基于Elman神经网络的橡胶介质复合胀形多通管坯料参数优化  

Optimization of blank parameters of rubber media compound bulging for multi-pass tubes based on elman neural network

作　　者：陈志忠[1] 刘斌[1] 隗平平[1] 

机构地区：[1]华侨大学机电及自动化学院,泉州362021

出　　处：《机械设计与制造》2011年第8期188-190,共3页Machinery Design & Manufacture

基　　金：福建省科技计划重点项目(2008H0085)

摘　　要：探讨了运用人工神经网络技术进行复合胀形多通管坯料参数快速预测的方法。选择应用广泛的三通管件为例,以管坯长度、管坯壁厚和模具过渡圆角半径为网络输入参数,壁厚减薄不超过30%时的最大支管长度为输出参数,建立了人工神经网络的预测模型。结合正交设计和有限元数值模拟的思想获取训练、测试样本;在MATLAB软件平台上完成预测模型的训练、测试和管坯参数预测。预测结果和数值分析结果表明,利用Elman神经网络进行多通管的管坯参数优化的方法是可行的。The method of building tubular blank parameters forecasting model for compound bulging multi-pass tubes by using artificial neural network is discussed.The widely used typical tee-pipe fittings are selected to build the fast forecasting model of artificial neutral network,in which the tubular blank parameters of tubular length,tubular thickness and die knuckle radius are selected as network inputs,and the max branch length as the thickness reduction less than 30 percents is selected as the network output.By combining the method of orthogonal test and the finite element numerical simulation,training samples and testing samples are obtained.The training,testing for the forecasting model and the tubular blank parameters forecasting are completed on the software platform of MATLAB.The results of forecasting and the numerical analyzing verify that the Elman neural network is capable of optimizing the tube blank parameters of compound bulging multi-pass tubes.

关 键 词：管坯参数 有限元数值模拟 ELMAN神经网络 参数优化 

分 类 号：TH16[机械工程—机械制造及自动化] TG335.83[金属学及工艺—金属压力加工]