BP改进算法神经网络的保护渣性能预测模型  

Predictive model of mould flux properties based on improved BP algorithm of neural network

作　　者：向嵩[1] 王雨[2] 刘国权[1] 

机构地区：[1]北京科技大学材料科学与工程学院,北京100083 [2]重庆大学材料科学与工程学院,重庆400044

出　　处：《炼钢》2006年第3期45-48,共4页Steelmaking

基　　金：国家自然科学基金;上海宝钢集团联合基金资助项目(50274078)

摘　　要：采用混料回归设计原理设计试验保护渣组成。以试验为基础,针对常用BP算法的不足采用动量因子与自适应学习速率相结合的BP改进算法建立神经网络保护渣性能预测模型。研究结果表明该模型预测精度高,适用组元多、成分变化范围大;对保护渣的性能预测取得了较好的效果,能为保护渣设计提供理论指导。Compositions of the mould flux are designed according to the principle of regression design on the blended mix. In light of deficiency of the normal BP algorithm, a neural network predictive model for the mould flux properties is established on the basis of the improved BP algorithm in combination of momentum factor and self-adaptive learning rate. Experimental results show that this model is highly accurate and suitable for the mould flux with multi-components and the wide range of compositions. Better results have been achieved in prediction of the mould flux properties by this new model, thus providing a solid theoretical basis for mould flux design.

关 键 词：保护渣 预测模型 BP神经网络 

分 类 号：TF044[冶金工程—冶金物理化学]