Convolutional neural network-based model for predicting the performance of steel fiber reinforced concrete

Journal Information

J. Hefei Univ. Tech. (Nat. Sci.), May.2026, Vol.49 No.5: 711-720

DOI: 10.3969/j.issn.1003-5060.2026.05.019

Authors

ZHU Lingling $ ^{1} $, HONG Li $ ^{1,2,3} $, TAN Yuqing $ ^{1} $, ZHAO Dan $ ^{1} $, LI Hongcen $ ^{1} $

(1. School of Civil and Hydraulic Engineering, Hefei University of Technology, Hefei 230009, China; 2. Anhui Key Laboratory of Civil Engineering Structures and Materials, Hefei 230009, China; 3. Engineering Research Center of Low-carbon Technology and Equipment for Cement-based Materials, Ministry of Education, Hefei 230009, China)

Abstract and Keywords

Abstract: network(BPNN), support vector regression(SVR), random forest(RF), and radial basis function neural network(RBFNN) in terms of accuracy and generalization capability. Finally, the SHapley Additive exPlanations (SHAP) method was employed to scrutinize the influence of each variable on the target performance attributes, namely the slump, compressive strength, and splitting tensile strength. The insights garnered from this study provide a foundational basis for the optimization of steel fiber reinforced concrete mix designs to meet multifaceted performance demands, thereby contributing valuable insights to the domains of engineering practice and structural design.

Keywords: convolutional neural network(CNN); steel fiber reinforced concrete; slump; strength; performance prediction

作者信息

朱玲玲 $ ^{1} $，洪丽 $ ^{1,2,3} $，谭雨晴 $ ^{1} $，赵丹 $ ^{1} $，李虹岑 $ ^{1} $

（1. 合肥工业大学土木与水利工程学院，安徽合肥 230009；2. 土木工程结构与材料安徽省重点实验室，安徽合肥 230009；3. 水泥基材料低碳技术与装备教育部工程研究中心，安徽合肥 230009）

摘要和关键词

摘要: 文章基于卷积神经网络(convolutional neural network, CNN)建立钢纤维混凝土坍落度、抗压强度和劈裂抗拉强度的预测模型。基于文献调研建立了294组钢纤维混凝土材料组成及性能数据的数据库, 以水灰比、砂率、钢纤维体积掺量、钢纤维长径比和减水剂用量作为模型的输入变量, 以钢纤维混凝土的坍落度、抗压强度和劈裂抗拉强度为输出结果, 以均方根误差(root mean squared error, RMSE)、平均绝对误差(mean absolute error, MAE)、平均偏差误差(mean bias error, MBE)和决定系数 $ R^{2} $为评价指标, 衡量预测模型的准确性和可靠性。结果表明, 与反向传播神经网络(back propagation neural network, BPNN)、支持向量回归(support vector regression, SVR)、随机森林(random forest, RF)和径向基神经网络(radial basis function neural network, RBFNN)相比, 基于CNN建立的钢纤维混凝土预测模型精度最高、泛化能力最强。采用SHAP(SHapley Additive ex Planations)方法分析了钢纤维混凝土坍落度、抗压强度和劈裂抗拉强度的敏感因素, 确定每个特征对目标性能的具体影响。该研究结果为后续实现基于多性能需求的钢纤维混凝土配合比优化设计提供依据, 有助于指导工程实践和施工设计。

关键词: 卷积神经网络（CNN）；钢纤维混凝土；坍落度；强度；性能预测

Funding

国家重点研发计划资助项目（2020YFC1909902）

合肥工业大学学报(自然科学版)

Navigation