Why aren't surrogate reliability indices so reliable? Can they be improved?

Abstract

[EN] Water distribution networks are known to be costly infrastructures. A few decades ago, the research efforts concerning water distribution network design were focused on economic aspects and the goal was to obtain the least cost solutions. Beyond economics, these infrastructures must mostly be reliable since they provide an essential service to society. But reliability assessment is a complex task and involves various aspects: mechanical, hydraulic, water quality, water safety, among others. This paper focus is on the hydraulic reliability. As hydraulic reliability is computationally hard to measure directly, researchers came up with surrogate measures, like entropy and the resilience index. But these surrogate measures had some flaws and researchers quickly started suggesting new ones trying to avoid those known flaws, like the diameter-sensitive flow entropy or the modified resilience index. But even these new approaches are still not so reliable to be used in the design of water distribution networks. This paper presents a thorough analysis of the performance of these reliability surrogate measures, supported by illustrative examples, highlighting their pros and cons to help in deciding which one to use for design purposes. In the end, some guidelines will be presented in order to improve these indices performance, presenting some results obtained from these improvements.