Decision Support Tool for Design and Reliability Assessment of Water Distribution Networks - Pattern Recognition for Reliability Assessment of Water Distribution Networks - page 255

Figure 9.7c Sample of generated networks NfullC4(U): layout nr.1 - left, layout nr.1000 - right
Figure 9.7d Sample of generated networks N55-75C2(4): layout nr.1 - left, layout nr.1000 - right
Figure 9.7e Sample of generated networks N55-75C4(U): layout nr.1 - left, layout nr.1000 - right
It shows that an adaptation of one or two connections can significantly increase the number
of matching layouts. Furthermore, while the random number generator uses the same pattern
regardless the selected number of networks, the total number of possibilities in non-random
generation is huge. Consequently, the stricter the template is (i.e. with the lesser number of
connections) the larger the sample of generated networks should be.
9.4.3 Network Initialisation and Optimisation
To provide a variety of layouts, the optimisation on the least-cost diameter has been
conducted for all five categories of the filtered networks shown in Table 9.9. The default GA-
settings shown in Tables 9.2 and 9.3 have been used with the following diameter range (total
15, in mm): 80, 100, 125, 150, 200, 250, 300, 350, 400, 450, 500, 600, 700, 800, and 1000.
Total 1817 filtered networks ranging between 55 and 79 pipes have been optimised in
approximately 400 minutes i.e. in less than seven hours, as shown in Table 9.10. The
hydraulic simulations have been conducted with the nodal elevations and demands shown in