模糊神经网络在电子装备故障智能诊断中的应用

Application of Fuzzy NN in Fault Intelligent Diagnosis for Electronic Equipment

据故障现象与原因模糊集合及其隶属度,建立总模糊关系隶属度矩阵。导出故障原因计算模型和故障诊断数学模型,并采用改进的 5 层 BP 模糊神经网络求解。通过 BP 网络学习,若输出层出现误差,把误差逐层反向传播至各节点,算出各节点的参考误差,并据此自动调整模糊推理模型中的隶属函数,即可量化输出层输出模糊化数值。

Subjection matrix of fuzzy relation was established according to fault phenomenon and causation fuzzy set and its grade of membership. Calculation model of fault causation and mathematical model of fault diagnosis was educed, these models were calculated with improved BP fuzzy neural network with 5 layer. After BP network learned, if error was appeared in output layer, the error was sent to every node reversed layer by layer and reference error of every node was calculated. The fuzzy value of output layer was estimated according to the subjection functions of fuzzy reasoning model.