摘要
针对传感器输出易受温度、湿度等因素影响的问题,提出利用RBF神经网络良好的学习、泛化和非线性逼近能力改善传感器的输出特性。RBF神经网络采用不需要事先确定隐层单元个数、可在线自适应学习的最近邻聚类学习算法。将该算法用于易受温度影响的压力传感器的非线性校正,仿真结果表明RBF神经网络在对传感器输出信号的补偿精度和网络训练速度方面均优于BP神经网络和传统的非线性补偿方法。该方法可推广应用于其他传感器输出特性的优化。
Aiming to the problem of sensors＇ output performance easily interfered with measure circumstance such as temperature, humidity etc. RBF neural network which have good abilities such as self-learning, generalizing and nonlinear approaching is proposed to optimize the sensors＇ output performance. Nearest Neighbor Clustering Algorithm （NNCA）, one of on-line adaptive learning arith- metic, which the unit numbers of hidden layer can be automatically determined is used to correct the output of pressure-sensor which is easily interfered with temperature. The simulation results show that the calibrated precision and training speed are much better than BP neural network and other traditional nonlinear compensation methods. The model can be also used to optimize other sensors ＇ output performance.
出处
《微计算机信息》
北大核心
2008年第28期237-238,233,共3页
Control & Automation
基金
多传感器数据融合在瓦斯检测系统中的应用
山西高校科技研究开发项目(山西省教育厅颁发)(20051216)
关键词
径向基函数
传感器
输出特性
Radial basis function
Sensor
Output performance
