基于径向基神经网络的薄板平整轧制力预报模型  

A Model to Predict Temper Rolling Force of Thin Gauge Strip with RBF Neural Networks

作　　者：段雪厚[1] 王石刚[1] 徐威[1] 唐成龙[1] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《上海交通大学学报》2011年第6期924-928,共5页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金资助项目(50975171)

摘　　要：冷轧薄板在平整轧制时具有轧件厚度薄、压下率小的特点,其平整轧制力往往计算困难,精度难以保证.针对上述情况,提出一种基于参数修正的轧制力数学模型来计算其平整轧制力.同时,为进一步提高计算精度,运用RBF(Radial Basis Function)神经网络来预测该平整轧制力数学模型的计算误差,并将该误差与数学模型的计算值相结合,完成对其的修正.离线仿真结果表明,薄板平整轧制力数学模型在经过自身修正参数及RBF神经网络的2次修正后,计算精度可达到6%以内,具有较高的工程应用价值.The temper rolling process of thin gauge strip has the features of thin thickness and small reduction,thus it is difficult to calculate its temper rolling force and get a high-precision value.Aiming to solve this problem,this paper established a mathematical model to calculate its temper rolling force based on parameters modification.Meanwhile,the paper predicted the calculation error of the mathematical model above by using RBF neural networks in order to improve its precision further,and completed the modification for temper rolling force model by combining the calculation error and calculated values of original mathematical model.The off-line simulation indicates that the precision of new model of temper rolling force for thin gauge strip can reaches 6% after twice modification by its modified parameters and RBF neural networks,which makes it have high value of application in real project.

关 键 词：平整轧制力预报 薄板 径向基神经网络 参数修正 

分 类 号：TG335.5[金属学及工艺—金属压力加工]