基于BP神经网络的20CrMnTiH钢淬透性预报模型  

BP neural network-based prediction model for quenching degree control of 20CrMnTiH

作　　者：林俊[1,2] 廖俊[2] 

机构地区：[1]上海交通大学自动化系,上海200030 [2]宝山钢铁股份特殊钢分公司,上海200941

出　　处：《炼钢》2008年第4期60-62,共3页Steelmaking

摘　　要：建立一个三层BP神经网络模型,利用冶炼过程的化学成分预报20CrMnTiH钢的J9端淬值,实现±3HRC误差范围内96%的准确率,同时分析了误差存在的原因。A three level BP neural network prediction model has been built and the J9 top quench value for 20CrMnTiH predicted by analysis of the chemical compositions in the process of refining. The hitting rate of the report has reached 96% with the tolerance controlled within the limit of 3HRC . The cause to the deviation is also analyzed.

关 键 词：20CrMnTiH 淬透性 人工神经网络 

分 类 号：TF764[冶金工程—钢铁冶金]