基于PSO-RBF神经网络的模拟电路诊断

Analog circuit diagnosis based on particle swarm optimization radial basis function network

为了提高径向基神经网络(radial basis funtion neural network,RBFNN)进行模拟电路故障诊断的速度与准确性,提出了一种基于粒子群算法(particle swarm optimization,PSO)优化RBFNN的故障诊断方法。该方法利用PSO优化RBFNN的结构参数,克服了神经网络中模型结构和参数难以设置的缺点,避免了参数选择的盲目性;同时对模拟电路的响应信号采用小波包分解,提取有效故障特征。仿真结果表明,方法具有更高的诊断精度和更快的收敛速度,能有效地实施模拟电路的故障定位。

In order to improve the speed and accuracy of analog circuit fault diagnosis using radial basis funtion neural network（RBFNN）,this paper proposed a new fault diagnosis method based on RBFNN optimized by particle swarm optimization（PSO）.Trained RBFNN by the PSO algorithm which overcame the shortcomings that structure and parameters of neural network were hard to be set.Preprocessed the response signals of analog circuit by wavelet packet transform as the fault feature.The simulation result shows that this method which has higher diagnostic accuracy and faster convergence speed is effective for fault location.