Diagnostics of Regular Performance of Water Distribution Networks and its Relation to the Network Reliability - Pattern Recognition for Reliability Assessment of Water Distribution Networks - page 174

7.1 INTRODUCTION

That the early diagnostics of diseases is essential for successful cure is all too obvious in
medicine but is generally true for treating any other problem. Water distribution networks
(WDNs) are complex systems that can be compared with live organisms and their failure to
provide expected level of service can result from combination of factors related to mistakes
made in any single step of the system development, such as improper planning, design,
construction, operation and maintenance.

Hydraulic models can point problems caused by poor planning, design and operation. In early
days of hydraulic computer modelling this would be done for limited number of scenarios or
after some damage has already been inflicted. With the development of faster algorithms and
tools, more substantial reliability analyses could be conducted for preventive purposes,
usually by assuming a number of possible irregular scenarios, and then prepare an adequate
response in advance; yet, bearing in mind that such a 'library of calamities' is never
complete. This is to large extent still the practice and there are several reasons for that. Most
importantly, every WDN is different, more or less, in the way it looks and is operated.
Consequently, the attempts to categorise WDNs according to their reliability resulting from
the way they have been designed and operated under regular conditions are rare in literature.
Next to the uniqueness of WDNs, the latest optimisation techniques are still short of
computational power to process large network models (say, over 1000 pipes) within
reasonable period of time. Furthermore and also important, the libraries of real case networks
for reliability analyses are rarely large and consistently variable, as far the properties, to lead
to credible conclusions out of hydraulic simulation results. It is mostly the synthetic networks
used for benchmarking, or isolated case studies used to illustrate particular concept or
method, which are available in the literature. The need therefore exists to develop a large
sample of networks where the impact of alterations of network simulation parameters can be
monitored continuously, which has been discussed in Chapter 4. Last but not the least, quite
some failures result from poor workmanship in the construction and maintenance phase,
which is practically impossible to build accurately into a hydraulic model input. In addition, it
is often that the failure records are maintained with little care or are insufficiently long to
draw firm conclusions about vulnerability of particular network components.

The concept of network resilience developed by Todini (2000) and improved by Prasad and
Park (2004) deals with the capacity of network to sustain certain level of calamity based on
the network configuration and provision of energy at the sources or booster stations, where
applicable. This concept, discussed in Chapter 5, is initially promising but relies too much on
the supplying heads and not enough on the network configuration i.e. the geometry. It
appears therefore to be less successful in the description of network buffer than the proposed
network buffer index (NBI). The derivation of NBI however asks for pressure-demand driven
hydraulic simulations under the failure conditions. The attempt in this chapter was to explore
other measures of network buffer, which can be derived based on statistical analyses of the
common parameters under regular operation. These measures have been mostly based on the
hydraulic properties of the network, taking also into consideration the energy balance
between the source(s) and discharge point(s).