基于特征选择的支持向量机在故障诊断中的应用

机械设备因为其本身结构的复杂性,故障很难简单地进行诊断,所以智能诊断成为一个热点的研究方向。以前的工作中多是通过神经网络甚至支持向量机等方法进行诊断,本文提出了基于支持向量机集成的特征选择算法,通过该算法可以有效去除故障数据集中所提取的不相关特征,并在新的更少特征的数据集上进行建模。在实际某柴油机故障数据上的计算表明:在通过特征选择后的数据集上利用支持向量机集成的方法建模可以得到比不进行选择更好的结果,也得到了比单个支持向量机建模更好的结果。

Fault Diagnosis Using Support Vector Machines Based on Feature Selection

Because of the complexity of mechanical structure,the diagnosis of mechanical structure is not very easy,intelligent diagnosis has become one focus of the fault diagnosis area.The previous work mostly uses the neural network or the SVM to diagnose.This paper introduces the SVM algorithm based feature selection.This algorithm can discard the irrelevant features effectively.Then the new model is trained.The process of training and testing on one set of some diesel engine indicates that the SVM algorithm based feature selection can get the better results than the algorithm not using feature selection.