Leveraging External Data for Enhanced Understanding and Causal Attribution of Anomalies in Water Network Systems – LUCAS

This project seeks to leverage external data to represent, analyze, understand, and identify the causes of anomalies observed on water network systems. It aims to break with traditional approaches to problem analysis based on a single source, and to take advantage of independently designed databases in order to better address the problems encountered in urban networks. We will use the field of water network systems as a guideline to illustrate the generic methodologies developed in this project. Water network data are rich, including Geographic Information Systems (GIS) that provide information about network infrastructures, as well as videos from televised inspections (ITV), which are crucial for annotating various anomalies observed in the pipes. Our main objective is to go beyond water network-specific data to include contextual information on the network's vicinity, such as buildings, road traffic, population density and so on.
The aim of this project is to propose innovative solutions to the challenges posed by these massive and diverse data. It comprises the task of collecting and representing data, as well as the task of completing and selecting data relevant to the analysis of targeted anomalies. This is followed by the tasks of developing effective methods for combining these data in order to identify, understand and explain the causes of the observed anomalies, and most importantly, to predict them. Attention is given to causal ascription, which consists in determining which of a series of events spread over time is responsible for the appearance of an anomaly. The developed methods must take into account data incompleteness and uncertainty, which are intrinsic to analyzing water network data. The consortium for this project is multidisciplinary, bringing together researchers from water modeling, data science, and artificial intelligence to address the challenges identified in this project.