Enhancing Concrete Performance: Experimental And Neural Network-Based Optimization Of Admixtures For Improved Workability And Flexural Strength

Abstract

This study explores enhancing concrete performance through experimental and Artificial Neural Network (ANN) optimization techniques. Concrete mixes incorporating five different admixtures—fly ash, blast furnace slag, rice husk ash, silica fume, and eggshell powder—were prepared and analyzed. The methodology involved a comprehensive approach: first, concrete mix designs were developed based on standard calculations to achieve target mean strength and appropriate water-cement ratios. Experimental batches of concrete were then cast and cured, with properties such as workability and flexural strength assessed through slump cone tests and Universal Testing Machines over various curing periods. Concurrently, an ANN model was constructed to predict concrete performance based on the proportions of admixtures. The ANN model was designed with an input layer for admixture proportions, hidden layers for learning complex patterns, and an output layer for predicting workability and flexural strength. The model was trained using the experimental data and validated with a separate dataset to evaluate its predictive accuracy. Although the ANN model faced challenges, with high error metrics and weak correlation between experimental and predicted values, it provided insights into optimal admixture proportions. The study highlights the effectiveness of combining experimental research with neural network-based optimization to develop high-performance and sustainable concrete formulations, advancing the field of concrete technology