Decision Support Tool for Design and Reliability Assessment of Water Distribution Networks - Pattern Recognition for Reliability Assessment of Water Distribution Networks - page 269

design for large number of layouts that can be manually adapted in the final stage.
Additional information obtained from the corresponding HRD is further instrumental in
assessing and solving the bottlenecks After all, the finishing touch should always be done
by humans; computers do only what they are programmed to do.

- In the analysis of large number of layouts, the network buffer index, NBI , and the

corresponding HRD, have proven to be good indicators of network resilience, which can
distinguish between the layouts that may look pretty similar. The other two measures, the
pressure buffer index, PBI , and the network resilience, I n , have shown less compliance
with the loss of demand. Especially, the latter coefficient gives pretty conservative values,
significantly lower than the other indices, which does not reflect the resilience as high as
it really appears to be.

In the absence of sound theory emerging from the research on pattern recognition for
reliability assessment of water distribution networks, NEDRA is the concrete outcome that
can be effectively used in practice, to efficiently analyse numerous alternatives and scenarios.
The real development and testing of the package has just started with this research.
Throughout a huge number of simulations conducted, a couple of ideas have already been
considered how to improve it further. Most of them deal with the network generation module
that is presently the biggest bottle-neck of NEDRA. For instance:
- The list of input settings can be simplified. Some combinations of the settings yield pretty

similar configurations.

- The existing network generation algorithm can be upgraded by introducing a condition

that the node can be connected with another one only if that connection is available in the
template.

- More radically, an alternative algorithm of network generation developed specifically for

design purposes could be based on building the library of layouts applying gradual
decomposition of the template.

Furthermore, in the process of GA-optimisation some diameters could be preset e.g. the
skeleton of secondary mains, and kept fixed. This feature exists in some other GA-software
but has not been built in the EO yet.

Last but not the least, NEDRA should pass through a much more rigorous testing before
being integrated, made more user friendly, and possibly distributed for public use.