基于递归神经网络模型的传感器非线性动态补偿

讨论了递归神经网络模型在传感器非线性动态补偿中的应用，给出了递归神经网络模型的结构及相应的训练算法．递归神经网络模型本身具有动态映射能力，其结构仅与输入层和中间层的节点数有关，且不需要知道被补偿传感器的结构特性(如输出、输入的最大延迟)等先验知识，简化了动态补偿器的结构设计．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．实验结果表明，经过补偿后的传感器具有期望的输入输出特性，应用递归神经网络对传感器进行非线性动态补偿是一种行之有效的方法．

Nonlinear Dynamic Compensation of Sensors Based on Recurrent Neural Network Model

A new approach based on recurrent neural networks model to correct dynamic measurement errors of sensors was investigated. The desired characteristics can be obtained for a compensated sensor. The recurrent neural networks whose structures are determined by the nodes of input and middle layers possess the ability of dynamic mapping. The dynamic compensator can be designed without knowing structure characteristics of a compensated sensor. A recursive prediction error algorithm which converges fast is applied to train the recurrent neural network. The experimental results show that the dynamic compensation method is effective.