Hybrid machine learning for elastic modulus prediction in recycled aggregate concrete

被引：0

作者：

Linhua Huang ^{[1
]}

Hongna Cui ^{[2
]}

Song Li ^{[3
]}

Enping Guo ^{[4
]}

机构：

[1] Hunan University of Science and Engineering,Hunan Provincial Key Laboratory of Intelligent Protection and Utilization Technology in Masonry Artifacts

[2] Hunan University of Science and Engineering,School of Civil and Environmental Engineering

[3] Hebei University of Architecture,School of Science

[4] Hebei University of Architecture,Zhangjiakou City Key Laboratory of Engineering Mechanics Analysis

来源：

Multiscale and Multidisciplinary Modeling, Experiments and Design | 2024年 / 8卷 / 7期

关键词：

RAC; Concrete; Elastic modulus; Compressive strength; Machine learning algorithms; Regression analysis;

D O I：

10.1007/s41939-025-00909-w

中图分类号：

学科分类号：

摘要：

The elastic modulus of recycled aggregate concrete (RAC) is a basic mechanical property indicating the stiffness as well as the deformation behavior of a material under stress. The precise prediction of this property is very important in designing sustainable and durable concrete structures, especially where utilizing recycling materials is involved, as they introduce mechanical variability. The paper discusses a machine learning-based method of projecting the elastic modulus of RAC with good accuracy. A multi-layer perceptron regression (MLPR) model was utilized and improved with two different optimization algorithms: the Mayfly Optimization Algorithm (MOA) and the Chef-Based Optimization Algorithm (COA). Two hybrid schemes were formulated and tested. Out of them, the one optimized by the Mayfly achieved the highest possible accuracy with an R2 of 0.962 on the test database. These findings confirm the power of combining machine learning strategies to capture complicated relationships between the input variables as well as the elastic modulus, presenting a trusted and applicable tool for encouraging the employment of RAC in sustainable construction.

引用