基于支持向量机的混合电路故障诊断  

Mixed circuit fault diagnosis based on SVM

作　　者：李桥敏[1] 王怀龙[1] 朱善林[1] 

机构地区：[1]海军工程大学电子工程学院电子工程系,湖北武汉430033

出　　处：《现代电子技术》2013年第5期131-134,共4页Modern Electronics Technique

摘　　要：针对混合电路故障检测率低、诊断速度慢等问题,采用一种全新的诊断方法来应对这些故障诊断技术发展中的瓶颈问题。通过提取有效的故障特征和选取不同的核函数,支持向量机(SVM)能够很好的改善BP神经网络的收敛速度慢和容易陷入局部极小值等不足,具有更好的模式识别能力。以动态电流IDDT为参数的数/模混合正弦波信号发生器为例,建立了以支持向量机为基础的混合电路故障诊断模型,说明了支持向量机在小样本、非线性等情况下具有更强的泛化能力,得出了支持向量机方法的故障诊断率明显高于BP神经网络,为混合电路故障诊断提供一个新的方向。Mixed circuit has low fault detection rate and slow fault diagnosis speed. Here, the paper presents a novel fault diagnosis technology to settle the bottleneck in the development of fault diagnosis technologies. By introducing the extraction of fault feature and optional kernel functions, support vector machine （SVM） can rationally overcome the problems of slow conver- gence speed and local minimum, etc, and has better performance of pattern recognition. An example of sine wave signal genera- tor with dynamic current DDT as the parameter is utilized to establish the mixed circuit fault diagnosis model based on SVM, which explains that SVM has strong generalization ability of learning method in small samples and nonlinear cases. Results show that the proposed method based on SVM has higher fault diagnosis rate than BP neural network method, which is a new orienta- tion in the fault diagnosis of mixed circuit.

关 键 词：支持向量机 混合电路 故障诊断 动态电流 

分 类 号：TN710-34[电子电信—电路与系统]