基于神经网络的热轧卷取温度模型辨识  

Identification of Coiling Temperature Model of Hot Roll Based on Neural Network

作　　者：王益群[1] 王海芳[1] 孙旭光[1] 张伟[1] 

机构地区：[1]燕山大学,秦皇岛066004

出　　处：《中国机械工程》2006年第1期71-74,共4页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(60474044);河北省自然科学基金资助项目(E2004000221)

摘　　要：针对热轧带钢卷取温度模型具有高度非线性的特点,利用神经网络具有逼近任何非线性函数及预报的性质,采用附加动量BP算法,准确预报卷取温度,进而应用最小二乘辨识方法对卷取温度统计模型进行参数辨识,辨识结果与设定结果的比较表明此方法行之有效。这种神经网络预报与最小二乘线性辨识相结合的方法为热轧带钢卷取温度模型的辨识优化提供了新的途径。The model of conventional coiling temperature is highly nonlinear. The coiling temperature of hot roiled strip was exactly predicted based on neural network（NN） by means of its approximation to any non-linear system and its ability of prediction. Additional momentum method, which is an improved BP algorithm, was used in the NN. And the predicted coiling temperature was applied to identify characteristic parameters of the static model of coiling temperature with aid of traditional least square method. Finally, identification results are illustrated and verified by referenced values.

关 键 词：热轧带钢 卷取温度 神经网络 BP算法 最小二乘法 

分 类 号：TG333.5[金属学及工艺—金属压力加工]