基于改进型松散小波神经网络的软故障诊断方法

Soft fault diagnosis method based on improved loose wavelet neural network

提出了一种基于改进型松散小波神经网络的模拟电路软故障诊断方法.首先应用多分辨率分析和小波包分析将采集得到的电路故障原始数据进行故障特征提取,通过主元分析对提取的故障特征进行降维处理,并作为BP神经网络的输入对其训练,训练的结果即为具有故障模式识别能力的BP神经网络,然后将其应用于待测电路进行电路软故障诊断.仿真实验结果表明,本文提出的改进型松散小波神网络实现了待测电路的故障元件定位,是一种有效的模拟电路软故障诊断方法.

A soft fault diagnosis method of analog circuit based on improved loose wavelet neural network is proposed.Firstly,the multiresolution analysis method and wavelet packet analysis method are applied to extract the fault features from the collected fault data of circuit;secondly,the dimension of the extracted fault features is reduced by PCA analysis method;thirdly,the BP neural network is trained by the fault feature data after dimension reduction,then the trained BP neural network result is saved for fault diagnosis after test;fourthly,for the circuit under test,the BP neural network is applied for soft fault diagnosis after test.Finally,simulation results are demonstrated to venify the efficiency of the proposed improved loose wavelet neural network method.