Abstract: An analog circuit fault diagnosis system is designed based on information fusion and neural network theory. By measuring accessible node voltages and peak of output voltage under different excitation frequencies with microcontroller-based signal acquisition system, two types of fault characteristic parameters of voltage and magnification are acquired. Preliminary diagnosis is performed separately by BP neural networks with these two characteristic parameters. Fused diagnosis and fault location is accomplished by using fuzzy transform method. Experimental results show that this system can locate different types of component fault with higher accuracy compared with diagnosis system based on single test information.

Key words: information fusion, fault diagnosis, neural network, analog circuit, single chip microcomputer, signal collection

摘要： 设计一种基于信息融合和神经网络理论的模拟电路故障诊断系统。采用基于单片机的信号采集系统，对电路可及点电压和不同激励频率下的电路输出电压峰值进行采样，得出电压和电路放大倍数2类故障特征参数，利用其进行基于BP神经网络的初步诊断，运用模糊变换方法进行融合诊断及故障定位。实例诊断结果表明，与基于单一信息的诊断系统相比，该系统能定位不同类型的元件故障，诊断准确率较高。

关键词: 信息融合, 故障诊断, 神经网络, 模拟电路, 单片机, 信号采集

CLC Number: 

• TP391