PhD thesis - The contribution of artificial intelligence to improve asset management and the performance of drinking water supply networks
- Strasbourg, Bas-Rhin
- CDD
- Temps-plein
How can artificial intelligence be used to exploit existing data collections or information systems within water services in order to improve asset management and the performance of drinking water supply networks? Over and above the scientific aspect, this project will strengthen the research capabilities of Engees and Icube in the field of drinking water and offers operational advantages that could lead to negotiation and arbitration tools for water utility managers to support public decision-making objectively. For years, water utilities have been encouraged to instrument their networks, but they now find themselves with vast quantities of monitoring data that are not used effectively to support decision-making. AI can therefore be a tool for exploring and exploiting this data in order to objectify and support public decision-making. Several operational and scientific challenges will need to be addressed. The first concerns the contribution of artificial intelligence to the analysis of monitoring data and their relevance as explanatory variables through the identification of performance indicators that can be considered as variables to be explained. It is not enough to have data, beyond its completeness, it is necessary to establish causal links with the performance indicators to be predicted. A business dimension needs to be integrated into the understanding and analysis of the available data. The second scientific challenge concerns the establishment of links between monitoring data from sensors scattered across drinking water networks and asset management actions that may involve maintenance or investment. In other words, how can we establish the link between consumption, flow or pressure measurement data and water losses? How can causality be established between observed data and asset management actions such as pipe repairs? How can the effectiveness of actions on the network be predicted ex ante? The third issue concerns the choice of artificial intelligence approaches to be used and their level of complexity. This raises questions about the learning methods to be used and the data processing or prediction algorithms to be employed. One of the challenges is to develop approaches that are not too complex and precise, so as to encourage the involvement and participation of service managers. Experience shows that if the models developed are too complex, their appropriation and use will be limited within the services. The fourth issue concerns the identification of potential links between monitoring data and asset data. This involves studying possible causalities between data. In the case of our research, we want to cross-reference these data in order to build a decision-support model capable of exploiting the heritage data and the monitoring data. The aim is to be able to cross-reference operational data and asset data.RequirementsResearch Field Engineering » Computer engineering Education Level Master Degree or equivalentSkills/Qualifications
- Data analysis
- Data science : capacity to understand and program methods by python
- Understand the operation of a drinking water service
- Experience in mathemaical modelling
- Skills in asset management of infrastructures
- Basic backgroung in economics and management
- Good knowledge of machine learning approches & methods
- Background in the assessment of key performance indicators ( KPIs) of water services
- Capacity of adaptation
EURAXESS