基于D-S理论的故障诊断融合算法及应用研究

Research on Application of Fault Diagnosis Fusion Algorithm Based on D-S Evidence Theory

针对工业现场传感器状态类型复杂多变、被测参量难以准确可靠获得等问题,提出了一种基于RBF神经网络和证据理论的两级信息融合方法。利用RBF神经网络实现特征层数据融合,并建立基本信任分配函数,解决了D-S证据理论确定基本信任分配函数困难的问题;基于D-S证据理论的传感器故障诊断方法,可用来判断出工业现场传感器的有效工作状态。木材含水率检测结果表明,基于RBF神经网络和证据理论的两级信息融合方法可正确定位并准确分离出失效传感器。

To solve the problems of sensors used in industrial fields, including complicated types of status and hard to acquire proper measured parameters, etc. , the method of two-level information fusion based on RBF neural network and evidence theory is proposed. The data fusion on feature layer is realized by using RBF neural network, and the basic belief assignment function is established, the difficult issue of determining basic belief assignment function with D-S evidence theory is solved. The fault diagnostic method based on D-S evidence theory can be used for judging effective operation status of the sensor. The detection result of the moisture in lumber indicates that this two-level information fusion method is able to determine and accurately separate the failure sensors.