Predictive Analytics for Sewer Corrosion

Sulphide induced corrosion of concrete sewers is a serious problem in wastewater systems worldwide, particularly in countries with a warm climate like Australia. This project proposes a machine learning approach to predictive analysis for sewer corrosion based on Bayesian nonparametric methods. The proposed approach provides more modelling capacities and flexibilities for incorporating various factors (e.g., H2S, temperature, and humidity), geological constraints, and temporal dynamics. The resulting predictions will lead to more effective chemical dosing, sewer pipe rehabilitation, and sensor deployment based on need rather than time.