Decision Support Tool for Design and Reliability Assessment of Water Distribution Networks - Pattern Recognition for Reliability Assessment of Water Distribution Networks - page 242

saved in the EPAinpfilter.txt file. The user can also decide to add the networks below certain
number of mismatched links to this list, qualifying them as 'passed'. The rest of the files will
be eliminated from further analyses but they are not deleted from the folder, for possible re-
use at later stage.
9.3.3 Network Initialisation Module
The major network parameters will be assigned either directly in EPANET and/or during the
process of network generation. The network initialisation module leaves a possibility to
further modify the model input by comparing the content of the files with the name on the list
of EPAinpfiles.txt file with the template used for the filtering of generated files. The template
is a network whose basic properties can be assigned to the generated networks if this was not
done during the process of generation, or their data have been modified in the course of
various analyses, for instance during the optimisation. The data conversion will also include
pumps and valves if existing in the template. This module can therefore be run before or after
the optimisation.
Next to the practical reasons, this module also serves to correct what seems to be a bug in the
EPANET toolkit function ENsaveinpfile . This function omits the nodal coordinates from the
saved INP-file, restricting visual representation in EPANET of such-saved file. The
coordinates will be reinstated in the process of data initialisation, as well as can be the case
with nodal demands, elevations and emitter coefficients. The module further restores the pipe
lengths, diameters, k-values and in addition, the status of deliberately closed pipes, if existing
in the template. Currently, the contents of INP-files will be over-written, which has to be
taken into consideration for later use (no back-up file is made).
9.3.4 Network Optimisation Module
The generated- and possibly filtered/initialised networks can be further optimised by using
the Evolving Objects (EO) optimiser, which is a single objective oriented GA-optimization
tool developed by Keijzer et al. (2002), and distributed under the GNU Lesser General Public
License by SourceForge (http://eodev.sourceforge.net).
The goal of the optimization module is to get optimal solutions for the network pipe
diameters. The main reason to use the EO has been the fact that it is an open source package
for evolutionary computations, which can be adapted for specific research purposes and used
for relatively large number of nodes and links in a network. Hence, the EO has proven to be
convenient choice for the specific purpose of this research, too. Furthermore, an investigation
on the best performing GA-optimiser to be built into NEDRA, or on the list of optimal
default GA-parameters, was not seen as an objective of this research.
The networks listed in the EPAinpfiles.txt file are optimised to keep the selected threshold
pressure at each demand node; this is done with help of information available in SNxxx files
prepared in the network generation process, and their list stored in the Inp.txt file, which can
be also adapted during the network filtration process. In case the pressure at node is below
the threshold pressure, the penalty will be added. After a number of iterations, the GA-
process will provide optimal diameters for the least cost design. The default boundary
conditions of the objective function, constraint and the penalty cost are shown in Table 9.2.