基于主元分析和D-S证据理论的传感器故障诊断与应用

摘要

Because the type sensor for underground complex, the measured parameters of the data were huge, used principal component analysis to reduce the dimensions of the data. Used RBF neural network to carry out the feature level data fusion,and established distribution function of basic trust, further used the evidence theory advantage of representation and reasoning to inaccurate information, realized the fault detection and isolation capabihties effectively. The simulation shows that the use of principal component analysis and D-S theory can be correctly located and accurately isolate the failure of sensors.%针对井下传感器状态类型复杂多变、被测参量数据庞大等问题,采用主元分析法对数据进行降雏处理.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,再以证据理论对非精确信息的表示和推理优势,有效实现了故障检测和分离.实例仿真表明,利用主元分析和D-S理论能正确定位并准确分离出失效传感器.