Compétition L’INTELLIGENCE ARTIFICIELLE au service du GÉNIE CIVIL
Modélisation prédictive de la résistance en compression du béton à base de cendres volantes et de laitier de haut fourneau à l'aide de techniques d'apprentissage automatique
Le sujet
This study introduces a Machine Learning-based methodology for predicting the compressive strength of concrete and optimizing its mix design. A comparative study of numerous algorithms is conducted to find the best algorithm providing the best prediction. We then propose an inverse system to predict the optimal proportions of concrete components to achieve the desired mechanical properties. A dataset consisting of 1802 data lines is used to train and test the different algorithms. The proposed approach demonstrates that the XGBoost algorithm achieves better performance with a minimum R2 of 0.98 and a RMSE of 2.28 in terms of the average deviation between predicted and actual values. As for the inverse system, experiments have shown that the KNN algorithm provided the most efficient predicting the optimal combination of components at a R2 of 0.69 and a RMSE of 0.55. Ultimately, this study contributes to advancing a construction industry that is more cost-effective, environmentally sustainable, and performance-driven.
- Équipes
- 17
- Participants
- 34
- Établissements
- 1
Le podium
Islem Baccar (GC)
Ons Frikha (TI)
Yosr Ben Saïd (GC)
Hamza Grioi (TI)
Amal Khochteli (GC)
Iskandar Sassi (TI)