A Risk-Based Decision Support System for Failure Management in Water Distribution Networks
Thesis or dissertation
University of Exeter
The operational management of Water Distribution Systems (WDS), particularly under failure conditions when the behaviour of a WDS is not well understood, is a challenging problem. The research presented in this thesis describes the development of a methodology for risk-based diagnostics of failures in WDS and its application in a near real-time Decision Support System (DSS) for WDS’ operation. In this thesis, the use of evidential reasoning to estimate the likely location of a burst pipe within a WDS by combining outputs of several models is investigated. A novel Dempster-Shafer model is developed, which fuses evidence provided by a pipe burst prediction model, a customer contact model and a hydraulic model to increase confidence in correctly locating a burst pipe. A new impact model, based on a pressure driven hydraulic solver coupled with a Geographic Information System (GIS) to capture the adverse effects of failures from an operational perspective, is created. A set of Key Performance Indicators used to quantify impact, are aggregated according to the preferences of a Decision Maker (DM) using the Multi-Attribute Value Theory. The potential of distributed computing to deliver a near real-time performance of computationally expensive impact assessment is explored. A novel methodology to prioritise alarms (i.e., detected abnormal flow events) in a WDS is proposed. The relative significance of an alarm is expressed using a measure of an overall risk represented by a set of all potential incidents (e.g., pipe bursts), which might have caused it. The DM’s attitude towards risk is taken into account during the aggregation process. The implementation of the main constituents of the proposed risk-based pipe burst diagnostics methodology, which forms a key component of the aforementioned DSS prototype, are tested on a number of real life and semi-real case studies. The methodology has the potential to enable more informed decisions to be made in the near real-time failure management in WDS.
PhD in Engineering