基于DS证据理论的SVM分类模糊域数据修正

在介绍支持向量机(SVM)和DS证据理论的基础上,提出了一种利用DS证据理论对SVM分类模糊域数据进行分类修正的方法。该方法首先利用SVM对测试样本进行分类,对SVM分类输出模糊域的样本使用隶属度函数将SVM的输出距离转换成样本对各状态的隶属度;其次利用DS证据理论融合其他传感器信息,对各状态下的隶属度进行适度修正,从而实现该区域数据的重新合理排布;最后将该方法应用于高压断路器故障诊断,以验证其诊断性能。大量的实验结果表明,该方法可以利用断路器操作线圈电流数据,合理修正振动数据分类结果,实现断路器机械故障的准确检测。

Modification of SVM classification fuzzy area based on DS evidence theory

SVM(Support Vector Machine) and DS evidence theory are introduced and the method to modify the data of SVM fuzzy classification areas is proposed based on DS evidence theory.The test samples are classified by SVM and the distance of its output fuzzy area samples is transformed into their membership grades to each state by the membership function.The diagnosis information of other sensors is fused together based on the DS evidence theory to modify the sample membership grades to each state and the data of this area is redistributed.The proposed method is applied in the fault diagnosis of high voltage circuit breakers for verifying its diagnostic performance.Experimental results show that,it uses the data of coil current to modify the results of vibration data classification to realize the accurate detection of mechanical fault.