基于神经网络的传感器非线性误差校正方法

为对传感器进行非线性校正以进一步提高其测量精度,提出了基于神经网络的校正办法。理论分析了传感器非线性误差的复杂性,并以位移传感器标定为例,详细介绍了传感器非线性校正的过程和方法。采用了最小二乘拟合、BP神经网络以及RBF网络三种方法进行校正,设计并实现了RBF网络的校正模型。实验结果证明,RBF网络的校正方法比BP网络校正方法精度提高了约44%,其补偿效果更优,且其在传感器种类变化或环境影响较大的情况下比最小二乘拟合更具非线性补偿优势。

Nonlinear errors correction method of sensors based on neural network

In order to further improve measurement accuracy of sensor, a non-linear errors correction method for the sensors based on neural network be proposed. Theoretical analysis of the complexity of the sensor nonlinearity error, took example as displacement sensor calibration, introduced the details of the non-linear sensor calibration process and methods. Three methods including Least Squares, BP Neural Network and RBF Network have been used for errors correcting, designed and implemented a calibration model of RBF Network, and the results shows that the accuracy of RBF Network has been increased by about 44%than the accuracy of BP Network, and it has more nonlinear compensation advantage than the Least Squares in complex environment and various types of multi-sensor application.