摘要
为实现路用水性环氧树脂力学参数极值寻优,以进一步科学优化其材料组成及配比,基于不同类型原材料自身属性及其掺量的调控,制备了一系列路用水性环氧树脂;以拉伸强度为例,开展了力学参数标定试验并获取了75组有效数据,以此构建了基于小样本数据的传统多元回归方程、典型前馈式及反馈式神经网络等不同预测模型,对比分析了不同模型的拟合效果与预测精度;采用遗传算法对优选的预测模型进行优化,实现了拉伸强度极值寻优功效,回溯确定了原材料各项属性及掺量的最佳取值,为路用水性环氧树脂材料组成及配比的科学优化提供有益借鉴。结果表明:相比五元非线性回归方程和反向传播(Back Propagation,BP)模型,Elman模型的预测误差降低了25.94%~37.5%,具有较好的预测精度和稳定性;综合考虑预测精度和运算速度,推荐采用Elman神经网络(6×4×1)作为路用水性环氧树脂拉伸强度的预测模型;采用遗传算法可快速获得最优拉伸强度和相关材料组成及配比,降低试验成本,但其准确度仍需进一步优化。
The purpose is to obtain the extreme value of mechanical parameters of road-use waterborne epoxy resin for further scientific optimization of its material composition and proportion.According to the adjustment of self-properties and proportion of different types of raw materials,a series of road-use waterborne epoxy resins were prepared.Taking tensile strength as an example,the mechanical parameter calibration test was carried out and 75 groups of effective data were obtained.Based on small sample data,different prediction models such as traditional multiple regression analysis,typical feedforward and feedback neural networks were established.The fitting effect and prediction accuracy of different mo-dels were compared and analyzed.Genetic algorithm and particle swarm optimization were used to optimize the selected prediction model,and the optimization effect of tensile strength was realized.The optimum values of property and proportion of raw materials were determined retrospectively.It provided a useful reference for scientific optimization of composition and proportion of road-use waterborne epoxy resin.The results are as follow.Compared with the nonlinear regression equation with five independent variable and back propagation(BP)model,the prediction error of Elman model is 25.94%—37.5%lower.It has significant prediction accuracy and stability.According to prediction accuracy and operation speed,Elman neural network(6×4×1)is recommended as the prediction model of tensile strength of waterborne epoxy resin for road.Genetic algorithm can quickly obtain the optimal tensile strength and its related material composition and proportion,and reduce test cost.But its accuracy still needs to be further optimized.
作者
陈谦
王朝辉
傅豪
樊振通
刘鲁清
CHEN Qian;WANG Chaohui;FU Hao;FAN Zhentong;LIU Luqing(School of Highway,Chang’an University,Xi’an 710064,China)
出处
《材料导报》
EI
CAS
CSCD
北大核心
2021年第16期16172-16177,共6页
Materials Reports
基金
山东省交通运输厅科技计划项目(2018B50)
中央高校基本科研业务费专项资金(300102219314,300102219701)。
关键词
道路材料
水性环氧树脂
力学参数
拉伸强度
小样本
极值寻优
road material
waterborne epoxy resin
mechanical parameters
tensile strength
small sample
extreme value optimization
作者简介
陈谦,长安大学公路学院,博士研究生,主要从事绿色道路新材料及新技术的研究,发表学术论文23篇,其中SCI检索16篇,获国家授权专利8项;王朝辉,长安大学公路学院教授、博士研究生导师,交通运输科技青年英才,美国俄克拉荷马州立大学访问学者,国际稀浆罩面协会专家委员会委员。2008年7月毕业于长安大学,获工学博士学位。同年加入长安大学公路学院道路研究所工作至今,主要从事绿色道路新材料与新技术的开发及应用、道路预防性养护与决策优化技术等领域的研究。以第一作者/通讯作者发表学术论文110余篇,其中SCI/EI 70余篇,获国家授权发明专利70余项,wchh0205@chd.edu.cn。
