基于CMAC神经网络的热连轧精轧温度预报模型  

Temperature Predictive Model of Hot Tandem Mill Based on CMAC Neural Network

作　　者：王莉[1] 王冬青[1] 王二元[1] 

机构地区：[1]北京科技大学钢铁流程先进控制教育部重点实验室,北京100083

出　　处：《控制工程》2011年第2期188-190,197,共4页Control Engineering of China

基　　金：北京市重点学科建设项目(XK100080537)

摘　　要：温度是带钢热连轧过程中几个最重要的工艺参数之一,由于温度将直接影响到热轧轧制力,因此精确预报各道次,特别是精轧机组各机架的轧制温度,是保证厚度、板形及宽度数学模型命中率的关键。而精轧温度预报技术是热连轧的核心技术。由于传统的模型技术已经不能进一步提高精轧温度的预报精度,针对带钢热连轧精轧温度传统模型的固有缺陷,根据CMAC神经网络具有很强的泛化能力,并且误差收敛速度快的特点,提出了基于CMAC神经网络的热连轧精轧温度预报模型。运用实际生产数据对该网络进行了训练和测试。结果表明,该CMAC模型能准确、实时地预报精轧温度,实现了通过提高精轧温度预报精度来达到提高带钢终轧温度命中精度的目的。通过CMAC预报精轧温度方法与传统的经验模型预报相比,CMAC系统误差的响应速度快、稳定性好,此模型具有良好的在线应用前景。The temperature forecasting problem of the hot strip rolling process is discussed. Because the traditional modeling technology could hardly improve the predictive precision any more, against the inherent defects of traditional finishing temperature models, a CMAC neural network method in hot strip steel is put forward. To make full use of the approximation capability of the CMAC and the quickly convergent velocity, a CMAC finishing temperature predictive model is established. The CMAC neural network is trained and checked with actual production data. The result indicates that the CMAC model could predict the finishing temperature real-time precisely and improve the finishing temperature precision by developing the forecasting effect of finishing temperature. The CMAC finishing temperature forecast method is compared with the traditional experienced forecasting model. The CMAC system error responds quickly and is stable. The method can be used for on-line predicting.

关 键 词：精轧温度 CMAC神经网络 温度预报 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]