DiAMANT-Water: Data of the Future

The integrity of the water infrastructure is hard to determine using a directly measurable parameter, such as the condition of the underground mains network for instance. In the event of a change or deterioration in the condition of a water main, or a sudden variation in water quality, it is often difficult to identify the (possible) causes. The use of big data techniques uncovers some relevant correlations.


The idea of smartly combining data from various sources using big data techniques has already proved successful in other sectors. With the knowledge produced by the analyses, the water company can further research the failure sensitivity of mains and the occurrence of customer notifications, as well as more optimally design their management and investment plans.


The realisation of a methodology to unlock data contributes to more effective operational management because knowledge is available. With statistical, knowledge-discovery and machine-learning techniques, connections are established in a few use cases related to the integrity of the mains network.


Data inventory and the information management of components of the water cycle should be given more priority in order to provide support to operational management. Brabant Water therefore committed to streamlining its data management system. KWR worked on techniques to smartly combine data and analyse results, Nelen & Schuurmans built the platform for data visualisation. Witteveen + Bos had an advisory role.


These are the two key results of the research:

  • Production data versus failure data (USTORE) case: various datasets were aggregated into 55 supply areas in Brabant, where a specific water pressure is maintained by the pumping stations. By combining these with failure data it was discovered that failures in cement-based mains are much more closely correlated with high (maximum) water pressure than is the case of mains made from other materials. The results were visualised in a demo. This knowledge can contribute to establishing priorities in the rehabilitation of the mains network or in the control of the pumping stations.
  • Customer notifications versus temperature case: in this second case, the mining of customer notifications and environmental factors for possible correlations showed that notifications of discoloured water correlate in a significantly different manner with environmental temperature than do other types of notifications. This outcome can contribute to better capacity planning in the Customer Contact sector.
    Both cases show that domain-specific knowledge, which is frequently based on experience, is confirmed by correlations which use several datasets. The findings offer a starting point for the use of domain-specific knowledge and the further extension of the cases.