Impacts of Node Connectivity on Reliability of Water Distribution Networks - Pattern Recognition for Reliability Assessment of Water Distribution Networks - page 151

be too coarse to distinguish between the networks that actually look pretty similar. The
example of good correlation is network SN73 that is visibly the network of the lowest
connectivity, which has been captured by all the measures except the NCI3 . Furthermore,
networks SN13 , SN23 , SN26 , SN31 , SN33 , SN34 , SN70 , SN72 and SN79 all qualify for the
ranking amongst the last 10, which is hinted by comparing Table 6.12 and Figure 6.10
although not unanimously. On the other hand, SN80 and SN6 appear to be ranked as well
connected networks, which can also be anticipated for SN37 , SN53 and SN59 . Finally, a
couple of networks in the medium range of connectivity also shows reasonably uniform
ranking, such as SN14 , SN40 or SN77 , but there are networks where some measures rank
them highly while the others low. The example is SN88 , where possible reason is in more
heterogeneous layout (some nodes with one connection, but also several of those with four or
five connections).

All the results have been further compared with the reliability measures, discussed in Chapter
5, namely the network buffer index ( NBI ), the network resilience of Prasad and Park ( I n ) and
the average available demand fraction ( ADF avg ). For this purpose, a series of hydraulic
simulations has been conducted in EPANET, for the head at the source of 160 msl. All nodal
elevations have been set to 50 msl, creating elevation difference with the source of 110 m.
The uniform k-value of 0.5 mm has been applied for all the pipes. Two rounds of calculations
have been executed to assess the correlation with the reliability measures: the first one for all
the configurations having uniform pipe diameter of 500 mm and variable demand multiplier
adjusted by NDT to reach the selected pressure threshold; this has resulted in the range of
demands between 1995 l/s (in network SN2 ) and 5837 l/s (SN34). The second run has been
conducted for GA optimised diameters at minimum pressure of 20 mwc, which was in both
runs also taken as the threshold used in PDD calculations (with the exponent of 0.5). The
range of diameters used for GA optimisation is between 100 and 750 mm, in increments of
50 mm; the EO-optimiser has been used for the calculations. The total demand of 2280 l/s has
been kept fixed for all the networks.

The results in Figures 6.13 to 6.16 show insufficiently strong correlation between the
connectivity and reliability measures, which is slightly better and more consistent in case of
the networks with fixed D/variable Q and NodeXL results.

1.20

1.00

0.80

NGI
NCF3/3
NCFavg/3
NCI3
NSI3
ADFavg
NBI
In

0.60

0.40

0.20

0.00

2 6 8 12 13 14 16 17 19 21 23 26 31 33 34 37 40 46 53 59 69 70 72 73 77 79 80 82 85 88

Figure 6.13 Reliability measures and connectivity indices calculated by NDT (fixed D, variable Q)