A data-driven approach to predicting multifactor-influenced flexural size effect and fracture behaviors of concrete

作     者：Ye, Junhong Uddin, Md Nasir Yu, Jie Xu, Tengfei Zhan, Yulin Zhang, Dong Weng, Yiwei 

作者机构：Southwest Jiaotong Univ Inst Smart City & Intelligent Transportat Chengdu Peoples R China Hong Kong Polytech Univ Dept Bldg & Real Estate Hong Kong Peoples R China Southwest Jiaotong Univ Dept Bridge Engn Chengdu Peoples R China Fuzhou Univ Coll Civil Engn Fuzhou Peoples R China Hong Kong Polytech Univ Shenzhen Res Inst Shenzhen Peoples R China 

出 版 物：《ENGINEERING FRACTURE MECHANICS》 (Eng. Fract. Mech.)

年 卷 期：2025年第315卷

核心收录：

学科分类：08[工学] 0801[工学-力学（可授工学、理学学位）] 

基　　金：National Natural Science Foundation of China Guangdong Basic and Applied Basic Research Foundation [2024A1515011870] Hong Kong Polytechnic University [P0038966, P0046543] 

主　　题：Size effect Machine learning Flexural strength Fracture toughness Gene expression programming 

摘      要：Flexural size effect, originating from the fracture characteristics of materials, is a common phenomenon in concrete. Conventionally, time-consuming and labor-intensive experiments are required to investigate the flexural size effect and fracture behaviors of concrete. To tackle the limitations, a data-driven approach was adopted to predict the multifactor-influenced flexural size effect and fracture behaviors of concrete by gene expression programming (GEP) due to its capability of addressing non-linear problems and developing empirical equations with multiple input variables. Results show that the GEP models can accurately predict nominal flexural strength (R2, 0.890) and fracture toughness (R2, 0.946). Parametric analysis reveals that the compressive strength and tensile strain capacity positively impact the nominal flexural strength and fracture toughness of concrete. Based on the GEP model, a multifactor-influenced size effect law (SEL) is proposed to predict the nominal flexural strength by incorporating both material and geometric parameters, removing the need for extensive experimental investigations. The findings provide generalized models to predict the nominal flexural strength and fracture toughness of various materials at different sizes.