摘要
为实现对粗骨料UHPC的抗压强度的预测和配合比设计方法的优化,搜集了国内外文献中168组粗骨料UHPC配合比和标准养护28 d抗压强度实测值,给出了各材料组分和抗压强度频数分布,并基于灰色关联分析法分析了各材料组分与抗压强度的关联关系,通过神经网络参数分析,建立了基于遗传算法的前馈神经网络,相比普通的BP神经网络具有更好的预测精度和泛化能力。最后基于建立的GA-BP神经网络给出了不同强度等级粗骨料UHPC配合比设计中粗骨料/胶凝材料、钢纤维体积掺量、砂胶比的建议取值范围。
In order to predict the compressive strength and to optimize the mix design of ultra high performance concrete(UHPC)incorporated with coarse aggregates,168 mix proportions and corresponding compressive strength of UHPC after 28 days standard curing were collected from domestic and foreign literature.Charts of frequency distribution of each material component and compressive strength were given,and correlations between each material component and compressive strength were analyzed based on grey relational analysis(GRA)method.Through parameter analysis,the back propagation(BP)neural network based on genetic algorithm(GA)was established,which achieved better prediction accuracy and generalization ability than common BP neural network.Finally,based on the established GA-BP neural network,the recommended ranges of coarse aggregates-binder ratio,volume fraction of steel fiber and sand-binder ratio were given in the mix design for preparing UHPC incorporated with coarse aggregates of different strength grades.
作者
周靖宜
蔡自伟
李凌志
俞可权
ZHOU Jingyi;CAI Ziwei;LI Lingzhi;YU Kequan(College of Civil Engineering,Tongji University,Shanghai 200092,China)
出处
《混凝土》
CAS
北大核心
2024年第2期11-19,共9页
Concrete
基金
国家自然科学基金(51778497,51778496)。
关键词
超高性能混凝土
抗压强度
粗骨料
前馈神经网络
遗传算法
ultra high performance concrete(UHPC)
compressive strength
coarse aggregates
back propagation neural network
genetic algorithm