基于神经网络的BFe30-1-1铜合金的本构关系模型  

Model of constitutive relationship of BFe30-1-1 copper alloy based on neural network

作　　者：马艳霞[1] 韩茂盛[1] 刘乐乐 梁晨 MA Yanxia;HAN Maosheng;LIU Lele;LIANG Chen(Luoyang Ship Material Research Institute,Luoyang 471000,He＇nan China)

机构地区：[1]中国船舶重工集团公司第七二五研究所,河南洛阳471000

出　　处：《锻压装备与制造技术》2018年第5期65-69,共5页China Metalforming Equipment & Manufacturing Technology

摘　　要：采用Gleeble-3500热模拟实验机研究BFe30-1-1铜合金在变形温度为800～950℃、应变速率为0.1～20s-1下压缩过程的高温流变行为。结果表明,BFe30-1-1铜合金的流变应力随着变形温度的增加而减小,随着应变速率的增大而增大。根据BP人工神经网络算法原理,建立了实验合金高温压缩过程中真应力与应变、应变速率和变形温度关系的神经网络预测模型,预测值与实验值对比表明BP神经网络具有很高的预测精度,所建立的本构模型平均相对误差为1.65%。该模型可以很好的描述BFe30-1-1铜合金在高温变形时各热力学参数之间高度非线性的复杂关系,为本构关系模型的建立提供了一种准确有效的方法。The high temperature flow stress behavior of the BFe30-1-1 copper alloy has been studied during compression experiment within a deformation temperature range of 800-950℃ and a strain rate range of 0.1~20s-1 by use of Gleeble-3500 simulator. The experimental results show that the flow stress of BFe30-1-1 copper alloy decreases with the increase of deformation temperature and enhances with the increase of the stain rate. The predicted model of neural network among true stress ＆ strain, strain rate and deformation temperature for experimental alloy has been established according to BP Artificial Neural Network algorithm principle. By comparison the predicted value with experimental value, it is indicated that BP neural network is of high prediction accuracy and the average relative fault of the constitutive relationship model is 1.65%. This model can well describe the complicated nonlinear relationship of thermo dynamical parameters. It pro- vides a more convenient and effective way to establish the model of constitutive relationship for BFe30-1-1 copper alloy.

关 键 词：BFe30—1—1铜合金 流变行为 本构关系 BP神经网络 

分 类 号：TG146.1[一般工业技术—材料科学与工程]