摘要
采用9×27×9×1四层拓扑结构,以C、Cr、V、Ti、Mo、W、Ni、Cu含量和Cr/C比值为输入参数,以磨损体积为输出参数,构建了钒钛高铬铸铁耐磨损性能的神经网络优化模型,并进行了模型预测验证和铸铁试样的显微组织、物相组成和耐磨损性能的分析。结果表明,该神经网络模型预测精度较高,输出的磨损体积相对预测误差在1.1%~2.9%;优化出的高铬铸铁成分(wt%)由3C、18Cr、1.2V、0.4Ti、1.5Mo、1W、0.1Ni和0.2Cu组成。
By adopting four layers of 9 x27 x9 x l topolologieal structure, the neural network optimized model for high chromiumcastironwithV-Ti was built by taking C, Cr, V, Ti, Mo, W, Ni, Cu content and the value of Cr/C as input parameters and wear resistance as output parameters. The prediction precision of the model was verified, furthermore, the microstructure, phase composition and wear resistance of the samples were analyzed. The results show that the neural network model has high prediction precision and the prediction error of wear volume is between 1.1% and 2.9%. Moreover, the microstructure of optimized high chromium cast iron is composed of 3C, 18Cr, 1.2V, 0.4Ti, 1.5Mo, 1 W, 0.1Ni and 0.2Cu(wt%).
出处
《热加工工艺》
CSCD
北大核心
2016年第4期67-69,71,共4页
Hot Working Technology
基金
全国教育科学规划教育部青年课题(E-CA090455))
关键词
神经网络
钒钛高铬铸铁
耐磨损性
性能优化
预测精度
neural network
high chromium cast iron with V-Ti
wear resistance
performance optimization
prediction precision