Developing Decision Tree Models to Create a Predictive Blockage Likelihood Model for Real-World Wastewater Networks
This is an open access article under the CC-BY-NC-ND licence (https://creativecommons.org/licenses/by-nc-nd/4.0/). The final version is available from Elsevier via the DOI in this record.
To reduce the blockages occurring on wastewater networks, reducing costs, customer and environmental impact, greater levels of proactive maintenance are being conducted by water and sewerage companies. For effective prioritisation of this maintenance, an accurate model of blockage likelihood is required. This paper presents the development of a model, for provision of a blockage likelihood level and verification using unseen data, based on previous decision tree models constructed using the asset and historical incident data from the wastewater network of Dŵr Cymru Welsh Water. The model has been developed here using the geographical grouping of sewers and the application of ensemble techniques, with the results illustrating the potential benefits which can be derived from these techniques.
The work has been conducted as part of a Knowledge Transfer Partnership (KTP) with funding provided by Innovate UK and Dŵr Cymru Welsh Water (DCWW), working in collaboration with the University of Exeter’s Centre for Water Systems (CWS).
Vol. 154, pp. 1209-1216.