According to the 2013 ASCE infrastructure report card, the USA potable water system needs
an investment of $384.2 billion over the next 20 years to meet current and future needs. The
American infrastructure, which has a GPA of D+ and rated as poor, needs an investment of $3.6
trillion for all infrastructure assets by 2020. One of the key proposed solutions was to prioritize
the maintenance of infrastructure considering the Level of Service (LOS). Most of research works
on water distribution network (WDN) focused on the condition or performance of water systems
creating a gap between municipal goals and public expectations. It is evident that there is a lack of
research in the area of LOS and its link with WDN condition. There is a vital need in municipalities
to link renewal plans with the Level of Service. Therefore, the main objectives of the present
research are to: 1) identify and study the factors that impact the LOS, 2) establish an assessment
model for LOS in the WDN, and 3) map the LOS to WDN condition.
Building upon recent work on the LOS of drinking water supply systems, the present research
identifies LOS factors based on the review of water supply system (WSS) performance indicators
from literature and experts in the domain. It consequently develops a framework that is dependent
on two main models: (1) Best-Worst Method (BWM) model that determines the LOS of a WSS
considering the relative weight of importance of the identified LOS factors and (2) Artificial Neural
Network (ANN) model that maps the WDN condition to LOS. Using the water network data set
of the city of Montreal, the framework is tested and the impact of pipe material and environmentalconditions on breakage rate is studied. This research proves that breakage rate varies significantly
for different pipe materials and neighborhood areas with different environmental conditions. Questionnaire
responses from the industry experts show that supply pressure and continuity, quality of
supplied water, and customer complaints are the main factors that govern the quality of service.
They also show that water quality is the most important factor to the LOS among the other significant
factors. The relationship between WDN condition and LOS is determined considering the
metrics of water quality, customer complaints, as well as pressure and continuity of water supply.
An Artificial Neural Networks (ANN) model is developed in which the above metrics are considered
the input variables and the LOS total score resulting from the developed BWM model is the
output variable. The model is cross-validated using the embedded validation in the used software
resulting in an R2 value of 0.871, which reflects a good representation of the relationship between
the inputs and the outputs. Municipal management teams will be able to connect the technical
world of condition assessment of WDN to the customer world by adopting a customer-oriented decision
making process. This enables them to understand the customer perception of the provided
service, optimize the budget allocation process and forecast the LOS based on the network condition.
It also opens perspectives to key issues for future research work to diagnose the customer
perception of municipal infrastructure performance.
iv