Abstract: This paper proposes a new method for simulating circuit fault diagnosis. The method includes wavelet decomposition of Haar, normalization of data, and optimization of RBF neural network by wolf pack algorithm. Firstly, the original fault data set of the obtained circuit is transformed by Haar wavelet, and then the transformed data is normalized, and finally the input data needed for RBF neural network training is obtained. This paper uses wolf pack algorithm to optimize RBF neural network to improve the stability of network training and the success rate of diagnosis for RBF neural network. Finally, the implementation process of these methods is discussed through the diagnosis of two circuits, and the feasibility of the simulated circuit fault diagnosis is verified by using this method.

Key words: analog circuit, fault diagnosis, RBF neural network, wavelet decomposition, wolf pack algorithm

摘要： 提出了一种新的方法来进行模拟电路故障诊断。该方法包括Haar的小波分解，对数据的归一化处理，以及用狼群算法优化RBF神经网络。用Haar小波对所得的电路原始故障数据集进行变换，然后对变换后的数据进行归一化处理，最终得出RBF神经网络训练所需的输入数据。针对RBF神经网络中隐层节点中心、基函数宽度及权值选取困难问题，使用狼群算法来优化训练RBF神经网络，以提高网络训练稳定性与诊断成功率。通过两个电路的诊断实例，来论述这些方法的具体实现过程，验证用该方法进行模拟电路故障诊断的可行性。

关键词: 模拟电路, 故障诊断, RBF神经网络, 小波分解, 狼群算法