基于统计模式识别的优化BP网络在20MnVB钢冲击韧度预测中的应用  

Application of Optimized BP Network Based on SPR for Predicting Impact Toughness of 20MnVB Steel

作　　者：刘晓燕[1] 赵西成[1] 董洁[1] 何晓梅[1] 

机构地区：[1]西安建筑科技大学冶金工程学院,陕西西安710055

出　　处：《热加工工艺》2009年第6期5-8,共4页Hot Working Technology

基　　金：陕西省教育厅专项基金资助项目(03JK134)

摘　　要：用模式识别和人工神经网络建立了20MnVB钢的化学成分与其冲击韧度之间的PLS-BP降维网络模型,用这种方法预测了20MnVB钢的冲击韧度,并与PLS回归方程的预测结果进行了比较。结果表明:所建PLS-BP网络能较准确预测20MnVB钢的冲击韧度,这为研究20MnVB钢冲击韧度提供了一种有效的方法。此外,用训练好的PLS-BP网络模型分析了合金元素B和V的含量对20MnVB钢冲击韧度的影响。A PLS-BP dimension reduction network model used to study the relationship between impact toughness and chemical composition of 20MnVB steel was established using pattern recognition and artificial neural network. The impact toughness of 20MnVB steel was predicted using the network and compared with predicting results acquired fi＇om the PLS regression equation. The results show that the impact toughness can be predicted more accurately using PLSoBP artificial neural network. The network provides an effective method for studying impact toughness of 20MnVB steel. In addition, the influences of alloying element B and V on the impact toughness of 20MnVB steel were analyzed with the PLS-BP network model.

关 键 词：PLS-BP人工神经网络 20MNVB钢 冲击韧度 合金元素 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]