Optimizing Waste Foundry Sand in Concrete Considering Strength Properties for Sustainable Green Structures

Sustainable Green Structures Waste Foundry Sand Concrete Strength Machine Learning

Authors

  • Néstor Ulloa
    nestor.ulloa@espoch.edu.ec
    1) Facultad de Mecánica, Escuela Superior Politécnica de Chimborazo (ESPOCH), Riobamba 060155, Ecuador. 2) Grupo de Investigación y Desarrollo de Nanotecnología, Materiales y Manufactura (GIDENM), Escuela Superior Politécnica de Chimborazo, ESPOCH, Riobamba, Ecuador https://orcid.org/0000-0001-7819-6670
  • Kerly Mishell Vaca Vallejo Dipartimento Di Ingegneria Informatica, Modellistica, Elettronica E Sistemistica-DIMES, University of Calabria, Rende, 87036, Italy
  • Ana Marí­a Bucheli Campaña Escuela Superior Politécnica de Chimborazo (ESPOCH), Sede Orellana, El Coca 220150, Ecuador
  • Mery Mendoza Castillo Escuela Superior Politécnica de Chimborazo (ESPOCH), Sede Orellana, El Coca 220150, Ecuador
  • Byron Gabriel Vaca Vallejo Facultad de Informática y Electrónica, Escuela Superior Politécnica de Chimborazo (ESPOCH), Riobamba, 060155, Ecuador

Downloads

Incorporating waste foundry sand (WFS) into concrete is a sustainable approach to enhance green construction practices. Waste foundry sand is a byproduct of the metal casting industry and is often discarded in landfills, posing environmental concerns. Using it as a partial replacement for natural sand in concrete addresses both waste management and resource conservation. In this research paper, advanced machine learning models have been reported on the soft computing of the optimal waste foundry sand in concrete based on strength properties for sustainable green structures. The machine learning techniques such as “Group Methods Data Handling Neural Network (GMDH-NN)”, “Support Vector Machine (SVM)”, “K-Nearest Neighbors (KNN)”, “Tree Decision (Tree)” and “Random Forest (RF)” were applied on a database for the compressive strength containing 397 records, for elastic modulus containing 146 records, and for split tensile strength containing 242 records. Each record contains C-Cement content (kg/m³), WFS-Waste foundry sand content (kg/m³), W-Water content (kg/m³), SP-Super-plasticizer content (kg/m³), CA-Coarse aggregates content (kg/m³), FA-Fine aggregates content (kg/m³), TA-Total aggregates content (kg/m³), and Age-The concrete age at testing (days), considered as the input parameters and CS_WFS-Compressive strength of waste foundry sand concrete (MPa), E_WFS-Elastic modules of waste foundry sand concrete (GPa), and STS_WFS-Split tensile strength of waste foundry sand concrete (MPa), which are the output parameters. A 75/25 partitioning pattern for train/test of the database was used in line with established rules. At the end of the model operation, it can be observed that kNN, SVM, and RF were paramount in terms of performance and therefore outclassed the other models in the three-state strength condition of the WFS cement concrete. Hence, these were selected as the decisive models for the prediction of the compressive strength, elastic modulus, and splitting tensile strength of the WFS cement's concrete. The sensitivity analyses showed that Age, WFS/C and CA/C are more impactful on the compressive strength, Age, FA/TA, and W/C are more impactful on the elastic modulus; and 1000SP/C, WFS/C, and W/C are more impactful on the splitting tensile strength of the WFS cement concrete. Generally, these models provide a foundation for optimizing material use, ensuring quality, and meeting environmental goals. Industries leveraging these tools can produce eco-friendly, high-performance concrete while addressing waste management challenges and reducing their carbon footprint.