基于BP神经网络的传感器非线性补偿

Nonlinear Compensation of Sensors Based on BP Neural Network

由于传感器本身的非线性特性以及传感器在测量过程中外界环境因素的影响,使得传感器的输入输出特性呈现出非线性.讨论了BP神经网络模型在传感器非线性补偿中的应用.给出了相应的补偿方法,即采用两个相同的传感器对同一被测量进行测量,其测量结果作为神经网络模型的输入,经过补偿后的传感器具有线性的输入输出关系.采用递推预报误差算法训练神经网络,具有收敛速度快、收敛精度高的特点.试验结果表明,应用神经网络对传感器的非线性进行动态补偿是一种行之有效的方法.

Many sensors have nonlinear input-output characteristics because of their inner nonlinear properties and the influences of measurement circumstance. BP （Back Propagation） neural network models are applied to the nonlinearity compensation of sensors in this paper. Two sensors with the same type are used to measure the same measurand and the outputs of the sensors are imported into the trained neural network model to obtain the linear input-output characteristics. A RPE （recursive prediction error） algorithm with the advantage of fast convergence is applied to training the recurrent neural network. Experimental results show that the nonlinear compensation of sensors based on neural network models is effective.