Experimental and artificial intelligence approaches for predicting the strengths of ternary blended concrete incorporating recyclates

Oyebisi, Solomon and Shammas, Mahaad I and Ikotun, Jacob (2025) Experimental and artificial intelligence approaches for predicting the strengths of ternary blended concrete incorporating recyclates. Discover Applied Science, 7 (1381). pp. 1-28. ISSN 3004-9261

Text - Published Version Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (3MB)

Abstract

Predicting concrete strengths with traditional techniques, primarily based on empirical equations and statistical analysis, is deficient in accuracy and efficiency due
to the complexity and nonlinear relationship between concrete mix proportions and strength. This study uses artificial intelligence techniques (gene expression
programming (GEP) and deep neural networks (DNN)) to forecast the compressive strength (CS), flexural strength (FS), and split tensile strength (STS) of recyclate-based ternary blended concrete (TBC). The recyclates, such as corncob ash (CCA) and ground oyster seashells (GOS), were prepared and used as Portland limestone cement (PLC) replacements at 5–15 wt% for TBC production. The concrete samples were made and tested for CS, FS, and STS after 3, 7, 28, 60, 90, and 120 curing days. For learning purposes, 70% of the experimental datasets were allocated for training, while the other 30% of samples were designated for validation and testing. The experimental results revealed optimally improved strengths at 10 wt% CCA and
10 wt% GOS replacements with PLC. The 8-20-20-20-1 DNN structures exhibited a strong correlation for predicting the CS with 99.98% R as well as the 8-25-25-25-1 for forecasting the FS and STS with 99 and 97.06% R. For training and validation, GEP models yielded 99.61 and 99.33% R values for predicting CS, 99.61 and 99.33% R values for predicting FS, and 99.61 and 99.33% R values for predicting STS. The
validity of the developed DNN and GEP models with raw experimental datasets produced a strong correlation. However, the DNN technique outperformed the GEP model. A sensitivity analysis indicated that the curing age was the most significant parameter in predicting the CS of TBC incorporating CCA and GOS. The deployment of DNN and GEP in the building and construction sector can help obtain more
accurate strength analysis of recyclate-based TBC.

Item Type:	Journal Article
Uncontrolled Keywords:	Artificial intelligence, Compressive strength, Modelling, Recyclates, Recycling, Ternary blended concrete
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) T Technology > TH Building construction
Divisions:	Faculty of Science, Technology and Environment (FSTE) > School of Engineering and Physics
Depositing User:	Solomon Oyebisi
Date Deposited:	07 Dec 2025 23:43
Last Modified:	07 Dec 2025 23:43
URI:	https://repository.usp.ac.fj/id/eprint/15207

Actions (login required)

View Item