组合网络与证据理论相结合的电路故障诊断

针对神经网络中两种常用的网络类型BP和Elman,在模拟电路故障诊断中的应用过程的特点——各自网络都有自身缺陷,提出基于BP-Elman神经网络与证据理论相结合的故障诊断方法.首先BP-Elman网络得出其各自的初步诊断结果,经过必要的转换将其转换成证据理论的概率赋值,作为证据组合的依据;证据理论组合规则将初步诊断结果融合得出决策级诊断结果.通过对某装备位置调节器板的故障诊断过程表明,该模型经过尽可能融合有效故障信息,大大提高了系统的诊断精度和诊断正确率,而且有效降低了系统不确定性.

Study of Analog Circuit Fault Diagnosis base on Integration of Combined Neural Network with D-S Evidence Theory

There are two kings of normal network style-BP and Elman, they have their self defect in the application process of analog circuit fault diagnosis, a fault diagnosis new approach for analog circuit diagnosis faults is presented based on Integration of BP-Elman neural network with D-S evidence theory. First BP and Elman network get their self preliminary diagnosis results, which is changed to probability evaluate of evidence theory with essential means and is taken as basis of evidence combination; then combination rule of evidence theory obtain decision-making diagnosis results from preliminary diagnosis result. The fault diagnosis process of some equip position adjustor indicate that the approach can fused efficient fault information as far as possible, increased the precision and correctness of system fault diagnosis greatly and reduced unsureness of system.