带钢纠偏系统的神经网络辨识

Strip Steel Correction System Identification Using Neural Network

针对退火炉内连续带钢纠偏系统控制和改善纠偏系统性能领域建立准确系统动态模型的需要,建立一种回归神经网络辨识非线性水平摆动式带钢纠偏系统数学模型,研究具有内部状态反馈的神经网络和误差能量最小的网络权重训练算法。利用辨识实验获得输入/输出数据动态调整网络权值。辨识结果表明:神经网络描述的带钢纠偏系统数学模型有较高精度,权重训练算法具有良好全局逼近能力。

According to the requirement of the controlling of the correction system and accurately dynamic modeling in the field of improving correction system performance,a method of identifying was built to identify mathematical model of nonlinear strip steel correction system.The neural network having internal state feedback and network weight training algorithm with minimum energy error are researched.The weights of neural networks were dynamic adjusted by input/output data of process that were acquired by experiments of identifying.The experimental results show that strip deviation rectification system mathematical model described with neural network has a high accuracy,and weight training algorithm has the ability of good universal approximation.