域波分量特征信息作为预测控制模型的输入量。该方法应用于工程实践中，有效地提高了预测精度，并为设备的工况和剩余寿命定位提供了依据。

局域波法, 支持向量机, 故障诊断, 趋势预测, 往复机械