传感器标定的非线性校正研究

Nonlinear Correction of Sensor Calibration

针对传感器输入输出存在的非线性误差;本文利用曲线拟合法中的最小二乘拟合、多项式曲线拟合及神经网络拟合三种方法来对某压力传感器标定实验所得的数据进行非线性校正,并对拟合后的数据进行非线性分析;结果表明,通过以上三种曲线拟合方法都可以对传感器标定的数据进行校正,但神经网络拟合后的线性度相对最小二乘和多项式拟合后的线性度要小,非线性校正效果显著.此类方法简捷,实用,可对所有非线性系统进行校正,具有较高的应用价值.

To satisfy the need of correcting the nonlinear characteristic of the sensor between the iv.put and output, this paper uses the least squares, polynomial curve fitting and the neural network fitting methods to do the nonlinear correction, and compares the linearities after fitting. The results show that all the three curve fitting methods can realize the nonlinear correction for the data of sensor calibration, but the linearity after neural network fitting is smaller than that of the other fitting methods. The method of neural network fitting is simple and practical, and it can be used for the calibration data processing of all types of nonlinear system and has significant impact on application areas.