提出了一种基于一类支持向量机的传感器故障诊断方法,它利用相空间重构技术将传感器本身获取的相关时序数据转化成矢量,再用一类支持向量机来训练正常行为模型,进而通过发现数据异常来进行故障诊断.通过连续输入测量数据,它还可以实现在线诊断.仿真实验表明:该方法在诊断传感器故障方面有良好的性能.
A method of fault-diagnoses for sensor based on one-class SVM is put forward,which transforms the timeseries data from sensor into vector by means of the technology about reproduction in phase-space ,and trains normalaction model with one-class SVM and implements fault-diagnoses by finding abnormity data.The online-diagnoses can be implemented by continuously importing measuring data.The simulation-experiments show,it has nicer performance in sensor fault-diagnoses.