Environmental Engineering Reference
In-Depth Information
to generate any such layout. Table 4.16 shows that the non-random generation has been more
consistent in the segment of 20 junctions but the running times would certainly increase with
the larger number of nodes and/or links, for the same reason as in case of randomly generated
networks.
The network layouts shown in the previous figures are just a fraction of several hundreds of
networks that have been generated in the process of the development and testing of NGT.
Some of these have been used in the research presented in Chapters 6 and 7. The initial
observations from running the tool in random and non-random mode are summarised in
Table 4.17.
Table 4.17
Comparison of random and non-random network generation
Performance
parameters
Random generation
Non-random generation
Effectiveness
Mostly effective for mid-complex
networks.
Effective for various degree of network
complexity.
Network type
Specific network configuration is not
easy to control; difficult to preset
branched or highly looped networks;
more difficult to increase the
complexity gradually.
Various configurations can be obtained with
gradual increase in complexity.
Complexity
Random and difficult to control; only
by defining the range of links, and
maximum number of nodal
connections.
Either defining the range of links and/or the
complexity and/or the column range, or using
all can fix this. Some practice is however
needed to arrive at ideal combinations of input
settings.
Running time
Sufficiently fast but inconsistent; not
clearly linked to the network size.
Sensitive on the ratio of the number of
links and junctions, and the maximum
number of nodal connections.
Fast for smaller networks (a few dozen
junctions). Slower for full range of complexity
due to large number of combinations; can be
reduced by selecting specific column (range).
Can be highly influenced by the (awkward)
selection of input settings.
Seed value
Network generation is dependent on
random generator seed value; may
generate similar variety of networks
regardless the selected number of
layouts.
Not applicable.
4.8
CONCLUSIONS
The network generation tool (NGT) presented in this chapter has been developed based on
the principles of graph theory by using the EPANET toolkit functions, with the idea of
generating smaller, yet sufficiently large samples of synthetic networks with properties that
resemble real water distribution networks. NGT uses the information about the set of
junctions prepared in the EPANET input file (in INP-format) and generates the networks
using the principle of connecting up to three closest junctions, starting from the source and
avoiding pipe crossings and duplications. The pipe connections are created either by random
Search WWH ::
Custom Search