Volume 1, Issue 2 — Year 2024 — Article e100014

Hamed Mahmodi¹; Karim Samadzamini²

1. Department of Civil Engineering, University Collage of Nabi Akram (UCNA), Tabriz 5183919611, Iran
2. Department of Computer Engineering, University Collage of Nabi Akram (UCNA), Tabriz 5183919611, Iran

This study investigates the application of artificial intelligence in predicting the uniaxial compressive strength of concrete using experimental data. A multilayer perceptron (MLP) neural network was developed using TensorFlow and Keras in Python program. The dataset includes 150 concrete cube samples tested after 28-days of curing, divided into training (70%) and testing (30%) sets. The results show that the proposed model significantly improves prediction accuracy, achieving over 80% recall and maintaining consistent performance through 500 iterations, with an accuracy range of 80-85%. Comparative analysis with algorithms such as SVM, k-NN, RF, DT, and Adaboost indicates that the MLP model provides superior accuracy in predicting concrete strength. The confusion matrix reveals 91% accuracy and 99% precision. Evaluation using MSE, MAE, and RMSE metrics confirms that the MLP model has lower error rates than other algorithms, demonstrating its effectiveness in predicting uniaxial compressive strength.

Artificial intelligence, Uniaxial compressive strength, Concrete prediction, MLP, Neural networks