摘要
提出了一种采用最小二乘支持矢量机构造异步电机转子多故障分类模型的方法。首先通过对采样的定子电流进行快速傅里叶变换,所得频谱经一致化处理后作为支持矢量机的输入参数,然后利用1对1策略构造了转子多故障分类器,经训练后可以对四种不同转子故障进行识别。文中还分析了惩罚因子、核函数和子分类器输出融合策略对分类准确性的影响,指出高斯径向基函数和混合矩阵融合策略可以提高诊断精度。实验结果表明,该模型具有很好的分类精度和泛化能力。
A multi-class classification model for induction machine rotor fault diagnosis based on LSSVM （Least Square Support Vector Machine） is proposed. At first, the signals of the sample stator current are analyzed with FFT, and the normalized spectral characteristics are used as the inputs of SVMs, then the multi-class LSSVM classifiers are constructed according to the one-against-one strategy, finally, the four rotor fault types of induction machine are identified by the trained classifiers. The influences of some factors such as the error penalty, kernel function, and the coupling schemes of sub classifiers, on the accuracy, are also discussed. It is pointed out that the model can be improved by adopting RBF kernel function and mixture matrix coupling scheme. The experimental results show that the proposed approach has good accuracy and generalization performances.
出处
《电工技术学报》
EI
CSCD
北大核心
2006年第5期92-98,103,共8页
Transactions of China Electrotechnical Society
基金
国家自然科学基金(50477010)
福建省青年科技人才创新基金(2004J032)资助项目。
关键词
最小二乘支持矢量机
异步电机
转子故障诊断
分类
Least square support vector, induction machines, rotor fault, classification
