Predicting Degradation Patterns in Water Distribution Networks and Their Influence on Pipe Failure

This paper presents a machine learning approach to model degradation (corrosion) of cast iron pipes in water distribution networks by utilizing a large data set of acoustic condition assessment measurements. The model provides insights into degradation patterns on distribution networks in North America and the United Kingdom (UK). A statistical analysis was performed by comparing observed degradation with the historical break rates. The results showed that the likelihood of failure does not increase significantly if wall loss remains below 15%. However, once wall loss exceeds 30%, the likelihood of failure for a 1 km (0.62 mi) pipe segment rises to over 80% within the next 5 years.