摘要
针对模拟电路的故障诊断和定位问题,提出了一种改进支持向量机(Support Vector Machine,SVM)故障诊断方法。通过在标准SVM中加入了对数据流形局部分布的约束,设计了一种依赖于数据分布的新型SVM。相对于标准SVM方法而言,新方法有效融合了数据分布的先验信息,提高了模型的诊断精度。将其用于模拟电路的故障诊断,验证了所提方法的有效性。
Focusing on the problem of analog circuit fault diagnosis and location, an improved support vector machine （SVM） was proposed for fault diagnosis. A new type of SVM lying on data distribution was designed by joining the con- straint of distribution of the data manifold. Compared to standard SVM, the proposed method effectively combined prior dis- tribution information of the data to increase the diagnosis accuracy of the model. The simulation results showed the effec- tiveness of the algorithm.
出处
《海军航空工程学院学报》
2014年第2期117-121,共5页
Journal of Naval Aeronautical and Astronautical University
基金
国家自然科学基金资助项目(61203168)
关键词
故障诊断
模拟电路
支持向量机
数据流形
fault diagnosis
analog circuit
support vector machine
data manifold
