Mc, np). To find the proper heuristicwhich increases the solution quality
and improves the global searching capability, the algorithm was run
several times and the parameters are randomly generated. Further,
computer runs were performed nine times for each combination according
to the methodology proposed by Payá-Zarforteza et al. [48]
based on the extreme value theory. A span length of 35 m was considered.
Figs. 5 and 6 show, respectively, the best solutions for cost and
CO2 emissions and the computing time for each combination. Table 6
gives the results of a six case-study series whose results are optimal
when both CO2 emissions and computing time are considered. S1 is
the heuristic that provides a lower minimum emission (170,002.39 kg
CO2). Besides, the average emission and computing timeare reasonable.
The difference checked between the minimum CO2 emissions obtained
with the nine SAGSO runs and the extreme value estimated using the
three-parameter Weibull distribution that fits 50 SAGSO results is less
than 0.7%. So, this set of parameters is chosen. Similarly, Table 7 gives
the results for the parameter calibration using cost as objective function.
Note that our previous test for a span length of 35 m showed that the
best results for the GSO and SA algorithms were 218,264.59 kg CO2
and 182,652.04 kg CO2, respectively.