Geology Reference
In-Depth Information
Figure 5. The top node displacements of the optimal design found by (a) CGA+HNNS and (b) MCGA+HNNS
compared with the corresponding actual ones
FUTURE RESEARCH DIRECTIONS
The optimization task is achieved by CGA
and MCGA incorporating ENDA and HNNS. The
numerical results show that all of the solutions
are feasible and the stress-strain hysteretic curves
corresponding to the elements of the optimal struc-
tures imply that the structural behavior is highly
nonlinear. The results obtained by MCGA are
better than that of the CGA in terms of structural
weight, required generations and prediction ac-
curacy. The weight of the optimal structure found
by CGA using ENDA and its required generations
are 1099.6 and 74, respectively while these of the
MCGA are 1081.4 and 66, respectively. Also the
corresponding weight and required generation
to the CGA+HNNS are 1089.1 and 78 while for
MCGA+HNNS process are 1076.2 and 63, respec-
tively. Therefore it can be observed that MCGA is
superior to CGA and the best results are provided
by the MCGA+HNNS process. According to the
results the overall time of optimization using
HNNS is significantly reduced by a factor of 0.09
compared with the optimization using ENDA.
In this chapter, a single earthquake record is applied
to the structure. For the future researches, a bank
of natural or artificial records scaled according to a
seismic design code and considering hazard levels
may be used. Also, other optimization algorithms
with higher performance may be investigated. To
improve response prediction accuracy, other neural
network models such as fuzzy network can be used.
As the computational work of the methodology
is low, it also can be used to solve the massive
seismic reliability based optimization problems.
CONCLUSION
An efficient hybrid neural network based optimi-
zation algorithm is presented to achieve design
optimization of structures subjected to earthquake
time history loading with inelastic behavior. A
modified cellular genetic algorithm denoted as
Search WWH ::




Custom Search