支持向量机在模拟电路故障诊断中的应用

故障诊断发展的瓶颈之一是故障样本的缺乏，而不仅在于诊断方法本身。支持向量机是建立在结构风险最小原则基础上，专门针对小样本情况的，其目标是得到现在信息下的最优值而不仅仅是样本数趋于无穷大时的最优值。它能在训练样本很少的情况下达到很好的分类效果，从而为故障诊断技术向智能化发展提供了新的途径。介绍了支持向量机的二值分类算法，以支持向量机二值分类为基础，构建了基于支持向量机的多值分类器并应用于模拟电路故障诊断。以两管视频放大器的多种故障分类为例，进行了实际应用验证。结果表明，该诊断方法具有算法简单、可对故障在线分类，有很好的分类能力和较高的计算效率，不需要对原始数据进行预处理就可达到满意的效果。

Application of Support Vector Machine in Fault Diagnosis of Analog Circuits

One of the bottlenecks restricting the development of fault diagnosis lies in the lack of samples. Support Vector Machine (SVM) is a machine - learning algorithm based on structural risk minimization principle with small quantity of samples. Its target is to get superior solution not only in the case of infinite samples but also under existing information. SVM has desirable classification ability even with fewer samples, which provides us a new method to develop the intelligent fault diagnosis. The binary classification algorithm of SVM is introduced. Multi - fault SVM classifiers based on binary classification are developed and applied to fault diagnosis of Analog Circuits. Experiment of Multi - fault SVM classifiers applied to fault diagnosis of Analog Circuits by video frequency amplifier circuits is conducted. The results from the experiment show that the SVM method has such advantages as simple algorithm, online fault classification, a good classification ability and high efficiency, even for the cases without pre - processing of the original signals.