Environmental Engineering Reference
In-Depth Information
or non-random generation. The screened networks will finally have the set of properties that
are assigned in random or controlled way (fixed values, or calculated from coordinates, for
pipe lengths, or in the process of GA-optimisation, for pipe diameters).
The tool has been tested on three sets of junctions (20, 77, and 200) and the conclusions
about its functionality can be summarised in the following bullets:
1.
NGT has proven to produce network layouts that fairly resemble those seen in water
distribution practice.
2.
The complexity of network layouts can be controlled (better in the non-random mode), as
well as the running time to generate several dozens of networks up to a few hundred links
is fast enough (between a few seconds and a few minutes). The random generation is
showing faster running times for larger networks but with more unpredictable layouts,
both in terms of complexity and number of networks. Various calculations will show
different running times, which is highly sensitive to the selection of the number of links
for given number of nodes and specified maximum nodal connectivity.
3.
The library of case networks can be further diversified by the range of possibilities in
definition of network properties.
4.
During the development phase, the tool has not been tested for extreme conditions of
several hundreds of networks (of several thousands of pipes). The large samples of a few
thousand layouts have been generated in the research presented in Chapter 9, but the
testing of upper limits of the tool is actually yet to be done. That has not been seen as
immediate priority in this research, for two reasons. First, a few initial tests showed a
huge increase in running times (of several hours) without any guarantee of completing the
process; hence, the input settings are first to be refined, in fact simplified. Secondly, the
improved robustness that would enable generation of large number of big networks may
have little practical use: for design purposes very likely, and for research purposes,
possibly. Analyses of extremely large samples ask much more time, whilst the additional
benefit of the (large number of) results is yet to be proven compared to the analyses of
smaller number of alternatives.
5.
Nonetheless, it is believed that the NGT algorithm can be further improved to provide
smooth generation of larger number of bigger networks. For particular purposes, allowing
long running times of several hours or even days could be tolerated as the network
generation process will mostly be a 'one time task'.
6.
The room exists to improve the user friendliness of NGT by simplifying the programme
options e.g. to integrate three ways of column/complexity definition into one, in the non-
random generation mode.
The initial idea of NGT has been to serve the research in water distribution. With small
addition of filtering the networks on particular pipe/street routes, NGT can easily be
converted into a tool to be effectively used for design purposes, as well. This add-on has been
developed within the decision support tool discussed in Chapter 9.
Search WWH ::
Custom Search