Geology Reference
In-Depth Information
tion size, the population needs to be sorted
every generation
From Equation (6), it can be inferred that an in-
dividual with the lower rank is selected while if
two points are in the same front, the individual
with the larger local crowding distance is se-
lected.
Lack of elitism
Requirements of specifying the sharing pa-
rameter σ share
Hybrid NS2-IRR GA
With the non-dominated sorted current popula-
tion P , the non-dominated fronts will be added to
the parent population composed of non-dominated
individuals E until the size exceeds beyond the
specified population size in order to fill the popu-
lation for the next generation E . The individuals
in E are assigned via the crowding distance. To
estimate the density of individuals surrounding a
particular point in the phenotype non-dominated
Pareto front graph, the average distance of the two
points on either side of this point along with each
of the objectives is used as crowding distance. This
crowding distance is used for estimating the size
of the largest Cuboid enclosing the point i without
including any other point in the population. The
crowding distance,
The NS2-IRR GA (Cha 2008; Cha et al 2011b) is
developed through the integration of an advanced
selection method (i.e., NSGA-II) for multi-ob-
jective problems and a dynamic search encoding
policy (i.e., IRR GA) to consider high complex
control device layout optimization problem. The
proposed MOGA algorithm flowchart is shown
in Figure 4 in detail. This algorithm is composed
of mainly 4 steps:
1. Initialization of population ( P 0 ) by randomly
generated binary numbers (i.e., 0 and 1)
2. Evaluation of each binary of population ( P 0 )
3. Non-dominated sorting based on NSGA-II
algorithm using fitness values of current
population and previous non-dominated
population set ( Q t )
4. Genetic operation to generate child popula-
tion for the next generation
distance is (Deb et al. 2000)
I i
=
I i
I i
+
(
I i
+
1
m If i
1
m
)
(5)
distance
distance
For the second step, to calculate the fitness
values of the current population, the H LQG
where m is the number of objectives and I [ i ] distance
is the m th objective function value of the i th indi-
vidual in the set I . The non-dominated individual
E is also sorted according to the crowded com-
parison operator. The crowded comparison op-
erator n is (Deb et al. 2000)
2 /
control system is investigated by considering
controllability and stability of the closed loop
control system using control devices and sensors
layout information offered from each individual
binary of current population. For the third step
the non-dominated sorting is performed to calcu-
late rank ( F i ) of each non-dominated curve. If it
is the first generation, non-dominated sorted fronts
of the current population are assigned same fitness
values for each front. However from the second
generation, these non-dominated curves are filled
to P t +1 population without overflow of the pre-
defined size (N) of P t +1 population. To fill the
i
j
if
(
i
<
j
)
or
((
i
=
j
)
n
rank
rank
rank
rank
and
(
i
>
j
stance ))
distance
di
(6)
where n is the n th crowding selection; n is the
crowded comparison operator; i rank is non-domi-
nated rank; and i distance is local crowding distance.
Search WWH ::




Custom Search