基于径向基函数神经网络的探空仪湿度传感器曲线拟合

Curve Fitting of Humidity Sensor Used in Digital Sonde Based on RBF Neural Network

针对数字式探空仪上采用的XGH-02型高分子碳膜湿敏电阻湿度感应元件,采用正规化径向基函数(RBF)神经网络模型对其进行曲线拟合,与传统的曲线拟合效果相比较,寻求一种更加准确的湿度传感器标定和误差的校准模型。用训练样本和检验样本对建立的RBF模型分别进行训练和检验结果表明,建立的RBF模型有效提高了湿敏电阻的准确度,其测量的最大误差为2.0298%(RH),明显好于采用现用公式的测量准确度。

The curve fitting of XGH-02 macromolecule carbon hygristor used in digital sondes is studied,based on the model of RBF（Radial Basis Function） neural network.Compared with the traditional method of curve fitting,a more precise model of sensor and error calibration is presented.The training and testing of the RBF neural network the model with training and testing samples indicates that the model of the RBF neural network can improve the accuracy of humidity resistance effectively,and the maximum error of measurement is 2.0298%（RH）,which is smaller than that of the existing formula.