摘要
将模糊域分布和支持向量机相结合,提出了一种故障诊断的新方法,该方法将模糊域分布中的局部能量作为特征输入到支持向量机的多故障分类器进行故障识别.利用模糊域分布可以很好地刻画信号的时频局部化特征,与时-频平面特征提取相比,又可大大降低数据维数.对于不同类型的核函数分布,将其诊断结果进行比较,试验结果表明,基于模糊域的支持向量机故障分类无需核函数滤波就能取得最好的分类效果.
By combining ambiguity domain with support vector machine(SVM), a now method of fault diagnosis is presented. The proposed method used the local energy in ambiguity domain as a feature vector to input the SVM classifier to identify faults. The local information of signal can be ftdly reflected by using ambiguity domain distribution. Compared with the feature extracted from the time-frequency plane, the dimensions of feature vector can be greatly reduced. The recognition results are analyzed for different kernel functions. The experiment results show that the best classified efficiency can be obtained without any kernel function in ambiguity domain.
出处
《山东大学学报(工学版)》
CAS
2006年第6期116-120,共5页
Journal of Shandong University(Engineering Science)
基金
河南省杰出人才创新基金(No.0621000500)
国家自然科学基金(No.50675209)
河南省教育厅自然科学基金资助项目(No.2006460005)
关键词
模糊域
支持向量机
核函数
故障诊断
ambiguity domain
support vector machine(SVM)
kernel function
fault diagnosis