Artificial Neural Network Predictions for Properties of Lightweight Concrete Incorporating Polystyrene Particles

This study investigates the use of Expanded Polystyrene (EPS) particles in lightweight concrete to evaluate their compressive strength and density. The primary objectives are to assess the impact of varying EPS, cement, sand, and water proportions on these properties and to validate predictive models using Artificial Neural Networks (ANN). A dataset comprising 71 different mixes of concrete having strength up to 45 MPA from previous studies was utilized to build and train the ANN model. Moreover, an experimental program involved preparing and testing 120 different concrete mixes with various EPS contents was conducted. The results showed that increasing the cement content enhances both strength and density, while the higher EPS content provided a negative effect. A comparative study between the experimental results and ANN predictions revealed a close alignment, underscoring the ANN model's reliability and accuracy in predicting concrete properties, with an error margin of less than 1% and strong correlation coefficients of more than 94%. The findings highlight the potential of ANN for optimizing EPS lightweight concrete mix designs, providing a reliable alternative to extensive experimental testing.