寒冷地区腈纶纤维混凝土力学性能及多层感知器神经网络预测  

Mechanical Properties of Polyacrylic Fiber Concrete in Cold Areas and Prediction by Multilayer Perceptron Neural Network

作　　者：段明翰 覃源[1] 李阳[1] 耿凯强 DUAN Minghan;QIN Yuan;LI Yang;GENG Kaiqiang(State Key Laboratory of Water Engineering Ecology and Environment in Arid Area,Xi’an University of Technology,Xi’an 710048,China;School of Water Conservancy and Construction Engineering,Shihezi University,Shihezi 832000,Xinjiang,China)

机构地区：[1]西安理工大学旱区水工程生态环境全国重点实验室,西安710048 [2]石河子大学水利建筑工程学院,新疆石河子832000

出　　处：《材料导报》2025年第6期105-113,共9页Materials Reports

基　　金：国家自然科学基金面上项目(52279139)。

摘　　要：为延长寒冷地区混凝土材料的服役寿命,降低水利工程建设和维护过程中的资源损耗,通过添加腈纶纤维来制备腈纶纤维增强混凝土(PAN-FRC)。本工作主要探讨了寒冷气候条件下,养护方式、龄期及纤维掺量对PANFRC的动弹性模量、表面回弹硬度、抗压强度及劈裂抗拉强度等力学性能的影响。结果表明:冬季室外养护容易发生冻害,28 d龄期难以达到抗压强度设计值,纤维掺量在1.5~1.8 kg/m^(3)时对混凝土有增强效果,且抗拉性能提升效果最优,有望改善其抗裂能力。同时建立了适用于PANFRC的力学指标转换数学模型。此外,基于试验数据构建了关于PANFRC压、拉性能的神经网络预测模型(RBFN及MLPN),模型精度评估结果表明MLPN优于RBFN,测试集中两个模型的抗压强度和劈裂抗拉强度预测绝对误差分别可控制在3 MPa及0.17 MPa以内,相对误差分别可控制在9%及6%以内,相关系数R^(2)均在0.9以上。研究结果可为PANFRC进一步应用于寒冷地区渠道衬砌及其他水利设施建设提供理论依据。To extend the service life of concrete materials in cold regions and minimize resource expenditure during the construction and maintenance of water conservancy projects,polyacrylonitrile fiber reinforced concrete(PANFRC)has been developed through the incorporation of polyacry-lonitrile fibers.This work investigated the influence of curing methods,curing ages,and fiber content on dynamic elastic modulus,surface re-bound hardness,compressive strength,and splitting strength of PANFRC under cold climate conditions.The findings indicate that concrete cured in outdoor in cold region winter is susceptible to frost damage to hardly achieve the target compressive strength at 28 days.A fiber content of 1.5-1.8 kg/m^(3)can enhance the concrete’s properties,especially for tensile performance,and improve its crack resistance.Moreover,based on existing empirical correlations,a mathematical model for converting mechanical indices suitable for PANFRC had been established.Furthermore,based on experimental data,neural network prediction models(RBFN and MLPN)for the compressive and tensile properties of PANFRC had been developed.Model accuracy evaluations reveal that the MLPN model outperforms the RBFN model here.The absolute errors for com-pressive strength and splitting tensile strength predictions in the test set are confined to within 3 MPa and 0.17 MPa,respectively,with relative er-rors kept below 9%and 6%,respectively,and the correlation coefficient(R^(2))exceeding 0.9.This research provides a theoretical foundation for the broader application of PANFRC in canal lining and other water conservancy infrastructure construction in cold regions.

关 键 词：腈轮纤维混凝土 力学性能 强度预测 神经网络 渠道衬砌 

分 类 号：TV43[水利工程—水工结构工程]