基于BP神经网络的置氢TC21合金力学性能预测  

Artificial Neural Network Model for the Prediction of Mechanical Properties of Hydrogenated TC21 Titanium Alloy

作　　者：孙宇[1] 曾卫东[1] 赵永庆[2] 张学敏[1] 马雄[1] 韩远飞[1] 

机构地区：[1]西北工业大学凝固技术国家重点实验室,陕西西安710072 [2]西北有色金属研究院,陕西西安710016

出　　处：《稀有金属材料与工程》2012年第6期1041-1044,共4页Rare Metal Materials and Engineering

基　　金：国家"973"计划(2007CB613807);新世纪优秀人才支持计划(NCET-07-0696);凝固技术国家重点实验室开放课题(35-TP-2009)

摘　　要：基于神经网络的非线性映射和泛化能力,采用人工神经网络方法,建立了置氢TC21合金力学性能预测的BP神经网络模型。模型的输入参数包括高温拉伸试验温度和置氢含量,输出参数为合金的常用力学性能指标,即抗拉强度和屈服强度。通过检验样本验证了ANN模型的准确性。结果表明:该模型具有容错性好、通用性强等优点,可以预测置氢TC21合金在不同拉伸温度和不同置氢含量下的机械性能。同时,将神经网络技术应用于材料制备工艺设计领域,可以明显地提高工艺设计效率,缩短实验周期。Based on the ability of nonlinear mapping and generalization, an artificial neural network model for the prediction of mechanical properties of hydrogenated TC21 titanium alloy was established. The input parameters of the neural network model includes temperature tensile testing temperature and hydrogen content. The outputs of the model are mechanical properties namely ultimate tensile strength and tensile yield strength. The accuracy of ANN model was tested by the test sample. It is found that the predicted results are in good agreement with experimental value because of the characters of good fault-tolerance and strong commonality. The trained model can predict the mechanical properties of hydrogenated TC21 alloy under the condition of different experimental temperatures and contents. With the help of application of neural network technology in the field of material preparation process and design, the efficiency can be improved greatly, and the cycle of the actual experiment will be shortened obviously.

关 键 词：BP神经网络 置氢 力学性能 预测 

分 类 号：TG146.23[一般工业技术—材料科学与工程]