RESUMO
Waste Foundry sand (WFS), a major solid waste from metal casting industry, is posing a significant environmental threat owing to its disposal to landfills. In this research, an innovative artificial intelligence technique i.e. Multi-Expression Programming (MEP) is applied to model the split tensile strength (ST) and modulus of elasticity (E) of concrete containing waste foundry sand (CWFS). The presented formulations correlate mechanical properties with four input variables i.e. w/c, foundry sand content, superplasticizer content and compressive strength. The results of statistical analysis validate the model accuracy as evident by the low values of objective function (0.033 for E and 0.052 for ST). Moreover, the average error in the predicted values is significantly low i.e. 0.287 MPa and 1.75 GPa for ST and E model, respectively. Parametric study depicts that the models are well trained to accurately predict the trends of mechanical properties with variation in mix parameters. The prediction models can promote the usage of WFS in green concrete thereby preventing waste disposal and contributing towards and sustainable construction.
RESUMO
Waste foundry sand (WFS) is a major pollutant generated from metal casting foundries and is classified as a hazardous material due to the presence of organic and inorganic pollutants which can cause adverse environmental impact. In order to promote the re-utilization of WFS, gene expression programming (GEP) has been employed in this study to develop empirical models for prediction of mechanical properties of concrete made with WFS (CMWFS). An extensive and reliable database of mechanical properties of CMWFS is established through a comprehensive literature review. The database comprises of 234 compressive strength, 163 split tensile strength and 85 elastic modulus results. The four most influential parameters i.e. water-to-cement ratio, WFS percentage, WFS-to-cement content ratio and fineness modulus of WFS are considered as the input parameters for modelling. The mechanical properties can be estimated by the application of proposed simplified mathematical expressions. The performance of the models is assessed by conducting parametric analysis, applying statistical checks and comparing with regression models. The results reflected that the proposed models are accurate and possess a high generalization and prediction capability. The findings of this study can enhance the re-usage of WFS for development of green concrete leading to environmental protection and monetary benefits.