Investigation of mechanical properties of high-performance concrete via multi-method of regression tree approach

Concrete's workability and durability are influenced by its mechanical properties, including Tensile and Compressive strength. High-performance concrete exhibits non-linear relationships between its compressive and tensile strength characteristics and the proportions of its constituent material...

Full description

Saved in:
Bibliographic Details
Published in:Materials today communications Vol. 40; p. 109922
Main Authors: Qi, Rui, Wu, Haiyan, Qi, Yongjun, Tang, HaiLin
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.08.2024
Subjects:
Compressive and tensile strength
Decision tree
Ensemble learning
High-performance concrete
Meta-heuristic algorithms
B
W/B
High-performance concrete
Decision tree
DTCA
DT
DAOA
AOA
SF
COA
SI
DOA
W
Ensemble learning
SP
MK
DTAO
R2
SHAP
Meta-heuristic algorithms
RMSE
DTDA
Compressive and tensile strength
CS
MAE
NMSE
HPC
FA
DTDO
DTDDAC
Ca
TS
ISSN:2352-4928, 2352-4928
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Concrete's workability and durability are influenced by its mechanical properties, including Tensile and Compressive strength. High-performance concrete exhibits non-linear relationships between its compressive and tensile strength characteristics and the proportions of its constituent materials, such as water, aggregate, cement, and admixtures. The investigation of the relations between the constituents of concrete and its mechanical properties has posed a challenging issue. This paper seeks to develop a range of single, hybrid, and ensemble models to resolve the problem of accurately estimating the mechanical properties of concrete according to its constituent components as input variables. In this regard, various frameworks incorporating a machine learning technique called decision tree and four metaheuristic optimization algorithms were utilized in model development to emulate the values of tensile and compressive strength. The findings indicated that the decision tree-dynamic arithmetic optimization algorithm hybrid model produced results with an correlation of determination of 0.9919 in compressive strength and 0.9933 in tensile strength, which were on average 1–3 % higher than that of the decision tree-dandelion optimizer algorithm, decision tree-arithmetic optimization algorithm, and decision tree-coot optimization algorithm, indicating superior optimization performance of dynamic arithmetic optimization algorithm. Additionally, the powerful prediction capability of the decision tree + dynamic arithmetic optimization algorithm + dandelion optimizer algorithm + coot optimization algorithm + arithmetic optimization algorithm ensemble model was noticeable with R2 of 0.9882 and 0.9936 in compressive and tensile strength estimation reliable to be used for various data produced in the future. In general, the utilization of coupling techniques to generate hybrid and ensemble models can enhance the accuracy of predictions while reducing the time and costs associated with experimental procedures. [Display omitted]
ISSN:2352-4928
2352-4928
DOI:10.1016/j.mtcomm.2024.109922