基于BP神经网络的TB8钛合金流变应力预测模型  

Flow Stress of TB8 Alloy Prediction Model Based on BP Neural Network

作　　者：薛莹[1] 

机构地区：[1]陕西警官职业学院计算机应用与材料应用,陕西西安710043

出　　处：《热加工工艺》2015年第14期145-148,共4页Hot Working Technology

基　　金：重庆市“两江学者”计划专项经费资助;重庆市科技攻关计划项目(CSTC2011AB3005)

摘　　要：利用人工神经网络（ANN）模型来建立钛合金本构关系以及TB8钛合金热压缩试验数据,采用误差反向传播（Error Back-Propagation Networks）算法模拟了流变应力。结果表明：TB8钛合金在较宽泛温度948~1073K,应变速率在0.001~10s-1含有两个节点数为18的隐含层BP神经网络模型,这为研究TB8钛合金高温塑性变形行为提供了依据。对不同相区不同变形机制的TB8钛合金应力应变行为进行精确表征,训练阶段,最大绝对相对误差3.78%。在验证阶段,最大绝对相对误差4.06%,且大部分相对误差点分布在±2%的范围内,实现了较高的精度。The constitutive relation of titanium alloy was established by artificial neural network（ANN）, and the flow stress was simulated by experimental data from the isothermal compressions of TB8 titanium alloy and error back-propagation Networks. The results show that TB8 titanium alloy hidden the layer of BP neural network containing two 18 nodes at wide temperature range from 948 K to 1073 K and strain rate at 0.001 s-1to 10 s-1, which can offer foundation to study the high temperature deformation behavior of TB8 alloy. The strain and deformation behavior of TB8 titanium alloy with different phase region and deformation mechanism was accurately characterized. The maximally absolute and opposite is 3.78% at training stage and is 4.06 % at verifying stage, and most absolute and opposite error distribute at about 2% error, which come true the good precision.

关 键 词：BP神经网络 钛合金 流变应力模型 

分 类 号：TG146.2[一般工业技术—材料科学与工程] TP391[金属学及工艺—金属材料]