欢迎访问《硅酸盐通报》官方网站,今天是

硅酸盐通报 ›› 2026, Vol. 45 ›› Issue (2): 437-448.DOI: 10.16552/j.cnki.issn1001-1625.2025.0825

• 水泥混凝土 • 上一篇    下一篇

基于砂浆膜厚度的GA-BP神经网络混凝土抗压强度预测研究

余晗之1(), 周理安1,2(), 刘宇1, 周文娟1,2, 梁运健1   

  1. 1.北京建筑大学土木与交通工程学院,北京 100044
    2.北京节能减排与城乡可持续发展省部共建协同创新中心,北京 100044
  • 收稿日期:2025-08-15 修订日期:2025-09-29 出版日期:2026-02-20 发布日期:2026-03-09
  • 通信作者: 周理安,博士,副研究员。E-mail:zla@bucea.edu.cn
  • 作者简介:余晗之(2001—),男,硕士研究生。主要从事建筑垃圾、混凝土的研究。E-mail:864858565@qq.com
  • 基金资助:
    国家重点研发计划(2022YFC3803400)

Prediction of Concrete Compressive Strength by GA-BP Neural Network Based on Mortar Film Thickness

YU Hanzhi1(), ZHOU Li’an1,2(), LIU Yu1, ZHOU Wenjuan1,2, LIANG Yunjian1   

  1. 1. School of Civil and Transportation Engineering,Beijing University of Civil Engineering and Architecture,Beijing 100044,China
    2. Beijing Collaborative Innovation Center for Energy Conservation and Emission Reduction and Urban-Rural Sustainable Development,Beijing 100044,China
  • Received:2025-08-15 Revised:2025-09-29 Published:2026-02-20 Online:2026-03-09

摘要:

为了更加精准地预测现代混凝土抗压强度,以砂浆膜厚度替代混凝土各组分用量作为配合比设计的关键参数。通过理论计算与试验研究,分析砂浆膜厚度对混凝土抗压强度的影响,并采用遗传算法优化反向传播(GA-BP)神经网络分别构建基于传统混凝土配合比和砂浆膜厚度的两种抗压强度预测模型。结果表明,砂浆膜厚度由水泥砂浆总体积和粗骨料参数共同决定,随膜厚度增加,界面过渡区质量先改善后劣化,且粗骨料骨架支撑作用逐步减弱,导致混凝土抗压强度整体呈下降趋势;相较于传统预测模型,基于砂浆膜厚度的预测模型精度提高19.7%,均方根误差降低50.4%,平均相对误差为2.8%,决定系数达0.935。回归评价指标显示该模型可以更精准地预测混凝土抗压强度。

关键词: 混凝土, 砂浆膜厚度, 抗压强度, BP神经网络, 遗传算法

Abstract:

To achieve more precise predictions of modern concrete compressive strength, mortar film thickness was employed as the key parameter in mix design, replacing the quantities of individual concrete constituents. Through theoretical calculations and experimental studies, the influence of mortar film thickness on concrete compressive strength was analysed. Two compressive strength prediction models were constructed: one based on traditional concrete mix proportion and another based on mortar film thickness, utilising a backpropagation optimised by a genetic algorithm (GA-BP) neural network. Results indicate that mortar film thickness is jointly determined by the total volume of cement mortar and coarse aggregate parameters. As film thickness increases, the quality of the interfacial transition zone initially improves before deteriorating, while the skeletal support function of coarse aggregates progressively weakens. This leads to an overall decline in concrete compressive strength. Compared to the traditional prediction model, the mortar film thickness-based model achieves a 19.7% improvement in accuracy, a 50.4% reduction in root mean square error, an average relative error of 2.8%, and a determination coefficient of 0.935. Regression evaluation metrics indicate this model provides more precise predictions of concrete compressive strength.

Key words: concrete, mortar film thickness, compressive strength, BP neural network, genetic algorithm

中图分类号: