基于RBF网络的模拟电路故障诊断算法

Algorithm Based on RBF Artificial Network for Analog Circuit Faults Diagnosis

针对BP神经网络在模拟电路故障诊断上存在的收敛速度慢、易陷入局部最小等不足,提出了一种基于多层小波分解和RBF神经网络的模拟电路故障诊断算法。为提高诊断效率,用多层小波分解能有效提取电路故障特征;用RBF网络优良的泛化能力和快速的非线性逼近能力可以较好的解决模拟电路中存在的容差和非线性问题。故障诊断仿真实验表明,在保证较高故障诊断正确率的情况下,RBF网络的训练次数得到了极大地缩小,有效克服了基于BP网络算法存在的上述不足,极大地提高了模拟电路故障诊断的时间效率。

Aiming at faults diagnosis of analog circuit, an algorithm based on radial basis function neural network （RBF ANN） and multi -layer wavelet decomposition is discussed in this paper. The multi -layer wavelet decomposition can extract fault features, and meanwhile, the RBF ANN＇ s excellent generalization ability and nonlinear approximation ability can resolve analog circuit＇ s tolerance and nonlinearity effectively. Simulation results on benchmark circuits show that the training epochs of RBF ANN are less than that of BP ANN dramatically. The shortcomings of BPNN, such as low convergence speed and prone to fall into the local minimum points are overcame. And the diagnosis efficiency is enhanced.