This project uses machine learning to predict the impact resistance of fiber-reinforced concrete containing crumb rubber particles. The predictions focus on impact energy at the first crack (FCU) and ultimate failure (URU) based on laboratory testing data.
The dataset includes:
- Experimental results from concrete mixes with and without polypropylene fibers and crumb rubber.
- Tests conducted at two curing temperatures: -10°C and 25°C.
- Each sample measured:
- FC: Blows to first crack
- UR: Blows to ultimate failure
- PINPB: Percent increase in number of post-crack blows
- FCU: Energy absorbed until first crack (kN.mm)
- URU: Energy absorbed until failure (kN.mm)
| Feature | Description |
|---|---|
| FC | Number of blows to first crack |
| UR | Number of blows to ultimate failure |
| PINPB (%) | % increase in blows after first crack |
| CuringTemp | Curing temperature in Celsius |
| FCU | Impact energy to first crack (target) |
| URU | Impact energy to failure (target) |
A MultiOutputRegressor with XGBoost was trained to predict:
- FCU (First Crack Energy)
- URU (Ultimate Failure Energy)
- R² Score
- MAE (Mean Absolute Error)
- RMSE (Root Mean Squared Error)
fibercrete-impact-prediction/
├── impact_resistance.csv # Cleaned dataset
├── modeling.ipynb # Jupyter Notebook (Colab-compatible)
├── impact_model.pkl # Trained ML model
└── README.md # Project description
- This project is based on a Master's thesis in Civil Engineering – Construction Management.
- All experimental data were obtained through real lab testing.
- The goal is to help better design rubberized concrete with improved impact performance.
This project is open for academic and research use.