TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.4, April 201
4, pp. 2652 ~ 2
6
5
7
DOI: http://dx.doi.org/10.11591/telkomni
ka.v12i4.4807
2652
Re
cei
v
ed Se
ptem
ber 8, 2013; Re
vi
sed
Octob
e
r 16, 2
013; Accepte
d
No
vem
ber
5, 2013
Optimization Design of Cantilever Beam for Cantilever
Crane Based on
Improved GA
Shufang Wu
1
*, Tiexiong Su
2
1
School of Me
chan
ical En
gi
n
eeri
ng an
d Aut
o
matio
n
, Nort
h
Universit
y
of C
h
in
a, T
a
iyua
n, 030
05
1, Chin
a
2
School of Mec
hatron
i
c Eng
i
n
eeri
ng; North U
n
iv
ersit
y
of Ch
i
na; T
a
i
y
ua
n, 0
300
51, Ch
ina
*Co
e
rrspo
ndi
n
g
author, e-ma
i
l
:
sh
u
f
a
n
g
w
u66@
1
2
6
.
co
m
A
b
st
r
a
ct
Based
on in
d
epth study of
opti
m
i
z
at
ion
d
e
s
ign
meth
ods,
accord
ing to a
naly
z
i
n
g th
e fo
rces of
cantil
ever be
a
m
for Canti
l
ev
er Cr
an
e, w
i
th the feature
that I-bea
ms u
s
ed in
cantilever beam
m
o
stly,
combi
n
in
g stru
ctural
opti
m
i
z
at
ion
tech
niq
u
e
w
i
th discr
ete
v
a
ria
b
les
w
i
th G
A
, an
opti
m
i
z
a
t
ion
des
ig
n
met
hod
of cantilever beam
for Cantil
ever Crane based on im
prov
ed GA was propos
ed for the problem
s of huge
mater
i
al
re
dun
dancy
an
d
hig
h
pr
oducti
on
c
o
st. Mathe
m
ati
c
al
mo
de
l a
n
d
fitness fu
nctio
n
of th
e struct
ur
a
l
opti
m
i
z
at
ion w
i
th discrete v
a
riabl
es w
e
re b
u
ilt, opti
m
i
z
at
i
on d
e
sig
n
of cantil
ever b
e
a
m
structure w
a
s
achi
eved, the
efficiency of thi
s
opti
m
i
z
a
t
i
on
desi
gn
me
th
od
w
a
s validate
d
and th
e cons
u
m
pti
on of stee
l
for
prod
uction w
a
s
reduce
d
. T
h
is meth
od h
ad a
certain
g
u
id
in
g signific
anc
e for engi
ne
erin
g a
pplic
atio
n.
Ke
y
w
ords
:
in
dex ter
m
s-opti
m
i
z
at
io
n desi
g
n, improve
d
GA,
discrete vari
abl
es, fitness functio
n
, cantil
e
v
e
r
crane
Co
p
y
rig
h
t
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
The g
eom
etry and l
oadi
n
g
conditio
n
s of C
antilev
e
r
Crane
are quite
com
p
lex, so
stru
ctural parameter
s are prima
r
ily
dete
r
mine
d
by
ex
perie
nce,
no
pre
c
ise cal
c
ul
ation of stren
g
th
and
safety factor is
often a la
rge
r
value in
ord
e
r
to meet th
e relia
bility requireme
nts
in
traditional d
e
s
ign meth
od
s. Therefo
r
e, materi
al
wa
ste and cost im
proving a
r
e p
r
odu
ce
d.
At present, th
ere
are
many
algo
rithms a
nd t
heo
rie
s
a
bout optim
um
stru
ctu
r
al d
e
s
ign. A
s
a
global opti
m
ization se
a
r
ch algo
rithm
,
Genetic
Al
gorithm
(GA) is a ge
nera
l
frame
w
ork
for
solving
comp
lex optimizati
on problem
s
whi
c
h ha
s fe
w limitations
on se
archin
g
spa
c
e, no n
eed
of continuity of solution
s a
nd stro
ng rob
u
stne
ss.
Existing resea
r
ch on gen
etic al
gorithm i
s
ma
inly
about optimi
z
ation method
for continu
o
u
s varia
b
le
s. Due to cantil
ever bea
m with I-beam
s a
n
d
discrete
d
e
si
gn va
riable
s
,
discrete
resul
t
s which
obtai
ned
by adj
ust
i
ng the
optimi
z
ation
meth
o
d
s
of continuo
u
s
variabl
es a
nd need to test the
feasi
b
ility and reliability are usually infeasib
le
solutio
n
s. T
h
erefo
r
e, rese
arch o
n
cal
c
u
l
ation me
th
od
for o
p
timum
stru
ctural d
e
sign with
di
screte
variable
s
ha
s a certain p
r
a
c
tical valu
e.
By study on optimization te
c
hni
que of di
screte vari
abl
es an
d impro
v
ing stand
ard
genetic
algorith
m
in
this arti
cle,
mathemati
c
al
model
a
n
d
fitness fun
c
tion of the optimizatio
n
with
discrete va
ria
b
les
have buil
t
and ca
ntilever be
am st
ru
cture of Cantil
ever Cra
ne h
a
s o
p
timize
d
by
sele
ction,
cro
s
sover an
d
mutation. Finally, st
ru
ctural optimi
z
ation pro
b
le
m with discrete
variable
s
ha
s solved.
2. Simple G
A
GA is
a
kind
of glob
al o
p
t
imization
se
arch al
gorith
m
whi
c
h
sim
u
lates the p
r
oce
s
s of
geneti
c
and e
v
olution in na
ture an
d prop
ose
d
by
Holl
and at University of Michig
an in hi
s pap
er
for the
first i
n
19
75. Th
e
algorith
m
, firstly, enc
ode
s acco
rdin
g to
the
solutio
n
of the p
r
obl
e
m
,
transl
a
tes
so
lution space
into GA se
arching
sp
ace and p
r
o
d
u
c
e
s
an initia
l grou
p, the
n
,
simulate
s th
e phen
omen
on of bree
di
ng, cro
s
sove
r and ge
neti
c
mutation i
n
the pro
c
e
s
s of
natural
sele
ction an
d h
e
re
dity, improve
s
the
ada
pta
b
ility of individual in
the g
r
oup
thro
ugh
a
variety of genetic ma
nip
u
lation ope
rators, the
r
eb
y obtains be
tter individua
ls, use ge
ne
tic
operators (se
l
ection, cro
ssover and m
u
tation) to
co
mbine the
s
e
individual
s, prod
uce a n
e
w
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Optim
i
zation De
sign of Ca
ntileve
r Beam
fo
r Cantile
ve
r Crane Ba
se
d on… (Sh
u
fang Wu)
2653
can
d
idate g
r
oup, having
several gen
e
r
ation of gen
etic evolution
until after some conve
r
ge
nce
index is sati
sfied, that means get the op
timal
solution
of the proble
m
[1-3]. The pro
c
e
ss
sho
w
s
in Figure
1.
Simple gen
etic algo
rithm
s
can b
e
define
d
as a 7 - ele
m
ent array:
(,
,
,
,
,
,
)
cm
GA
M
F
s
c
m
p
p
(
1)
In this
type,
M
is pop
ulati
on si
ze,
F
is individual fitness evalu
a
tion functio
n
,
S
is selection
operator,
C
i
s
cro
s
sover o
perato
r
,
m
is
mutation o
p
e
r
ator,
P
C
is
cross
o
ver r
a
te,
P
m
is
mutation
rate.
Figure 1. Simple GA Flowchart
The key p
r
obl
ems in the p
r
oce
s
s of implementation a
s
follows:
(a) Codi
ng
Codi
ng i
s
th
e
process th
at re
pre
s
e
n
ting
the first g
e
n
e
ration
of i
n
d
i
viduals in th
e g
r
oup
with a
fixed
-
length
bin
a
ry encodin
g
string.
Usu
a
l
l
y codi
ng l
e
ngth i
s
d
e
te
rmine
d
by t
h
e
requi
rem
ent
of actual p
r
e
c
isi
on an
d th
e si
ze of the
grou
p is
gen
erated
by ran
dom meth
od.
In
gene
ral, the
large
r
the
popul
ation
size i
s
, the
b
e
tter it is. But oversi
ze
scale
will re
du
ce
computational efficiency
and small
size will
cause algorit
hm convergence i
n
advance, t
hen
optimal sol
u
tion co
uld not
be obtain
ed.
(b) Fitness
func
tion
Fitness fu
ncti
on i
s
a
corre
s
po
ndin
g
rela
tionshi
p b
e
tween i
ndividua
ls a
nd th
eir fitness i
n
a grou
p. In geneti
c
algo
ri
thm, the pro
bability of
individual he
re
dity to the n
e
xt generatio
n o
f
grou
p is d
e
ci
ded by the size of the in
dividual fi
tne
ss. Th
e grea
ter the indivi
dual fitness,
the
C
B
A
x
Q+G
1
y
P
1
P
2
N
1
N
2
L
L
1
q
Figure 2. Sch
e
matic Di
ag
ram of
Cantile
ver Beam Structure
and
L
oa
d
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 4, April 2014: 2652 – 2
657
2654
bigge
r the p
r
obability that inherit to the
next gene
ra
ti
on is. Oth
e
rwise, the le
ss i
t
is. In gene
ral,
take obj
ective
function a
s
the fitness evaluation fun
c
t
i
on in the ma
ximum probl
e
m
.
(c) Gen
e
tic o
perato
r
The sim
p
le g
enetic al
gorit
hm usu
a
lly includ
es three
geneti
c
ope
ra
tors:
1. Select o
p
e
r
ation: individ
uals
with hi
g
her
fitne
s
s se
lected f
r
om t
he current p
o
pulation,
put into the matchin
g
set
(buffer), p
r
e
pare
d
for late
r ch
romo
so
m
e
cro
s
sing
-ov
e
r, mutation
and
prod
uci
ng n
e
w
individ
ual.
Wheth
e
r
ea
ch in
dividual
will be
sel
e
cted a
nd
co
pied to the
next
gene
ration b
a
s
ed o
n
cum
u
l
a
tive proba
bil
i
ty mainly.
Ba
sic o
perator i
s
the ch
oo
sin
g
prob
ability:
)
(
)
(
i
i
x
F
x
F
s
(
2)
In this
formula,
i
x
is the i
-
th
chromo
som
e
in the po
pul
ation,
)
(
i
x
F
is
fitnes
s
func
tion,
)
(
i
x
F
is
the sum of all
individual fitness in the po
pulation.
2. Cro
s
sove
r operation: T
w
o in
dividual
s bin
a
ry cod
e
we
re
cho
s
en from th
e
curre
n
t
popul
ation a
nd one or m
o
re bits
were
exchang
e
from each oth
e
r, then ne
w individuals
were
prod
uced. T
a
ke two indi
viduals,
A
and
B
, as example in the
current po
p
u
lation, two
new
individual
s
A'
and
B'
we
re g
enerated thro
ugh si
ngle
-
po
int cro
s
sover.
01
00010100
11
10110111
11
00010100
01
10110111
’
’
B
A
crossover
B
A
(
3
)
3. Mutation operatio
n: take one bit of the cu
rrent bi
nary co
de in
dividual a
s
a variation
point, invert gene valu
es
of it. For example, take th
e forth bit of individual
C
as
mutation point,
new in
dividua
l
C '
wa
s gen
erated afte
r mutation.
1000010011
1001010011
’
C
mutation
C
(
4
)
(d) O
p
e
r
ation
param
eters
M
: pop
ulation
si
ze,
whi
c
h
i
s
the
nu
mbe
r
of individ
ual
s in the
g
r
ou
p, is 20
-10
0
u
s
ually;
T
:
terminate ev
olutional ge
n
e
ration of g
enetic
alg
o
rit
h
ms, usually
the value is 100
-50
0
;
Pc
:
crossover probability, usual
ly
the value is 0.4 -0.99;
Pm
: mutation probability, usually the val
ue
is 0.000
1-0.1.
3. Impro
v
ed
Gene
tic Alg
o
rithm of
Cantilev
e
r Beam for Cantilev
e
r
Crane
3.1. Cantilev
e
r Beam Str
u
ctur
e and S
t
res
s
Analy
s
is
Take
s the
col
u
mn ca
ntilever cran
e as i
n
stan
ce
with Lifting weig
ht
Q
=
3
t, Weight rotating
radiu
s
R
=400
0mm an
d Lifting heig
h
t
H
=3000m
m. At pre
s
ent, I iro
n
with the p
a
rticular type
3
6
c,
height
h
= 3
6
0
mm, width
b
= 14
0mm a
n
d
wei
ght is
a
bout 29
7kg served a
s
ca
n
t
ilever bea
m in
finalize
d
pro
d
u
ct [4]. Cantilever bea
m structure and lo
ad we
re sho
w
n in the follo
wing figu
re.
In Figu
re
2,
Q
a
s
m
a
ximu
m lifting lo
ad,
G1
a
s
trolley
dea
dweight,
q
a
s
ca
ntilever beam
dead
weig
ht load set,
L
as cantilever le
ngth,
L1
as
cantilever len
g
t
h of cantilever bea
m,
N1
as
the level
rea
c
tion of p
o
int
C,
N2
a
s
th
e
level re
actio
n
of poi
nt
A
,
P2
a
s
ve
rtical
reactio
n
of
poi
nt
A,
P1
as vertical re
actio
n
o
f
point
C
.
3.2. Cantilev
e
r Beam Str
u
ctur
e Opti
mized Model
Due to a
dopti
on of I steel i
n
cantileve
r b
eam, optimized po
ssi
ble value
s
we
re di
screte
variable
s
. Th
e ch
ara
c
te
rist
ics
of discrete variabl
e opt
imization
pro
b
lems
we
re t
hat the varia
b
le
values we
re discrete
ne
ss,
feasible sol
u
tion
set
was distrib
u
ted
m
a
inly
as scattering sp
ots
a
nd
obje
c
t fun
c
tio
n
s
and
con
s
t
r
aint fu
nctio
n
s
in
the
mat
hematical m
o
del
we
re
no
longe
r
had
the
contin
uity and differentiab
ility, so man
y
effectiv
e p
a
rsi
ng algo
rit
h
m in the original co
ntinu
ous
variable o
p
timization
coul
d not be appli
ed [5-7].
Structu
r
e o
p
timization
mat
hematical mo
del of
di
scret
e
varia
b
le
co
uld be
de
scri
bed a
s
follows
:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Optim
i
zation De
sign of Ca
ntileve
r Beam
fo
r Cantile
ve
r Crane Ba
se
d on… (Sh
u
fang Wu)
2655
12
mi
n
(
)
(
,
,
,
)
,
..
(
)
0
1
,
2
,
,
()
0
1
,
2
,
,
ni
i
i
j
k
FX
F
x
x
x
x
S
S
S
st
g
X
j
m
hX
k
p
(
5
)
In this
type,
()
F
X
is the obj
ecti
ve function,
X
is the de
sig
n
variable
s
,
i
S
is di
sc
ret
e
variable
s
val
ues set of
i
x
,
S
is s
e
t for
(1
,
2
,
,
)
i
Si
n
,
()
j
g
X
is the ine
quality
con
s
trai
nt fun
c
tion,
()
k
hX
is the equalit
y const
r
aint functio
n
.
(1) Obje
ctive
Function
The cantileve
r beam
of ca
ntilever cran
e
is a
wel
d
me
nt which was
con
s
i
s
ts
of I steel and
bra
c
ket. I ste
e
l was sta
n
d
a
rdi
z
ed
an
d
bra
c
ket
wa
s
weld
ed
with
steel plate
afte
r
cutting. Val
ues
of sectio
n pa
rameters an
d
sup
port pl
ate thickne
ss
p
a
rameters were
the set of discrete va
riabl
e
s
and o
p
timiza
tion de
sign
must corre
s
p
ond
with the
spe
c
ification
s
an
d sta
n
d
a
rd
s. The
r
ef
ore,
obje
c
tive function coul
d be
rewritten a
s
:
11
2
2
min
(
)
,
ii
i
F
XL
x
A
x
x
S
S
S
(
6)
In this
type, objec
t
ive func
tion
()
FX
is qu
ality of cantileve
r
beam,
1
is de
n
s
ity of I steel,
L is le
ngth o
f
I steel,
1
x
is p
l
ate are
a
of I steel,
2
is de
nsity of bra
cket plate,
A
is
ar
e
a
of
steel plate,
2
x
is thickne
ss of steel plate.
(2) Con
s
trai
nts
Strength con
s
traint
s:
11
;
22
(
7)
Stiffness con
s
traint:
11
f
f
;
22
f
f
(
8)
Among them:
1
is maximum stre
ss of the cantileve
r beam,
2
is max
i
mum stress of
cantileve
r bracket,
1
f
is m
a
ximum di
spl
a
cem
ent of I steel,
2
f
is maximum displacement of
brac
ket.
3.3. Impro
v
e
d
Gene
tic Al
gorithm of Discrete Varia
b
le
(1) Initial con
d
itions a
nd constraints
Geneti
c
algo
rithm took de
sign vari
able
s
co
ding a
s
operation obj
ects a
nd four design
variable
s
L
、
1
x
、
2
x
、
A
were
cont
ained
in th
e
obje
c
tive fun
c
tion
acco
rdi
ng to type
(6
). The
cantileve
r cra
ne had a fixe
d rotating rad
i
us in the current re
sea
r
ch (i.e.,
L
is a fixed value) a
nd
A
is
cro
s
s-se
cti
onal
are
a
of t
he b
r
a
c
ket .Compa
red
to
cantilever
bea
m and
up
righ
ts, length
an
d
width
of ca
ntilever
b
r
a
c
ket were small, so
they
were
can be
seen
a
s
con
s
tant va
lues. Be
ca
use
of little effect
on the
overall
quality,
2
x
was not
optimi
z
ed
.
1
x
wa
s
cross-section
a
l a
r
ea
of I-bea
m
whi
c
h
can b
e
se
arche
d
i
n
the man
u
a
l
x1={
2
1
.516,
26.131, 3
0
.765,
35.5
78,
39.578, 4
2
.128,
46.528, 4
8
.54
1
, 53.541, 5
5
.404, 61.
0
04,
67.156, 7
3
.55
6
, 79.956, 7
6
.480, 83.6
80}.
It can be
see
n
that
1
x
is
disc
ret
e
variables
optimiz
a
tion des
i
gn.
Available by the stiffness a
nd stre
ngth of
theoretical know
l
edge,
sel
e
ction of I ste
e
l
sho
u
ld satisfi
ed the followi
ng co
nstraint
s.
0
]
[
-
2
/
)
2
/
(
*
)
2
/
(
*
)
1
(
W
E
L
q
E
L
Q
G
(
9)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 4, April 2014: 2652 – 2
657
2656
0
]
[
-
8
/
3
)
2
/
(
*
3
/
3
)
2
/
(
*
)
1
(
f
EI
E
L
q
EI
E
L
Q
G
(
10
)
Available
by cantileve
r cr
a
ne d
e
si
gn
sp
ecification
s
,
mm
f
MPa
16
]
[
,
156
]
[
;
The valu
es of
W, I
were discrete numeri
c
al
and dep
end
on iteration
values of cross-se
ct
ional
area of I-be
am.
Becau
s
e
the
sup
port
plate
locate
d in
co
nstrai
nts sid
e
,
its st
re
ss an
d defo
r
matio
n
were
small,
so
con
s
trai
nt ca
n be omitted.
(2) En
cod
e
Whe
n
de
sign
variable
s
a
r
e
discrete n
u
m
e
rical
In the discret
e
variable
structural optimi
z
ati
on of ge
n
e
tic algo
rithm
,
the length of binary
cod
ed de
pen
ded on th
e numbe
r of di
screte vari
ab
le numb
e
r.
Each in
dividual ado
pt hybrid
cod
ed metho
d
whi
c
h incl
u
ded bina
ry coded, de
ci
ma
l integer cod
ed (serial n
u
m
ber of discrete
variable
valu
es) an
d real-cod
ed
(di
s
cre
t
e variabl
e va
lues). Bina
ry cod
ed
wa
s u
s
ed i
n
cro
s
so
ver
and mutation
operation
s
. Coded relation
ship
wa
s sh
o
w
n in Tabl
e 1
.
Table 1. Hyb
r
id Cod
ed Co
mpari
s
o
n
Tab
l
e
binar
y
coded
decimal in
teger coded
real-coded
0000
0
21.516
0001
1
26.131
0010
2
30.765
……
……
……
1111
15
83.680
(3) Fitness
func
tion
Simple gen
e
t
ic algo
rithm
usually u
s
e
d
for solvin
g the pro
b
le
m of unco
n
strain
ed
maximum pri
n
cipl
e, gene
rally took obj
e
c
tive func
tion
as fitness fu
nction a
nd fitness value
was
requi
re
d to b
e
greater tha
n
ze
ro. But
cantilever
bea
m optimize in
the arti
cle b
e
longe
d to sol
v
ing
the proble
m
of minimum
p
r
inci
ple
with
constrai
ns
. So s
o
me methods
were
adopted to trans
f
orm
gene
ral fitne
s
s fun
c
tion to
anothe
r ki
nd
of fitne
ss fu
n
c
tion
whi
c
h i
n
clu
ded
co
nstrained.
Refu
sal
strategi
es
we
re used he
re.
ma
x
1
1
2
2
1
1
2
2
1
1
2
2
11
2
2
1
1
2
2
'(
)
(
)
(
)
'(
)
0
(
)
F
X
C
L
x
A
x
i
f
and
and
f
f
and
f
f
F
X
i
f
or
or
f
f
or
f
f
In this
formula,
)
(
X
F
is
fitnes
s
func
tion,
max
C
is a large
r
given co
nstant.
(
4) G
eneti
c
o
perato
r
s
Selection
op
erato
r
:
Fitne
s
s valu
e p
r
o
portion
sele
ct meth
od
wa
s a
dopte
d
an
d
accompli
sh
ed
by roulette wheel metho
d
in sele
ction.
In this way, firstly, individu
al fitness was cal
c
ulate
d
, a
nd then th
e p
r
opo
rtion
of in
dividual
fitness in the
total popul
ation's fitness
was
cou
n
ted which re
pre
s
e
n
t
ed
the cho
s
e
n
proba
bility
in
the pro
c
e
s
s
of sele
ction.
Wheth
e
r the
indivi
dual was sele
cted or
not wa
s
d
e
termin
ed
by
a
rand
om num
ber, so as t
o
ensure the good g
e
n
e
s tra
n
smitte
d to the ne
xt generatio
n
of
individual
s.
Cro
s
sove
r op
erato
r
:
Due t
o
the fewer o
p
timal de
sign
variable
s
a
n
d the sm
alle
r desi
gn
variable
s
’
stri
ng le
ngth, the
crossove
r
probability
was
100 %
(
It me
ans that all
th
e chro
mo
som
e
s
were involved in the cr
o
ssover o
peratio
n) in
this opti
m
al de
sig
n
, the inte
rsectio
n
was
gen
era
t
ed
rand
omly.
Mutation op
e
r
ator
:
Vari
ation wa
s an i
m
porta
nt me
thod to prev
ent pre
c
o
c
io
us, the
numbe
r of
va
riation digits were
d
e
termi
ned by
t
he m
u
tation p
r
ob
a
b
ility, the len
g
th of e
n
codi
ng
string a
nd the
populatio
n si
ze, and the
n
do mutation o
peratio
n.
(5) O
p
e
r
ation
param
eters
The p
opul
atio
n si
ze
M
=50,
the ge
netic a
l
gebra i
s
5
0
, the
cro
s
sove
r
rate
Pc
=100
%, the
mutation rate
Pm
= 0.001,
Cm
ax
= 10
0.
The
kno
w
n
p
a
ram
e
ters of
cantileve
r :th
e lifting capa
city
Q
= 3t,
R
= 4000m
m,
H
= 3
000m
m, the type of I
-
beam
s i
s
3
6
c
,
h
=
3
60m
m,
b
=
140
m
m
,
x1
=
9
0
.
88
c
m
2,
G1
is the trolley’s self-weight,the ca
nt
ilever’s
self-weight load co
nce
n
tration
q
=
31.069
kg/m,
the cantileve
r’s total length
L
= 43
00mm,
the cantileve
r length
L1
= 3000mm , th
e
scaffold mate
rial is Q
235,t
he cro
s
s-sect
ional a
r
ea
A
=14
62.5
c
m2
、
the thickne
ss
of steel pl
ate
x2
= 26mm,
Mpa
156
]
[
2
,
mm
f
1
]
[
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Optim
i
zation De
sign of Ca
ntileve
r Beam
fo
r Cantile
ve
r Crane Ba
se
d on… (Sh
u
fang Wu)
2657
3.4. The Opti
mization De
cisions
This p
ape
r h
a
s too
k
the
cantilever’
s weight as
obje
c
tive function
with co
nst
r
ai
nts that
the maximal
displ
a
cement
wa
s
not mo
re that all
o
wa
ble valu
es, th
e maximum
stre
ss
wa
s le
ss
than 1
56Mp
a
and
the mi
nimum
safet
y
factor was less th
an
1
.
5. Then, th
e cantilever
had
optimize
d
. T
he converge
nce
re
sult was
sho
w
n in
Figure 3. T
he obje
c
tive
function b
e
g
an to
conve
r
ge to the optimal so
lution after 61
th
generation;
the target value wa
s 23
6.9
K
g.
The stress an
d displ
a
ceme
nt before an
d
after optimization we
re sh
own in Ta
ble
2.
Table 2. The
Values
Com
p
arison befo
r
e
and after Size Optimizatio
n
design ariables
w
e
ight
max
i
mum displacement
max
i
mum stress
Minimum
safet
y
factor
1
x
2
x
before
90.880
26
287.3 Kg
13.8 mm
92.3 MPa
2.01
after
73.556
24
236.9Kg
15.3mm
106.5MPa
1.98
4. Conclusio
n
Thro
ugh the
global optimi
z
ation a
b
ility
of genetic
al
gorithm, the tradition
al alg
o
rithm in
the di
screte
space o
p
timization p
r
obl
e
m
s
wa
s
so
lv
ed. Ta
king
cantilever
cran
e cantilever b
eam
stru
cture a
s
an exa
m
ple,
the
cr
oss-section
a
l a
r
ea
wa
s taken
as th
e
de
sig
n
vari
able
s
,
the
minimum wei
ght wa
s take
n as the obj
ective f
unctio
n
, and the di
screte vari
ab
le optimizatio
n
desi
gn ba
sed
on geneti
c
algorithm
wa
s carrie
d out. Accordi
ng to the data whi
c
h
were sh
own in
table 2, the o
p
timized
qual
ity is about 2
0
% lighter
th
an befo
r
e, th
e utilizatio
n rate of the ma
teria
l
has imp
r
ove
d
,
the size of the stru
ctu
r
e p
a
rt
s ha
s rea
s
onably match
ed and the re
sea
r
ch provid
e
d
a theoreti
c
al
basi
s
for the l
i
ghtwei
ght op
timization de
sign of the can
t
ilever crane.
Referen
ces
[1]
T
A
O Yuan-fan
g
, SHI Xia
o
-fe
i
. Optimization
desig
n for crane g
i
rders b
a
sed o
n
impr
oved p
a
rticl
e
s
w
arm algorithm.
Chines
e Jo
urna
l of Constr
uction Mac
h
in
e
r
y
. 2012; (1): 50-53.
[2]
Z
H
ANG Sicai,
Z
H
ANG F
ang
-xi
ao. Ap
pl
icati
on of
A
l
gorithms in Structur
al
Optimization Design
w
i
t
h
Discrete Var
i
ables.
Jour
nal
of Southw
est Jiaotong U
n
verstit
y
. 2003; (4):1-5
.
[3]
Z
H
ANG Yan-
ni
an, LIU
Bi
n, GUO Pen
g
- fe
i.
H
y
brid
Genetic Algo
rithm for
Optimum
D
e
si
gn of
Bu
ild
in
g
Structure.
Jour
nal of North
eas
tern Univ
ersity
. 2003; (10): 9
9
0
-99
3
.
[4]
XU Ge-n
in
g, LU F
eng-
yi, Z
H
ANG Lian
g–
yo
u. Res
earc
h
o
n
Comp
leted O
p
timize
d Desi
g
n
of Structure
for Sle
w
i
ng Ca
ntilev
e
r Cran
e.
Journ
a
l of Ton
g
ji Un
iversity
. 2
001; (12): 1
476
-148
0.
[5]
ZHAO Yan-min, HUO Da.
Opti
mal D
e
sig
n
of Stee
l T
r
uss Structure B
a
sed
on G
ene
tic Simul
a
t
e
d
Anne
ali
ng Al
go
rithm.
Journa
l of Z
heng
z
h
o
u
Univers
i
ty.
201
1; (6): 54-57.
[6]
GUO Peng-fei,
HAN Ying-sh
i. An Imitative Full-St
ress Gen
e
tic Algor
ithm for
Structueal Optimizatio
n
w
i
t
h
Discret
e Variables.
En
gin
eeri
ng Mech
an
ics.
2003; (4): 95-9
8
.
[7]
Z
hu Cha
o
y
a
n
l,
Z
hang Yan
n
i
a
n, Guo Pengfe
i
l, W
ang
Xuezh
i
. An H
y
br
id Genetic Al
gorith
m
for Shap
e
Optimization of
Structure
w
i
t
h
Discrete Variables.
Jo
urn
a
l
o
f
W
uhan
Univ
e
r
sity of T
e
chn
o
l
ogy.
20
11
;
(2): 156-1
59.
Figure 3. The
Optimization
Gene
ration a
nd
Obje
ctive Function
Evaluation Warning : The document was created with Spire.PDF for Python.