改进神经网络在模拟电路故障诊断中的应用

Application on Fault Diagnosis of Analog Circuits Based on Improved Neural Network

研究模拟电路故障诊断准确性问题。电路故障与引起故障因素之间呈高度非线性,传统故障识别方法无法识别其非线性特点,导致传统故障方法的诊断精度低。为了提高电路故障诊断的精度,提出一种遗传算法优化BP神经算法的模拟电路故障诊断方法。首先对故障电路样本进行特征提取和归一化处理,然后采用遗传算法对BP神经网络参数进行优化,最后利用最优参数BP神经网络对电路故障样本进行训练和建模,获得电路故障诊断结果。在MATLAB平台上对模拟电路故障进行仿真测试,仿真结果表明,与传统模拟电路故障诊断方法相比,提高了模拟电路故障诊断精度,缩短了故障诊断时间,在模拟电路故障中有着广泛的应用前景。

Study the BP neural network application in the circuit fault diagnosis.Aiming at the shortcoming of overfitting and poor generalization ability in BP neural network analog fault diagnosis,a analog circuit fault diagnosis method is proposed based on genetic algorithm and BP algorithm.The genetic algorithm global optimization characteristics are used to optimize the BP neural network weights.The simulation experiment is carried out with MATLAB using a typical analog circuit.Compared with the other circuit fault diagnosis methods,the training speed of the improved BP neural network algorithm is faster and the accuracy is higher.It is a new effective analog fault diagnosis method.