非线性动态传感器系统Hammerstein神经网络补偿法

Compensation for the nonlinear dynamic system of a sensor using the Hammerstein neural network

针对传感器动态特性中存在非线性的问题,提出一种基于Hammerstein传感器模型的非线性动态神经网络补偿法。先将补偿模型分解为与Hammerstein模型对应的线性动态与非线性静态2个环节;再设计一种新型的神经网络结构,使网络权系数对应于相应的Hammerstein补偿模型参数,并推导反向传播的网络权系数调整方法;最后通过网络迭代训练,求得补偿模型的线性动态与非线性静态两个环节。仿真与实际实验结果均表明该传感器非线性动态补偿方法使传感器具有理想的输入输出特性。

Aiming at the nonlinear problem existed in dynamic characteristics of a sensor , a new approach to correc-ting dynamic measurement errors of nonlinear sensor systems based on Hammerstein model is investigated .The com-pensation model of Hammerstein is expressed as the accordant linear dynamic subunit and nonlinear static subunit . Secondly , a new neural network structure is designed , the weights in which are corresponding with the parameters of the Hammerstein compensation model , and the method of adjusting network weight coefficients based on back-propagation is deduced .Finally, the nonlinear static part and linear dynamic part of the compensator are identified simultaneously by iterative training .Simulations and experimental results show that the sensor obtains desired input and output characteristics through the nonlinear dynamic compensation method .