Predictive risk modelling of real-world wastewater network incidents

Abstract Due to growing pressure on wastewater network operators to deliver improved serviceability and lower costs to customers, there is a real need for greater understanding of the factors which influence incident rates, enabling effective prioritisation of proactive maintenance. This paper applies decision trees to investigate both static factors, such as sewer material and diameter, and derived factors, such as sewer velocity, for the prediction of blockages on the network of Dŵr Cymru Welsh Water. The results obtained illustrate the effectiveness of the proposed approach when identifying important explanatory factors and predicting sewers that are likely to block.