模拟电路单故障与多故障诊断的提升小波和RBF方法

Analog Circuit Single and Multiple Faults Diagnosis Based on Lifting Wavelet and RBF Neural Networks

为了更高效、更准确地诊断模拟电路的单故障和多故障,提出了提升小波和RBF神经网络相结合的方法。该方法用提升小波系数表征故障电路的特征,训练RBF神经网络,将训练好的神经网络作为分类器,对故障电路进行诊断。通过对比,提出的提升小波方法诊断效果明显优于传统小波,准确率达到99.2%,用时更长。结果表明,基于提升小波和RBF神经网络的模拟电路单故障与多故障诊断方法可以有效地提取故障电路的特征并准确快速地对故障进行分类。

In order to diagnosis the analog circuits of single fault and multiple faults more efficiently and accurately,an approach based on lifting wavelet （LW） and RBF neural networks is proposed.The impulse response signal of circuit under test （CUT） is sampled and decomposed to form fault features,training RBF neural networks as classifier to di-agnose the CUT.The filter experiment has proved that presented approach can diagnose single fault and multiple faults of CUT more efficiently. Compared with the traditional wavelet diagnosis rate of 96.8%,diagnosis rate of lifting wave-let can reach 99.2% and with less time.The results showed that analog circuit single and multiple faults diagnosis using lifting wavelet and RBF neural networks can effectively extract the fault features and carry on the diagnosis for the fault accurately and rapidly.