一种新的模拟电路故障诊断小波神经网络方法

A new wavelet and neural network based method for fault diagnosis of analog circuits

结合小波变换和神经网络二者之间的优点,提出基于小波神经网络的模拟电路故障诊断方法。采用能量分布特征提取方法和改进BP算法,用正弦信号仿真模拟电路,应用小波变换对模拟电路的采样信号进行多尺度分解,再进行能量分布特征提取,然后利用神经网络对各种状态下的特征向量进行分类识别,实现模拟电路故障诊断。相对于传统的故障诊断方法,用小波变换对故障信号进行预处理,大大减少了神经网络的输入数目,从而简化了神经网络的结构和减少了它的训练时间,并提高了辨识故障类别的能力。对实例电路仿真结果表明,该方法能正确识别各种故障状态,准确率高。

Combining the time-frequency location and multiple-scale analyzation of Wavelet transform （WT） with the nonlinear mapping and generalizing of Neural Network, a method of fault diagnosis in analogue circuits is proposed. Sinusoidal input to the analog circuit was simulated and its output was sampled in time domain to collect training data for neural network. The collected data was processed by WT to draw energy features, ie. , generate fault features. Feature vectors under certain states could be classified using neural network with improved BP algorithm. Using wavelet decomposition to process the impulse response drastically reduce the number of input fed to the Neural Network, simplifying its architecture and mininizing its training and processing time. Simulation results show that proposed fault diagnosis approach is feasible.