基于SVM与多振动信息融合的齿轮故障诊断

Gear fault diagnosis based on SVM and multi-sensor information fusion

针对齿轮振动信号故障特征微弱以及单个传感器故障诊断可靠性与准确性低等问题,采用多传感器信息融合方法,利用支持向量机(SVM)对8路齿轮振动信号进行特征级融合,实现故障诊断。研究结果表明:基于多个传感器单个特征量信息融合的齿轮故障诊断率比常规的基于单个传感的多个特征量的诊断准确率更高,诊断结果更可靠;峰值因子对齿轮故障最敏感,以峰值因子为特征量的多传感器信息融合,诊断准确率达93.33%。

To solve the problems that the vibration signals from a gearbox are usually noisy and it is difficult to find a potential failure in a gearbox by a single sensor,using support vector machine（SVM） as a tool for feature-level information fusion,eight gear vibration signals for fault diagnosis were investigated.The results show that the method for gear fault diagnosis based on multi-sensors information fusion has higher reliability and accuracy than that based on a single sensor.The crest factor is the most sensitive character for gear failure and the diagnostic accuracy rate reaches 93.33% by using the character to perform multi-sensor information fusion.