基于径向基神经网络的电涡流传感器输出特性拟合研究  

Research of output characteristic fitting of eddy-current sensor based on radial-basis function neural network

作　　者：尤文坚[1] 梁兵[2] 李荫军[3] 

机构地区：[1]广西农业职业技术学院电子信息工程系,广西南宁530007 [2]广西机电职业技术学院电气工程系,广西南宁530007 [3]广西电网公司南宁供电局,广西南宁530031

出　　处：《工矿自动化》2013年第2期47-50,共4页Journal Of Mine Automation

基　　金：广西省教育厅自然科学科研项目(200911LX547)

摘　　要：针对电涡流传感器的输出特性参数非线性较大,不能精确地反映被测物理量的问题,提出了一种采用径向基神经网络对电涡流传感器的输出特性参数进行拟合的方案。该方案采用newrb函数创建一个径向基神经网络,以被测物理量作为输入矩阵、电涡流传感器输出电压作为输出矩阵,对该径向基神经网络进行训练,从而可得到均方根误差小且光滑的电涡流传感器输出特性拟合曲线。实验结果表明,只要选择合适的创建函数和扩展系数,径向基神经网络能有效地实现电涡流传感器输出特性的拟合。In view of problem that eddy-current sensor cannot reflect measured physical quantity accurately caused by higher nonlinear of output characteristic parameter, the paper proposed a scheme of using RBF neural network to fit output characteristic parameter of eddy-current sensor. The scheme uses newrb function to create RBF neural network, and takes measured physical quantity as input matrix and output of eddy-current sensor as output matrix to train the RBF neural network, so as to obtain low root- mean-square error and smooth output characteristic fitting curve of eddy-current sensor. The simulation result showed that RBF neural network can effectively realize fitting of output characteristic of eddy- current sensor by selecting proper creating function and expanding coefficient.

关 键 词：电涡流传感器 输出特性拟合 径向基神经网络 newrb函数 扩展系数 

分 类 号：TD67[矿业工程—矿山机电]