基于IDDT和SVM的混合电路故障诊断探究

混合电路待测数据受限，存在故障诊断速度较慢、效率有限等问题，提出了一种基于动态电流测试结合支持向量机的混合电路故障诊断方法，其基本思想是运用小波分解提取混合电路动态电流的有效信息，再融合SVM进行故障诊断。采用标准样本Iris数据集研究、确定了多类支持向量机的算法，采用高斯径向基核函数，运用改进的网络搜索方法进行了粗搜索和细搜索，以确定出SVM的最佳参数对。PSPICE及MATLAB软件对混合电路实例的仿真表明，该方法模式识别能力较强，可改善BP神经网络的收敛速度慢和容易陷入局部极小值等不足，适用于混合电路故障的快速准确诊断。

Fault Diagnosis of Mixed Circuit Based on IDDT and Support Vector Machines

Mixed circuit test data is limited; its fault diagnosis is slow, limited efficiency and so on. Proposed a mixed circuit fault diagnosis method based on dynamic current testing and support vector machine. The basic idea is the use of wavelet decomposition to extract the dynamic current of the mixed circuit, and combined with SVM for fault diagnosis. By studying the standard sample Iris data set classification problems to determine the multi-category support vector machine algorithm, using Gaussian radial basis kernel function, the use of improved methods for Web search coarse search and fine search in order to determine the optimum parameters of SVM. By PSPICE software and MATLAB software simulation analysis of the mixed circuit, simulation results show that this method can improve the convergence speed of the BP neural network, and can make the BP neural network is not easy to fall into local minimum value, so that the network has a better pattern recognition capability. This laid the foundation for the completion of a more rapid and accurate mixed circuit fault diagnosis.