Advanced deep learning and machine learning method for predicting uniaxial compressive strength in recycled concrete

Yaser A. Nanehkaran , Yuan Xiaofeng , Tolga Pusatli , Yimin Mao

Abstract

Accurate prediction of uniaxial compressive strength (UCS) in recycled concrete aggregate (RCA) is essential for advancing sustainable construction practices. This study presents a data-driven predictive framework based on Deep Neural Networks (DNN) to estimate UCS using a comprehensive dataset comprising 326 literaturederived records and 50 experimentally validated samples. The model was trained using optimized hyperparameters over 1000 epochs and evaluated through a combination of statistical metrics and cross-validation techniques. The proposed DNN model demonstrated superior predictive performance compared to benchmark regression models, including Multilayer Perceptron (MLP), Support Vector Machine (SVM), and Decision Tree (DT), achieving an accuracy of 0.925 with significantly reduced error values. The robustness of the model was further validated using independent experimental data, confirming its generalization capability in real-world conditions. The novelty of this study lies in integrating heterogeneous data sources with a rigorous validation framework, enhancing both reliability and applicability. The findings highlight the potential of deep learning techniques to support efficient material design and decision-making in recycled concrete applications, contributing to more sustainable and resource-efficient construction practices.

Key Words

compressive strength; concrete aggregate; deep learning; recycled concrete; sustainable construction

Address

Yaser A. Nanehkaran, Yuan Xiaofeng — School of Artificial Intelligence, Yancheng Teachers University, Yancheng 224002, Jiangsu, China
Tolga Pusatli — Department of Management Information Systems, Cankaya University, Ankara, Türkiye
Yimin Mao — School of Information and Engineering, Shaoguan University, Shaoguan 512005, Guangdong, China

PDF Viewer

Preview is limited to the first 3 pages. Sign in to access the full PDF.

← Back to Volume 37, No. 6