Analysis of Different Deep Learning Networks for Estimating the Tensile Strength of Self-Compacting Concrete Containing Recycled Aggregates

Jesús de Prado Gil Prado Gil; Rebeca Martínez-García; Víctor Baladrón-Blanco; Covadonga Palencia; Pablo Gutiérrez-Rodriguez; Jesús Lozano-Arias; José Francisco Fernández-Órdas; Fernando J. Fraile-Fernández

doi:10.37934/arca.38.1.1222

Authors

Jesús de Prado Gil Department of Mining Technology, Topography, and Structures, University of León. Campus of Vegazana s/n, 24071 León, Spain
Rebeca Martínez-García Department of Mining Technology, Topography, and Structures, University of León. Campus of Vegazana s/n, 24071 León, Spain
Víctor Baladrón-Blanco Department of Mining Technology, Topography, and Structures, University of León. Campus of Vegazana s/n, 24071 León, Spain
Covadonga Palencia Department of Applied Physics, Campus de Vegazana s/n, University of León, 24071 León, Spain
Pablo Gutiérrez-Rodriguez Department of Management and Business Economics, University of León. Campus of Vegazana s/n, 24071 León, Spain
Jesús Lozano-Arias Department of Electrical and Systems Engineering and Automatics, University of León. Campus of Vegazana s/n, 24071 León, Spain
José Francisco Fernández-Órdas Department of Mining Technology, Topography, and Structures, University of León. Campus of Vegazana s/n, 24071 León, Spain
Fernando J. Fraile-Fernández Department of Mining Technology, Topography, and Structures, University of León. Campus of Vegazana s/n, 24071 León, Spain

DOI:

https://doi.org/10.37934/arca.38.1.1222

Keywords:

Machine-learning, Concrete, Tensile strength, SCC, recycling, materials

Abstract

This writing evaluates and contrasts the performance of various deep learning models in predicting the tensile strength of self-compacting concrete incorporating recycled aggregates. Experimental data sourced from existing literature were used to create test, training, and validation sets. A range of artificial intelligence models and optimization algorithms were explored to train these networks, with adjustments made to their architectures and parameters. The results revealed patterns that offer valuable insights into the relative efficacy of the models, advancing the understanding of how deep learning can be applied to predict concrete properties. This study serves as a strong reference point for both researchers and professionals in the building industry.