Data and AI · water and sewer engineering
Predicting sewer pipe deterioration with CCTV and AI
A machine-learning classifier that flags sewer pipes likely to be deficient, so inspection effort goes where it matters.
Results
- 25,674pipe records used for training
- 84%recall on test data
- 85%of deficient pipes flagged on 100 newly CCTV-inspected sections
Where it was presented
- XXXI Latin American Hydraulics Congress (IAHR, 2024): oral presentation and full paper in the proceedings, pp. 1928–1936, with M. L. Gallego Jaramillo.
- 10th Grupo EPM Suppliers and Contractors Meeting, Technical Sessions (2023): co-presenter of the model, with M. L. Gallego Jaramillo.
Tools
- Python
- pandas
- Jupyter
- Machine learning
- CCTV