基于Elman神经网络的三元乙丙橡胶胶料动态粘弹性能预测  

Dynamic Viscoelastic Properties Prediction of EPDM Compound Based on Elman Neural Network

作　　者：曾宪奎[1] 李营如 黄年昌 张杰[1] 鲍丽苹 ZENG Xiankui;LI Yingru;HUANG Nianchang;ZHANG Jie;BAO Liping(Qingdao University of Science and Technology,Qingdao 266061,China)

机构地区：[1]青岛科技大学机电工程学院,山东青岛266061

出　　处：《橡胶工业》2018年第11期1302-1305,共4页China Rubber Industry

基　　金：山东省自然科学基金资助项目(ZR2014EMM018)

摘　　要：研究三元乙丙橡胶(EPDM)胶料配方因子与动态粘弹性能之间的关系,通过建立Elman神经网络模型预测EPDM胶料在85℃,0. 1Hz,15%应变下的动态粘弹性能。通过正交试验设计18组数据样本,利用1—14号样本训练Elman神经网络、15—18号样本验证Elman神经网络的预测能力。结果表明:Elman神经网络对EPDM胶料动态粘弹性能预测结果的相对误差基本在4%以内,具有较高的预测精度。The relationship between the formula factors of EPDM compound and dynamic viscoelastic properties was investigated. The Elman neural network model was established to predict the dynamic viscoelastic properties of EPDM compound at 85 ℃,0. 1 Hz and 15% strain. Data were collected from 18 sets of samples by orthogonal experiment design,Elman neural network was trained with sample No. 1—No. 14,and the predictive ability of Elman neural network was verified by sample No. 15—No. 18. It was found that the relative error of dynamic viscoelastic properties of the EPDM compound predicted by Elman neural network was less than 4%,which demonstrated that Elman neural network model had high prediction accuracy.

关 键 词：ELMAN神经网络 三元乙丙橡胶 动态粘弹性能 样本 预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TQ333.4[自动化与计算机技术—控制科学与工程]