With the development of inspection robots and new kinds of sensors, and thanks to a growing wealth of data, light can be shed on the underground drinking water distribution network; something that has long been impossible. This project is developing a Proof of Concept for the automatic interpretation of visual data. For water utilities this means a significant improvement in the knowledge about the pipe system and in the quality of the available data, and therefore in their operational management.
Opening the black box
The underground drinking water pipe network has long been a black box. Our picture of the condition of pipes is still limited, and is based for the most part on statistical estimates. This also applies to the precise position of pipes and, to a lesser degree, to the connectivity as well.
Over the last few decades inspection tools have therefore been developed, and have succeeded in exposing small sections of this black box. This has given drinking water utilities access to increasing amounts of information about individual pipes. In the decade ahead these positive developments will be built upon thanks to the introduction of inspection robots and new sensors. This will bring to the surface a huge amount of data about the buried infrastructure. A comprehensive and current picture, for example, the exact location, condition and morphology, of the underground pipe network will give the drinking water utilities the opportunity to boost their operational processes and, for instance, improve the efficiency of their replacement planning and the guidance they give contractors.
Proof of Concept for visual data
Since water utilities increasingly base their decisions on information rather than experience, a reliable information basis is crucial. Thus, someone basing a decision on the result of a hydraulic calculation has to be fully confident about the input data in the hydraulic model, such as the exact topology, connectivity and coordinates. That is why this project focuses first on the development of a Proof of Concept of the automatic interpretation of visual data produced by sensors, inspection tools and robots. This involves the use of algorithms that have become available over the last few years in software libraries, such as Tensorflow. Follow-up research could then extend this research to include other data.
Automatic interpretation of visual inspection data
The automatic interpretation of data can significantly improve the knowledge about the pipe system and the quality of the available data. This research will also outline how widely applicable this interpretation of inspection and sensor data is for water utilities, and how this can lead to various operational management improvements. To make it possible to realise these opportunities in practice, a roadmap will be developed covering the course from data to asset management decisions and policy.