应用Elman网络优化非线性模拟电路测试激励

非线性模拟电路瞬态测试激励信号的参数对电路故障识别率影响很大,在搜索最佳激励信号的过程中,需建立非线性模拟电路的系统模型。Elman网络是一种递归神经网络,能逼近任意动态非线性系统。该文用一种改进的Elman网络建立故障电路和非故障电路的系统模型,用遗传算法搜索最佳瞬态测试激励信号参数,仿真实验结果表明经过该方法优化后的激励信号能大大提高非线性模拟电路的故障识别率。

Optimization of Testing Stimulus for Nonlinear Circuits by Applying Elman Neural Network

The parameters of transient testing stimulus for nonlinear analog circuits greatly influence the fault diagnosis efficiency. In the course of finding the optimum stimulus, we have to develop the nonlinear analog circuits' models. Elman network which is a dynamic recurrent neural network can approximate any dynamic nonlinear system. In this paper, Elman network is used to develop the models of the fault circuits and fault free circuit, the optimum stimulus is searched with genetic algorithm. Experiment results demonstrate that the optimum stimulus achieved has high fault identification.