20CrMnTiH钢端淬值数学模型  

Mathematical Models for Top Quench Value of Steel 20CrMnTiH

作　　者：林俊[1] 廖俊[2] 

机构地区：[1]上海交通大学,上海200030 [2]宝钢股份公司,上海200941

出　　处：《宝钢技术》2008年第4期42-45,共4页Baosteel Technology

摘　　要：开发了4种20CrMnTiH钢的端淬预报数学模型,并对2007年连续生产的50炉实绩数据进行预测对比。基于线性回归方程模型及基于加斯特方程模型结构简单,易控制但准确率较差;基于DI值模型以金属理论基础建模,但淬透性理论的不完善导致该模型也存在一定的误差;基于BP神经网络的端淬控制模型具有最高的准确性,适用于工业生产。分析了各模型误差产生原因,指出凝固偏析与测量误差是影响模型准确率的主要原因。Four kinds of mathematical models for predicting the top quench value of steel 20CrMnTiH were developed. The real data of 50 heats in successive production in 2007 were predicted and compared. Both the model based on the linear regression equation and the model based on the Jiasite equation are simple and easy to control, but not very accurate. The model based the DI value is guided by the metal theory, but there exist certain mistakes as its quenching degree theory is not yet perfect. The top quench control model based on the BP neural network possesses the highest accuracy, and it is suitable for the industrial production. With analysis of each model＇ s error causes, it was pointed out that the solidifiction segregation and errors in measurement were the main reason that affected the models＇ accuracy.

关 键 词：20CrMnTiH 淬透性 人工神经网络 

分 类 号：TG161[金属学及工艺—热处理]