基于支持向量机的液压泵在线故障预警

为了实现液压泵自动故障预警，提出了一种基于在线单类支持向量机的新方法.与离线单类支持向量机不同，该方法可根据输入样本的变化不断地及时调整自由参数，实现持续学习.同时提出了一种在线检测奇异值的鲁棒性算法.最后，从液压泵振动信号的时域信息中提取诊断特征参数，组成最小诊断参数组合，建立了液压泵在线故障预警系统，并进行了仿真研究.

Online fault early warning for hydraulic pump based on support vector machine

In order to realize hydraulic pump's auto fault early warning,a new method based on online one class support vector machines was presented.Contrasting to offline one class support vector machines,the model-free parameters obtained by using this method were refreshed as online monitoring data and the method implements continuous learning.At the same time,an algorithm aimed at detecting online abnormal events was designed so as to be more robust.Finally,five time-domain features of pump vibration as minimum combination of diagnosis parameters were extracted;an online fault early warning for hydraulic pump was established,and simulations were performed.