📄 Published: Springer Nature Book Chapter
🏛️ Presented at: HYDRO 2024 International Conference, Central Water & Power Research Station, Pune (Dec 18–20, 2024)
🏫 Institution: Chaitanya Bharathi Institute of Technology, Hyderabad
This project applies two AI modeling techniques to predict scour depth around bridge piers under clear water scouring (CWS) conditions:
- ANN-PSO (Artificial Neural Network with Particle Swarm Optimization)
- SVM (Support Vector Machine)
Scour—the erosion of sediment around hydraulic structures—is a leading cause of bridge failure globally. The 2022 and 2023 collapse of the Aguwani-Sultanganj Bridge over the Ganga River highlighted the critical need for accurate scour prediction tools.
| Metric | SVM | ANN-PSO |
|---|---|---|
| MAPE | 34.76% | 20.25% |
| RMSE | 0.521 | 0.496 |
| R² | 0.5145 | 0.5654 |
| Nash-Sutcliffe Efficiency | 0.620 | 0.708 |
ANN-PSO outperforms SVM across all metrics.
- 16 literature sources compiled (Ebrahimi, Khan, Chiew, Melville, Ettema, etc.)
- Non-dimensional parameters: b/y, V/Vc, Frc, b/d₅₀, σg → ds/y
- Split: 75% training / 25% testing
- Data collection from published CWS experimental datasets
- Normalization of all input variables (0–1 range)
- ANN architecture: feed-forward backpropagation with PSO weight optimization
- SVM: high-dimensional hyperplane separation for regression
- Evaluation: MAPE, RMSE, Nash-Sutcliffe Efficiency (E)
Python MATLAB ANN PSO SVM Statistical Error Analysis
Cherishma P., Sreeya M., et al. (2024). Experimental Investigations and AI Modelling of Scouring Depths for Improved Structural Integrity. HYDRO 2024 International Conference. Springer Nature (Book Chapter).
- Grant: CBIT/PROJ-IH/I024/Civil/D002/2024 (CBIT Seed Research Grant)