Modelling of the Mechanical Properties of High-Performance Concrete using Artificial Neural Networks

This study harnessed the possibilities of artificial neural networks (ANNs), through the prediction modelling of high-performance concrete and its behaviour, especially in terms of compressive, flexural, and split tensile strengths. This study evaluated concrete components that contained 6 phases and 24 mix series, after specimens had been subjected to 7, 21, and 28-day curing regimen. Thereafter, the ANN was meticulously trained with 7 in-put parameters, representing the materials and mix variations, while also configuring 5 output parameters for the experimental results. The ANN's performance displayed high accuracy that had good Mean Squared Error (MSE), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE) values. These also served as a further validation tool for the results obtained. With a regression, R value of 0.9825 and coefficient of determination (R2) of approximately 96.53%, the ANN's capacity for concrete strength prediction stands con-firmed. The study concludes that the ANN accurately predicted values which were very close to experimental data, especially compressive and flexural strengths, thereby paving the way for applications in other fields beyond concrete production. However, the flexural strength values did not meet minimum strength requirements.