Theoretical and Conceptual Framework - Pattern Recognition for Reliability Assessment of Water Distribution Networks - page 14

Ostfeld (2001) Probability of zero annual shortage Hydraulic analysis +

stochastic simulation

The regional water
supply system of
Nazareth, Israel

Ostfeld et al.
(2002)

Probability of Fraction of
Delivered Volume (FDV), Fraction
of Delivered Demand (FDD) and
Fraction of Delivered Quality
(FDQ)

Stochastic (Monte Carlo)
simulation

EPANET Examples 1
and 3 (Rossman, 2000)

HEURISTIC APPROACHES

Awumah et al.
(1990, 1991)

Entropy measures based on flow
and consumption

Simple entropy reliability
expression calculations

Illustrative examples

Awumah and
Goulter (1992)

Idem Entropy measures as

constraints in optimal design
of WDS; use of non-linear
optimisation

Illustrative examples

Tanyimboh and
Templeman
(1993)

Idem Tailored maximum entropy

flow algorithm for single
source networks

Small illustrative
example

Tanyimboh and
Templeman
(2000)

Summary of previous work Tailored maximum

constrained approach

Small illustrative
example

2.4 MODELLING FAILURES IN WATER DISTRIBUTION SYSTEMS

Generally, two distinct types of events can induce a water distribution system to a failure
state. The first one is termed as the hydraulic performance failure , which is related to the
situations where the demand imposed on a system exceeds the capacity of the system. The
second one is related to a component failure, or the mechanical failure , which can lead (but
not necessarily) to the hydraulic performance failure. It involves actual failures of the
network reducing its conveying capacity during the failure but also after the failed component
is isolated and undergoing a repair. This situation can be prevented through a selection of
larger diameter pipes and larger capacities for other components of the network (Lansey et
al., 2002; Trifunović, 2006). Many researchers have focused on the mechanical failures,
assuming their probability as a crucial aspect of the network reliability assessment.

2.4.1 Pipe Failures

Given their huge number and variety of conditions (different material, size, age, type and
frequency of maintenance), pipes are commonly analysed on mechanical failures. The
objective of modelling pipe failure rate is to reproduce adequately the average tendency of
the annual number of pipe breaks and to predict breakage rates in the future. According to the
Watson et al. (2001), the modelling of pipe failures can be mainly grouped into three
categories: (1) survival analyses, (2) aggregated (regression) models, and (3) probabilistic
predictive models.

Survival analyses focus on the lifetime of a pipe and are primarily used for a long term
financial planning. The pipe lifetime is treated as a random variable and a standard statistical