Experimental and deep neural network approaches on strength evaluation of ternary blended concrete

Oyebisi, Solomon and Alomayri, Thamer (2024) Experimental and deep neural network approaches on strength evaluation of ternary blended concrete. Experimental and deep neural network approaches on strength evaluation of ternary blended concrete, 439 . pp. 1-17. ISSN 0950-0618

Text Download (10MB)

Abstract

The manufacture of cement contributes significantly to carbon dioxide emissions; hence, the building and construction industry has focused on environmentally friendly cement substitutes. Supplementary cementitious materials (SCMs) such as calcite powder (CP) and Vitellaria paradoxa ash (VPA) offers sustainable substitutes. Thus, this study calcined shea nutshell at 700 °C for 3 h, obtaining VPA. Portland limestone cement was partially replaced by calcite powder and VPA at 5–15 wt% to produce 25 and 30 MPa concrete grades. Split tensile strength (STS), flexural strength (FS), and compressive strength (CS) of TBC samples were tested after 3–120 days of curing. Deep neural network (DNN) models, having 3 hidden layers with 5–30 nodes, were engaged to predict the strengths with respect to the concrete mix design proportions. For each strength, 108 datasets were obtained from the experimental data. Out of these values, 100 datasets were utilized to train the models, and the remaining 8 values were used to confirm the model's accuracy. The results revealed an improvement in concrete’s CS, FS, and STS at 10 % VPA and 10 % CP replacements. The 7–25–25–25-1 network topologies demonstrated robust correlation for training, validating, and testing the input and output variables of CS and FS with correlation coefficients (R) of 99.92 and 99.01 % compared to other architectures. However, 7-20–20–20-1 network structure exhibited the best performance metric for predicting the STS of TBC with 99.51 % R. Strong relationships were found between the created model's validity and the raw experimental datasets, with R2 values for CS, FS, and STS yielding 98.45, 99.75, and 99.35 %. By using this technique, TBC incorporating SCMs would be of higher quality.

Item Type:	Journal Article
Uncontrolled Keywords:	tensile strength (STS), flexural strength
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) T Technology > TD Environmental technology. Sanitary engineering T Technology > TH Building construction
Divisions:	Faculty of Science, Technology and Environment (FSTE) > School of Engineering and Physics
Depositing User:	Solomon Oyebisi
Date Deposited:	18 Jul 2025 00:26
Last Modified:	18 Jul 2025 00:26
URI:	https://repository.usp.ac.fj/id/eprint/15015

Actions (login required)

View Item