Predicting concrete corrosion of sewers using artificial neural network

• Service life of corrosion impacted sewer depends on the initiation time and corrosion rate.
• Artificial neural network was trained with 4.5 years of corrosion data obtained in lab.
• Trained ANN model predicts reasonably the corrosion observed in real sewers.
• ANN performs better than the multiple regression model in predicting sewer corrosion.
• ANN model can be improved with more training data or the inclusion of more factors.