Prediction of flexural strength of concrete with eggshell and glass powders: Advanced cutting-edge approach for sustainable materials

Abstract

Currently, there is a lack of research comparing the efficacy of machine learning and response surface methods in predicting flexural strength of Concrete with Eggshell and Glass Powders. This research aims to predict and simulate the flexural strengths of concrete that replaces cement and fine aggregate with waste materials such as eggshell powder (ESP) and waste glass powder (WGP). The response surface methodology (RSM) and artificial neural network (ANN) techniques are used. A dataset comprising previously published research was used to assess predictive and generalization abilities of the ANN and RSM. A total of 225 research article samples were collected and split into three subsets for model development: 70% for training (157 samples), 15% for validation (34 samples), and 15% for testing (34 samples). ANN used seven independent variables to model and improve the model, whereas RSM used three variables (cement, WGP, and ESP) to improve the model. The k-fold cross-validation validated the generalizability of the model, and the statistical metrics demonstrated favorable outcomes. Both ANN and RSM techniques are effective instruments for predicting flexural strength, according to the statistical results, which include the mean squared error, determination coefficient (R 2), and adjusted coefficient (R 2 adj). RSM was able to achieve an R 2 of 0.7532 for flexural strength, whereas the accuracy of the results for ANN was 0.956 for flexural strength. Moreover, the correlation between the ANN and RSM models and the experimental data was high. However, the ANN model exhibited superior accuracy.

Keywords

• waste glass powder
• eggshell powder
• artificial neural networks
• flexural strengths
• response surface methodology