基于D-S证据和PSO神经网络的电路故障诊断

Fault Diagnosis for Analog Circuits Based on D-S Evidence Theory and PSO Neural Network

针对传统故障诊断方法在进行电路故障诊断面临的不确定性,提出了一种基于D-S证据和PSO神经网络的故障诊断方法;首先,确定BP神经网络结构,并采用PSO算法对BP神经网络的各参数进行优化以提高诊断的精确性,然后,将PSO神经网络的诊断输出作为证据采用D-S证据合成规则进行融合,并根据阈值来确定具体故障;仿真实验表明,在阈值为0.75和训练目标误差为0.02时,文中的基于D-S证据理论和PSO神经网络的故障诊断模型,能准确地实现模拟电路的故障诊断,降低了系统的不确定性。

In order to conquering the uncertainty of the traditional fault diagnosis for analog circuits, the fault diagnosis method based on D-S evidence and PSO neural network is advanced. Firstly, the structure of BP neural network is assured and the PSO algorism is used to optimize the parameters of BP neural network to improve the diagnosis accuracy, then the output of the PSO neural network is used as the ev-idence for the D-S combination rule and the threshold is assured to get the final diagnosis. The simulation experiment shows the diagnosis method in this paper based on D-S evidence and PSO neural network under the threshold 0. 75 and the goal error 0. 02 can realize the diag-nosis for analog circuits, deducing the uncertainty of system.