基于自适应神经网络集成的电子设备故障预测方法

为了实现装备的主动维修,针对电子设备故障预测问题,提出了一种基于自适应神经网络集成(ANNE)的电子设备故障预测方法。首先利用FCM 聚类算法生成个体网络训练样本,从而确定了神经网络集成的规模。ANNE 根据故障序列样本与个体网络训练样本的相似度动态调整权值,自适应神经网络集成根据装备故障历史数据建立故障预测模型,根据当前时间预测故障间隔时间。仿真实例证明,该方法对平稳的故障间隔时间数据进行故障预测的精度较高。

Prognostic of electronic equipment based on adaptive neural network ensemble

To support the presentative mantainance of electronic equipment,a prognisitic method of electronic equipments was proposed based on adaptive neural network ensemble (ANNE),which uses fuzzy C-means(FCM) clustering algrithm to generate training samples of the individual neural network. The clustering number is the number of the individual of ANNE. ANNE adaptively adjusts the weights of ANNE. The prognistic model was bulit with historical fault data of electronic equipment by adaptive neural network ensemble. The prognistic model was used to predict fault occuring time by current time. The simulation example demenstrats that the proposed method can accuratly predict the fault time with smoothing fault interval time.