Enhancing Compressive Strength Prediction in Recycled Aggregate Concrete through Robust Hybrid Machine Learning Approaches

This research study delves into the domain of civil engineering, specifically focusing on the prediction of compressive strength (f'c) in Recycled Aggregate Concrete (RAC). As the construction industry seeks sustainable solutions, RAC has gained prominence due to its environmentally friendly na...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Advances in Engineering and Intelligence Systems Ročník 4; číslo 1; s. 21 - 37
Hlavní autoři: Samuel Keown, Dylan O’Dwyer
Médium: Journal Article
Jazyk:angličtina
Vydáno: Bilijipub publisher 01.12.2025
Témata:
compressive strength
dingo optimization algorithm
giant trevally optimizer
least square support vector regression
recycled aggregate concrete
ISSN:2821-0263
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:This research study delves into the domain of civil engineering, specifically focusing on the prediction of compressive strength (f'c) in Recycled Aggregate Concrete (RAC). As the construction industry seeks sustainable solutions, RAC has gained prominence due to its environmentally friendly nature. However, accurately predicting the f'c of RAC remains a complex challenge, owing to the inherent variability of recycled materials. To address this issue, robust hybrid machine learning (ML) approaches are employed, particularly emphasizing the Least Square Support Vector Regression (LSSVR) model. This investigation explores the integration of LSSVR with two innovative optimizers, namely the Giant Trevally Optimizer (GTO) and the Dingo Optimization Algorithm (DOA). The performance of the LSSVR model is enhanced through the utilization of these optimizers, resulting in its increased proficiency in predicting the f'c of RAC. Through comprehensive experimentation and rigorous analysis, this research showcases the effectiveness of the proposed hybrid method. The findings illustrate significant improvements in the accuracy and reliability of f'c predictions for RAC compared to traditional methods. Moreover, this investigation provides significant perspectives on the synergy between ML and optimization techniques in civil engineering. The most reliable outcomes in this study were achieved through the hybridization of the LSSVR model with GTO. The resulting LSGT model attained the highest R2 value of 0.989 and the lowest RMSE value of 1.618. Overall, the integration of LSSVR with GTO and DOA emerges as a promising methodology to enhance f'c prediction in RAC. This research enhances the current understanding of RAC and highlights its potential for robust hybrid machine-learning approaches in solving real-world challenges within civil engineering.
ISSN:2821-0263
DOI:10.22034/aeis.2025.505261.1286