Performance assessment of pressurized fluid-distribution networks

Pressurized fluid-distribution networks are key strategic elements of infrastructure. Drinking water is a precious resource and will become more and more important with the depletion of reserves. With the growth of the human population, challenges related to drinking water are destined to increase. On average, 20% of water is lost through leaks around the world. This illustrates that there is a need for methodologies to support more efficient management of pressurized fluid-distribution networks than are currently employed. In this thesis, a methodology for performance assessment of pressurized fluid-distribution networks is proposed. The methodology based model-falsification is evaluated for leak detection and demand estimation in water distribution networks. Comparisons of network reduction strategies are carried out leading to a framework that helps finding the most appropriate strategy. An electrical analogy for low cost case studies of water distribution networks is described. In addition, a sensor placement methodology using joint entropy is proposed. Four case studies illustrate the applicability of the performance assessment methodology. Results of the thesis lead to the following major conclusions: (1) model-falsification methodology demonstrates much potential for practical use; (2) the combination of reduction strategies with model-falsification successfully reduces computation time; (3) low cost electrical-network case studies can be built and used for testing; (4) Combining joint entropy and expected identifiability leads to an effective sensor placement methodology and (5) four case studies demonstrate that error-domain model-falsification is used successfully for leak-region detection and demand estimation.

    Keywords: leak detection ; system identification ; model falsification ; modelling ; uncertainty ; sensors

    Thèse École polytechnique fédérale de Lausanne EPFL, n° 6858 (2015)
    Programme doctoral en génie civil et environnement
    Faculté de l'environnement naturel, architectural et construit
    Institut d'ingénierie civile
    Laboratoire d'informatique et de mécanique appliquées à la construction
    Jury: Prof. Michel Bierlaire (président) ; Prof. Ian Smith (directeur de thèse) ; Prof. Maher Kayal, Prof. James Brownjohn, Dr Frédéric Bourquin (rapporteurs)

    Public defense: 2015-11-27


    Record created on 2015-11-17, modified on 2016-08-09


Related material