Decision Support Tool for Design and Reliability Assessment of Water Distribution Networks - Pattern Recognition for Reliability Assessment of Water Distribution Networks - page 241

by picking a combination out of total seven available matrix columns, some of which can be
skipped in the process of generation. That eventually leads to seven different levels of
complexity (1 - lowest, 7 - highest). Furthermore, the number of connections to the nodes can
be maximised, as well as the range in number of links to be generated in the process can be
specified. The obvious correlation between these settings asks for some caution while
choosing their combination, because some of these can lead to impossible assignment: for
instance, a high complexity with low maximum number of nodal connections and/or pipes, or
the other way round.
The random network generation is more straight forward process but also with more
unpredictable result in terms of configuration i.e. the complexity of generated networks. This
is logical, because the process is less controlled although the settings to influence the
complexity are also available in this mode, through the preset maximum and minimum
number of links and the maximum number of connections to the nodes, however with the
same concern as in case of non-random network generation. More detailed information about
the settings and the pros and cons of the non-random and random network generation has
been given in Chapter 4.
The generated links further receive the properties in the following way:
1. Lengths, L , which are calculated from the available nodal coordinates, or automatically
assigned as uniform- or randomised values (in specified range).
2. Diameters, D , automatically assigned as uniform- or randomised values (in specified
range). If the layouts are going to be GA-optimised, additional files will be created to
control the GA-simulation. Hence, the GA-optimisation does not take place within the
network generation module but is to be run, if preferred, after the modules of network
filtering and initialisation have been run (or skipped).
3. Pipe roughness, k , automatically assigned as uniform- or randomised values (in specified
range).
Hence, all generated links will be initially treated as ordinary pipes, meaning that the pumps
and valves have to be added separately, which can be done during the initialisation phase.
The output of the network generation process is a number of INP-files that can be directly
loaded and further analysed in EPANET. Their names (without the extensions) will be stored
in a separate file with unique name EPAinpfiles.txt which can be further used for handling of
multiple files in other modules in one go.
9.3.2 Network Filtering Module
The network filtering module is in fact an add-on to the network generation module, which
has been specifically developed for the design purposes. It is a short programme that
compares all the networks generated in the previous module with the network template
prepared by connecting the nodes of the initial INP-file with links that could suggest the
preferred or the only possible routes i.e. the streets. The degree of similarity with the template
file will be summarised in the specified output text file. This file lists the number of links in
each of the generated networks and the number of those that match the template. The IDs of
the mismatch links will be shown in the networks that have less than ten of these. Finally, the
list of all names of the files that match the template will be printed at the end, for possible
adaptation of the contents of the file EPAinpfiles.txt . For that purpose, the generated networks
that contain all the pipes corresponding to the links in the template shall also have their name