Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
8
, No
.
6
,
Decem
ber
201
8,
pp. 5
292~
5302
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v8
i
6
.
pp52
92
-
53
02
5292
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Negativ
e
Total Fl
oat to Im
prove
a Multi
-
o
bje
ctive Inte
ger
Non
-
l
ine
ar Pro
grammin
g for P
ro
j
ec
t
Sch
ed
uli
ng
Compressi
on
Fachr
urraz
i
1
, Abd
ull
ah
2
, Yu
w
aldi Awa
y
3
, T
euku
Bu
di A
uli
a
4
1
,2,4
Depa
rtment
o
f
Civi
l
Eng
ineeri
ng,
S
y
ia
h
Kual
a University
,
Indo
nesia
3
Depa
rtment of
El
e
ct
ri
ca
l
Eng
in
ee
ring
,
S
y
i
ah
Ku
al
a
Univer
sit
y
,
I
ndonesia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r
9
, 2
01
8
Re
vised
Ju
l
18
,
201
8
Accepte
d
J
ul
31
, 2
01
8
Thi
s
pape
r
pre
sents
Multi
-
Objec
t
ive
Inte
g
er
Non
-
Li
nea
r
Pro
gra
m
m
ing
(MO
INLP)
invol
ving
Neg
at
iv
e
Tot
al
Floa
t
(NT
F)
for
improvin
g
the
basic
m
odel
of
Multi
-
Objec
t
ive
Program
m
ing
(MOP)
in
ca
se
the
opt
i
m
iz
at
ion
o
f
the
addi
t
iona
l
c
ost
for
Proje
ct
Schedul
ing
Com
pre
ss
io
n
(PS
C
).
Us
ing
t
h
e
basic
MO
P
to
s
olve
the
m
ore
c
om
ple
x
proble
m
s
is
a
cha
ll
engi
n
g
ta
sk.
W
e
sus
pec
t
tha
t
Ne
gat
iv
e
Tot
a
l
Flo
at
(NTF)
havi
ng
an
indi
c
at
ion
t
o
m
ake
the
basic
MO
P
to
so
lve
the
m
ore
gene
ral
c
ase
,
both
sim
ple
and
complex
of
PS
C.
The
purpose
of
t
his
rese
arc
h
is
i
dent
if
y
ing
the
c
onfli
cting
object
ive
s
in
PS
C
proble
m
using
NTF
and
im
proving
MO
IN
LP
b
y
invol
v
i
ng
the
NTF
par
amete
r
to
sol
ve
the
PS
C
pro
ble
m
.
The
Solv
er
Applic
a
ti
on,
which
is
an
add
-
in
of
MS
E
xce
l
,
is
used
to
per
form
opti
m
iz
at
ion
proc
ess
to
th
e
m
odel
deve
lop
ed.
The
result
s
show
that
NTF
has
an
i
m
porta
nt
role
to
ide
ntif
y
the
conf
licti
ng
objecti
v
es
in
PS
C.
W
e
def
ine
NTF
is
an
aut
om
ati
c
m
axi
m
um
val
ue
of
the
a
ct
ivit
y
dura
ti
on
red
uction
to
ac
hi
eve
du
e
da
te
of
PS
C.
Furthermore,
th
e
use
of
NTF
as
a
co
nstr
ai
nt
in
MO
INLP
ca
n
solve
th
e
m
ore
gene
ra
l
ca
se
fo
r
both
sim
ple
and
complex
PS
C
proble
m
.
Base
on
the
condi
ti
on
,
we
st
at
e
tha
t
the
b
asi
c
MO
P
is
stil
l
signifi
c
ant
to
sol
ve
the
PS
C
complex
probl
e
m
s using
MO
IN
LP
as
a
sophisti
c
at
ed
MO
P te
chn
i
que.
Ke
yw
or
d:
Crit
ic
al
p
at
h
m
et
hod
In
te
ger
Mult
i
-
obj
ect
iv
e
Neg
at
ive
total
float
Nonlinea
r
Pr
oject
sc
he
du
l
e com
pr
essio
n
So
lve
r
a
ppli
cat
ion
Ti
m
e
-
cost fu
nc
ti
on
Copyright
©
201
8
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Fach
rurr
azi
,
Do
ct
or
al
Stu
dy
Program
o
f
E
nginee
rin
g,
Syi
ah
K
uala
U
niv
e
rsity
,
Ace
h, I
ndonesi
a
.
Em
a
il
:
fach
rurraz
i
@
unsyi
ah
.a
c.id
1.
INTROD
U
CTION
In
s
om
e
of
the
pro
j
ect
s
cases
e
m
ph
asi
zi
ng
the
sc
hedule
ob
j
ect
ive
rat
her
t
han
t
he
co
st
obj
ect
iv
e
as
a
strat
egic
pro
j
e
ct
req
uiri
ng
im
m
ediat
el
y
i
m
p
act
,
will
giv
e
t
he
co
ntracto
r
a
chan
ce
to
pro
po
s
e
the
com
pr
essi
on
pro
gr
am
m
e
of
the
pro
j
ect
sch
edu
le
.
The
c
on
tract
or
s
s
houl
d
hav
e
a
bargai
ning
abili
ty
reg
ar
ding
the
a
ddit
ion
al
costs
as
a
c
ons
equ
e
nce
of
sch
edu
le
c
om
pr
es
sion
t
hat
co
uld
have
im
plicatio
ns
f
or
win
-
win
s
olu
ti
ons
to
pro
j
ect
par
ti
es.
I
n
co
ntrast
toward
the
pr
oject
s
costs
as
the
m
ai
n
ob
j
ect
ive
(m
ai
n
pr
iority
),
the
sc
hedule
com
pr
e
ssion
as
init
ia
ti
ve
of
the
c
ontract
or
is
m
ai
nly
base
d
on
t
he
reasons
t
o
de
m
on
str
at
e
their
pe
rfo
rm
ance
to
the
owne
r
[
1]
an
d
to
purs
ue
c
omplet
ion
ti
m
e
un
de
r
c
on
t
ract
due
to
delay
in
pro
j
ect
i
m
plem
entat
ion
[
2]
,
[3]
[
4]
,
et
c.
Ge
ner
al
ly
,
the
P
SC
c
ou
l
d
be
done
by
usi
ng
t
wo
te
ch
ni
qu
e
s,
t
he
fast
track
[
5]
,
c
rash
i
ng
[
6]
,
or
com
bin
in
g
bo
t
h
te
c
hniq
ue
s
[7]
.
I
n
rece
nt
decad
es
,
opti
m
iz
at
ion
is
the
i
m
po
rtant
pro
cess
and
c
onti
nuously
i
m
pr
oved
to
so
lve
m
or
e
co
m
plex
prob
le
m
s
in
m
any
fiel
ds
that
nee
d
to
be
sat
isfie
d,
suc
h
as
busi
ness
portf
olio
[8]
,
pro
j
ec
t
m
anag
em
ent
[9]
an
d
e
nginee
rin
g
a
pp
li
cat
io
ns
.
I
n
pro
per
s
equ
e
nce,
it
wil
l
pro
vid
e
t
he
e
nd
use
r
sat
isfa
ct
ion
reg
a
rd
i
ng the
pro
blem
so
lving an
d decisi
on
m
aking
[
10
]
.
A
te
ch
nique
t
o
obta
in
the
opti
m
al
so
luti
on
i
nvol
ving
tw
o
or
m
or
e
confli
ct
ing
ob
j
ect
ives
i
s
know
n
as
Mult
i
-
Objecti
ve
Program
m
ing
(M
OP).
S
om
e
of
the
M
O
P
that
ha
s
bee
n
dev
el
op
e
d
t
o
so
lve
the
pro
blem
is
seen
as
ver
y
c
om
plex
an
d
le
ss
pract
ic
able,
li
ke
a
ge
netic
al
gorithm
-
based
opti
m
iz
a
ti
on
[
11]
.
But
in
ste
ad
,
us
in
g
a
basic
m
od
el
will
on
ly
so
lve
si
m
ple
pr
oble
m
s,
li
k
e
MOP
by
go
a
l
pr
ogram
m
ing
[12]
.
In
the
P
rojec
t
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Ne
ga
ti
ve
T
ota
l
Flo
at to Im
pro
ve a Mult
i
-
obje
ct
iv
e In
te
ge
r N
on
-
l
ine
ar
Pr
og
ra
m
ming
…
(
Fachrurr
az
i
)
5293
Sche
du
li
ng
Prob
le
m
(P
SC),
us
in
g
the
ba
sic
MOP
t
o
s
olv
e
the
m
or
e
com
plex
pro
blem
s
is
a
chall
en
ging
ta
sk.
We
ha
ve
the
oppo
rtu
nity
to
reali
ze
this
chal
le
ng
in
g
ta
sk
by
inv
olv
i
ng
Ne
gative
Total
Float
(N
T
F)
to
w
ard
the
basic
MOP
m
od
el
and
we
cal
l
thi
s
m
od
el
as
MOIN
L
P
(M
od
el
O
bject
ive
In
te
ge
r
N
on
-
L
inear
Pro
gr
am
m
ing
).
MOI
NLP
is
use
d
to
so
l
ve
t
he
P
SC
case
with
inte
ger
va
riables
on
the
high
de
gr
ee
of
no
n
-
li
nea
r
ti
m
e
-
cost
functi
on.
Pro
ble
m
-
so
lvin
g
the
PSC
hav
in
g
a
nonlinea
r
tim
e
cost
m
od
el
will
con
trib
ute
to
a
m
or
e
com
pli
cat
ed
and
dynam
ic
of
co
nf
li
ct
ing
obj
ect
iv
e
than
a
li
near
.
T
her
e
f
or
e
,
buil
ding
a
su
it
able
m
od
el
and
it
s
optim
iz
at
ion
al
gorithm
w
il
l
be
a
n
im
po
rtan
t t
ask
[
13]
.
Unde
rstan
ding
the
co
nf
li
ct
in
g
ob
j
ect
ive
a
nd
it
s
causes
on
the
PSC
pro
ble
m
will
m
ak
e
i
t
easy
fo
r
us
to
dev
el
op
th
e
in
novative
m
od
el
to
reac
h
t
he
pro
blem
s
so
lving
[
14]
,
es
pecial
ly
in
ca
se
of
the
PSC
op
ti
m
iz
ation
pro
blem
.
Ba
sed
on
the
ba
ck
groun
d,
the
pu
rposes
of
this
res
earch
are
t
o
id
entify
the
conf
li
ct
ing
obj
ect
ive
of
P
SC
pro
blem
a
nd
to
dev
el
op
MOI
NLP
base
on
the
basic
m
od
el
of
P
SC
pro
blem
.
A
da
ta
set
of
pro
j
ect
case,
a
s
show
n
in
Ta
ble
1,
will
be
a
pp
li
ed
t
o
c
onduct
the
te
sti
ng
of
t
he
pro
pose
d
m
od
el
in
this
pap
e
r.
We
us
e
the
Tot
al
Float
par
am
et
er
as
a
n
ap
proach
to
ide
ntif
y
the
con
flic
ti
ng
o
bj
ect
ives
of
act
ivit
ie
s.
The
Total
Float
[15]
in
sche
du
li
ng
pr
ob
le
m
cou
ld
hav
e
Ze
ro
T
ot
al
Float
(ZTF
)
are
us
e
d
to
perform
crit
ical
path
analy
sis
[16]
a
nd
t
o
ide
ntify
the
crit
ic
al
act
ivit
y
[17]
;
Po
si
ti
ve
Total
Floa
t
(P
TF
)
is
the
m
axi
m
u
m
al
lo
wab
l
e
value
of
act
ivi
ti
es
wh
ic
h
do
e
s
no
t
ca
us
e
del
ay
of
pro
j
ect
[
18
]
;
an
d
Ne
gat
ive
Total
Float
(N
T
F)
is
us
e
d
as
a
const
raint
of
t
he
propose
d
MOI
NLP
m
odel
.
De
velo
ping
MOI
NLP
in
t
his
pa
pe
r
has
been
ad
opte
d
f
ro
m
the
basic
m
od
el
in
Dec
kro
’s
pa
per
[19]
,
wh
ic
h
is
a
sim
ple/s
ta
nd
a
rd
m
od
e
l
to
so
lve
t
he
li
near
case
of
m
ul
ti
-
obj
ect
ive
.
De
ve
lop
m
ent
of
t
hi
s
prose
d
m
odel
is
associat
ed
wit
h
N
TF
a
s
the
a
dd
it
io
na
l
con
st
raints
m
od
el
.
Pr
oble
m
-
so
lvi
ng
of
the
PSC
with
a
nonline
ar
m
od
el
of
th
e
pr
oject
act
ivit
y
is
i
m
ple
m
ente
d
us
in
g
the
So
lve
r
Applic
at
ion
of
Mi
cro
s
of
t E
xc
el
ad
d
-
ins
.
Pr
oject
sc
he
duli
ng
c
om
pr
essing
is
a
m
ulti
-
ob
j
ect
iv
e
pr
ob
le
m
inv
ol
ving
th
e
co
nf
li
ct
ing
obj
ect
iv
e
of
sever
al
act
ivit
i
es
f
un
ct
io
n,
es
pecial
ly
on
c
riti
cal
act
ivit
ie
s.
The
c
onflic
ti
ng
obj
ect
ive
will
inc
rease
i
n
li
ne
with
the
sm
al
le
r
due
date
ca
us
e
d
by
PSC
.
T
his
researc
h
has
be
en
s
uccess
f
ully
involvin
g
NTF
to
i
de
ntify
t
he
confli
ct
ing
ob
je
ct
ive
an
d
t
o
i
m
pr
ov
e
t
he
pe
rfor
m
ance
of
t
he
basic
M
OP
m
od
el
cal
le
d
as
MO
IN
L
P.
This
researc
h
has
be
en
de
velo
ped
to
intro
duce
N
TF
of
act
ivit
y
as
an
ind
ic
at
or
of
act
ivit
ie
s
that
it
s
du
rati
on
can
be
reduce
d.
T
his
i
s
in
li
ne
with
t
he
L
IM
pa
pe
r
[16]
but
in
t
his
pa
per
,
we
ha
ve
bro
ught
the
NTF
a
s
the
im
portant
par
am
et
er
to
so
lve
the PSC
f
or
both
sim
ple
and
c
om
plex
pro
blem
.
This
re
search
delibe
r
at
el
y
eng
ineere
d
N
T
F
to
know
the
act
ivit
ie
s
to
be
a
ccel
erated.
F
urt
her
m
or
e,
we
de
fine
NT
F
as
an
autom
at
ic
maxim
u
m
value
of
the
act
ivit
y
du
rati
on
re
du
ct
io
n.
This
is
a
novel
ty
in
this
resea
rch.
T
his
resea
rch
al
s
o
util
iz
es
the
NTF
t
o
i
den
ti
fy
th
e co
nfl
ic
ti
ng
obj
ect
ive
of ac
ti
vity
f
un
ct
io
n i
n
MO
IN
L
P.
The
M
OINLP
is
a
de
velo
pm
ent
m
od
el
of
the
basic
M
OP
t
hat
re
fer
s
to
De
ck
ro’s
pap
e
r
[19]
by
involvin
g NTF
. W
e
ha
ve
sim
ulate
d
the M
OINLP
us
in
g
the
So
lve
r
A
ppli
cat
ion
to
s
olv
e t
he
PS
C p
roblem
w
it
h
the
tim
e
-
cost
act
ivit
ie
s
fu
nction
by
2
to 6
de
gr
ee
.
A
pp
ly
in
g
the
NTF
to
t
he
basic
MOP m
od
el
fo
r
the
c
om
plex
pro
blem
of
PSC
is
a
novelty
i
n
this
stu
dy.
R
el
at
ing
the
PS
C,
we
sta
te
tha
t
the
NTF
is
th
e
m
axi
m
u
m
al
l
ow
a
ble
value
t
o
r
ed
uc
e
the
act
ivit
y
durati
on
in
or
der
t
o
ac
hieve
the
du
e
date
of
t
he
P
SC.
T
he
re
su
lt
s
s
ho
wed
that
there
was
a
r
el
at
ion
sh
i
p
bet
ween
NTF
an
d
the
act
ivit
ie
s
wh
ic
h
s
houl
d
be
acce
le
rated
to
ob
ta
in
optim
al
so
luti
on
da
n
fe
asi
ble so
l
ution.
Besi
des
t
hat,
we have
a
pr
es
um
ption
that
N
TF w
il
l
im
pr
ove
the
s
pee
d up
o
f
t
he
op
ti
m
iz
ation
proces
s
as
a
fu
t
ur
e
re
searc
h
f
or
oth
e
r
resea
rc
her
s
.
W
e
ar
gu
e
that
NTF
has
an
i
m
po
rta
nt
r
ole
to
gen
e
rate
the
c
om
plex
prob
le
m
so
lving
of
MOI
NLP
(
non
-
li
near
functi
on
an
d
hi
gh
de
gr
ee
var
ia
bles)
.
Thi
s
conditi
on
can
be
a
sta
rtin
g
poi
nt
to
refres
h
the
us
e
of
a
sim
ple
MOP
m
od
el
,
as
Deckr
o's
researc
h
[19]
,
but
sti
ll
h
as the
po
wer t
o
s
olv
e
th
e lat
est
an
d m
or
e c
om
plex
PS
C pro
blem
s
.
2.
METHO
D
2.1.
Metho
d Cha
r
act
eri
s
tic
The
m
et
ho
d
of
identify
in
g
th
e
confli
ct
ing
obj
ect
iv
e
of
the
act
ivit
ie
s
in
the
PSC
pro
blem
is
propose
d
us
in
g
an
analy
sis
of
the
NTF
as
a
par
am
e
ter
of
T
otal
Float
of
act
ivit
ie
s.
The
Total
Floa
t
analy
sis
is
based
on
the
pri
nciple
of
the
Crit
ic
al
Path
A
naly
sis.
The
T
otal
Flo
at
of
eac
h
act
ivit
y
cou
l
d
ha
ve
a
ne
gative
va
lue
as
Neg
at
ive
T
otal
Float
(N
TF
),
a
po
sit
ive
val
ue
as
Po
sit
ive
To
ta
l
Float
(P
TF)
[20]
,
or
eve
n
a
zero
val
ue
as
Zero
Total
Float
(
Z
TF)
[20]
.
NTF
of
t
he
proj
ect
act
ivit
y
can
oc
cur
i
f
the
la
te
st
finish
(L
F)
is
le
ss
than
ea
rly
finis
h
(EF)
in
the
te
r
m
inal
no
de
of
the
CP
M
netw
ork
dia
gr
am
.
Wh
e
re,
t
he
LF
in
a
te
rm
inal
node
s
hould
be
equ
al
with
the
P
rojec
t
Crashi
ng
(
Pc
).
T
he
dif
fer
e
nc
e
betw
een
the
two
pa
ram
et
ers
(L
F
a
nd
EF
)
is
the
value
y
of
the
PSC, as
sho
wn in F
i
gure
3.
The
m
et
ho
d
t
o
so
lve
t
he
PSC is
MOI
NLP
m
od
el
.
The
MOI
NLP
is
a d
e
velop
m
ent
m
od
el
of
t
he
ba
sic
con
ce
pt
of
M
OP
ref
e
rr
i
ng
Deckr
o'
s
pap
e
r
[
19
]
.
This
MOP
m
od
el
will
be
de
velo
ped
an
d
inte
grat
ed
by
app
ly
in
g
NTF,
as
a
n
a
dd
it
io
na
l
co
ns
trai
nt
f
or
MO
IN
L
P
m
od
el
,
to
so
l
ve
com
plex
sche
duli
ng
pr
ob
le
m
s
(this
stud
y
will
te
st
the
act
ivit
ie
s
t
i
m
e
-
cost
functi
on
o
f
2
t
o
6
de
gr
ee
).
De
velo
pi
ng
the
MO
INLP
al
gorithm
n
eeds t
o
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
5292
-
5303
5294
descr
i
be
the
operati
on
al
perform
ance
of
the
CPM
con
ce
pt
(
Crit
ic
al
Path
Me
thod)
[
21]
.
The
il
lustrati
on
of
the
CPM
algorit
hm
is show
n
in
form
ula (
1) th
r
ough
(10).
Applyi
ng
the
MOI
NLP
m
odel
us
es
the
S
olv
er
A
pp
li
cat
io
n
of
Mi
cr
osoft
Excel
Add
-
ins
.
The
S
olv
e
r
Applic
at
ion
is
an
opti
m
iz
a
ti
on
to
ol u
si
ng
t
he
g
oal
pro
gr
am
m
ing
alg
or
it
hm
. Th
e ease of
the
So
l
ver
is
due to its
si
m
plici
t
y
in
im
ple
m
enting
m
at
he
m
at
ic
a
l
m
od
el
s
into
ta
bu
la
ti
on
a
nd
c
el
ls
fo
rm
ula
with
m
or
e
fle
xib
il
it
y.
Ev
olu
ti
onary
Mult
i
-
Objecti
ve
(EMO)
m
et
ho
d
of
the
So
l
ve
r
will
be
us
e
d
to
pro
blem
s
requirin
g
the
i
ntege
r
ou
t
pu
t
a
nd
sm
oo
t
h
s
olu
ti
on
[
22
]
.
EMO
ca
n
be
us
e
d
to
so
l
ve
sin
gle
-
obj
e
ct
ive
optim
iz
ation
pro
blem
s
with
a
fo
c
us
on
fin
ding a
sin
gle opti
m
al
so
luti
on
[
23]
.
The
us
e
of
th
e
popula
ti
on
i
n
e
vo
l
utionary
m
ult
i
-
obj
ect
ive
al
lo
ws
t
he
So
l
ver
to
ha
ve
par
al
le
l
searchi
ng
abili
ty
to
find
m
ult
iple
non
-
dom
i
nated
s
olu
ti
ons
in
a
sin
gle
it
erati
on
[
13
]
.
T
he
So
lv
er
al
go
rithm
s
a
re
base
d
on
a
go
al
pro
gra
m
m
ing
us
ing
the
par
am
et
ers
of
the
ob
j
ect
ive
f
un
ct
io
n
m
od
el
(as
ta
r
get
cel
l),
const
raints
m
od
el
(as
sub
j
ect
to
the
con
st
raints),
a
nd
fin
din
g
the
value
of
var
ia
bles
(as
changin
g
cel
ls)
[24]
.
The
fr
am
ewor
k
for
s
olv
i
ng
t
he
P
SC
pro
blem
us
ing
th
e
propose
d
M
OIN
LP
is
desc
ribe
d
in
the
f
ram
e
work
as
sh
ow
n
in
Fi
gure
1.
T
he
MO
I
NLP
m
od
el
wi
ll
be
app
li
ed
to
a
proj
ect
c
on
s
ist
ing
of
12
no
n
-
li
nea
r
f
unct
ion
s
of
pro
j
ect
acti
vity
(
the
fun
ct
io
n o
f 2 to
6 de
gr
ee
)
.
Figure
1.
The
fram
ewo
rk of
MOI
NLP
f
o
r P
roject
S
c
he
du
li
ng Com
pr
essio
n
(P
SC)
2.2.
The CP
M Alg
orithm
The
al
gorithm
for
CPM
ref
e
rs
to
the
for
ward
and
the b
ac
kward
p
r
oce
dure.
The
f
orwa
rd
p
r
ocedu
re
of
CPM
is
us
e
d
t
o
a
naly
ze
Earl
y
Start
(ES)
a
nd
Ea
rlie
st
Fini
sh
(EF),
w
hile
the
bac
kwar
d
proce
dure
of
C
PM
is
us
e
d
to
dete
rm
ine
Lat
est
Start
(LS),
Lat
est
F
inish
(L
F
)
.
Th
e
CPM
al
go
rit
hm
has
been
de
velo
ped
a
nd
a
pp
li
ed
in
m
any
fiel
ds
for
a
long
tim
e
and
wide
sp
rea
d
as
it
was
init
ia
te
d
by
Du
P
ont
Com
pan
y,
Kell
ey
,
and
Walke
r
[
25]
, a
s the
fo
ll
owin
g f
or
m
ula f
r
om
(1)
to
(1
0)
.
2.2.1.
T
he for
w
ard
pr
oced
u
re o
f CPM
a
l
gori
th
m
a.
The rule
of the
early
start
of
e
ach
pro
j
ect
acti
viti
es are:
ES
(
n
)
=
Max
{
ES
(
n
p
)
+
D
np
|
n
p
∈
set
of
imme
dia
t
e
pr
ede
cessor
s
of
act
iv
ity
n
}
(1)
b.
At
the
i
niti
al
node
of
a
pro
j
ect
(t
he
sta
rti
ng
tim
e
of
a
pro
j
ect
)
a
nd
the
init
ia
l
act
ivit
ie
s
ha
ve
no
pr
e
decess
or. T
hen, t
he
rule
of the ea
rly
start
tim
e fo
r
eac
h
i
niti
al
acti
vity
i
s 0
(ze
ro).
if
n
p
is e
m
pty t
hen
,
ES
(
n
_
in
itial
)
=
0
a
nd
,
ES
(
Proje
ct
)
=
0
(2)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Ne
ga
ti
ve
T
ota
l
Flo
at to Im
pro
ve a Mult
i
-
obje
ct
iv
e In
te
ge
r N
on
-
l
ine
ar
Pr
og
ra
m
ming
…
(
Fachrurr
az
i
)
5295
c.
At
the
en
d
node
of
a
pro
j
ect
wh
e
re
the
re
ar
e
no
su
cces
sor
s
of
act
ivit
ie
s.
The
n,
the
r
ule
of
ea
rly
finis
h
of
each last
ac
ti
vi
ti
es are:
EF
(
n
_
last
)
=
Max
{
EF
(
n
_
las
t
)
+
D
n
_
last
|
n
last
∈
set
of
th
e
last
act
iv
itie
s
of
a
p
roje
ct
}
(3)
EF
(
Proje
ct
)
=
Max
{
ES
(
n
_
last
)
|
n
last
∈
set
of
th
e
last
act
iv
itie
s
of
a
p
roje
ct
}
(4)
2.2.2.
T
he b
ac
k
w
ard proce
d
ure of
CP
M algo
ri
th
m
a.
At
the
e
nd
no
de
of
a
pro
j
ect
wh
e
re
the
re
a
r
e
no
su
c
cess
ors
de
pende
ncies
of
a
ct
ivit
ie
s.
The
n,
the
r
ule
of
the
la
te
st
finis
h o
f
eac
h
la
st ac
ti
viti
es are:
LF
(
Pro
je
ct
)
=
EF
(
Proje
ct
)
(5)
LF
(
n
_
las
t
)
=
LF
(
Pro
je
ct
)
(6)
b.
The rule
of
the
la
te
st
finish o
f
each p
roject a
ct
ivit
ie
s ar
e:
LF
(
n
)
=
Mi
n
{
LF
(
n
s
)
+
D
ns
|
n
s
∈
set
of
imm
edi
a
te
suc
cess
ors
of
act
iv
ity
n
}
(7)
c.
At
the
init
ia
l
node
of
the
pro
j
ect
(the
sta
rting
tim
e
of
a
pro
j
ect
)
an
d
the
i
niti
al
act
ivit
ie
s
hav
e
no
pr
e
decess
or. T
hen, the
rule
of the ea
rly
start
tim
e fo
r
eac
h
i
niti
al
acti
vit
y is
0 (ze
ro).
if
n
p
is e
m
pty t
hen
,
LS
(
n
_
in
itial
)
=
0
(8)
LS
(
Pro
je
ct
)
=
0
(9)
2.2.3.
T
he To
t
al Flo
at A
na
l
ysi
s o
f
e
ach acti
vit
y
in
CP
M
TF
(
n
)
=
LF
(
n
)
−
D
n
−
ES
(
n
)
(10)
The
m
ulti
-
obje
ct
ive
pro
ble
m
in
pro
j
ect
sche
du
li
ng
c
om
pr
ession
in
volvin
g
non
-
li
ne
ar
ti
m
e
/c
os
t
act
ivit
y
m
od
el
to
achie
ve
the
m
ini
m
u
m
ta
rg
et
cost
of
the
pro
j
ect
will
be
a
naly
zed
us
in
g
t
he
data
as
il
lus
trat
ed
in Ta
ble 1.
Table
1.
Data
of the
pro
j
ect
a
ct
ivit
ie
s f
or PS
C
No
Activ
ities
ID
Predecess
o
r
Su
ccesso
r
No
r
m
al
Duratio
n
(I
n
itial Dur
atio
n
)
The
m
ax
i
m
u
m
a
m
o
u
n
t o
f
Co
m
p
r
ess
in
g
Dura
tio
n
Activity
No
n
-
lin
ear
Ti
m
e/C
o
st
Mod
el of
eac
h
Act
iv
ity
(
n
)
(
n
p
)
(
n
s
)
(
D
n
)
(
X
n
ma
x
)
f
(
X
n
)
1
A
-
D,
E
12
9
37x
A
2
+
x
A
2
B
-
F,
H
12
9
30x
B
3
3
C
-
L
25
20
2x
C
3
4
D
A
G,
I
20
16
x
D
3
+
x
D
2
+
x
D
5
E
A
J
23
18
2x
E
3
−
x
E
2
6
F
B
G,
I
8
6
2x
F
5
−
x
F
4
7
G
D,
F
J
11
8
x
G
4
+
x
G
8
H
B
K
25
20
7x
H
2
9
I
D,
F
K
8
6
x
I
6
−
2x
I
5
+
3x
I
10
J
E, G
-
18
14
5x
J
2
+
5x
J
11
K
H,
I
-
15
12
2x
K
4
12
L
C
-
30
24
5
x
L
2
3.
RESU
LT
S
3.1.
Initial
Net
w
or
k D
i
ag
r
am
Sc
heduli
ng
Ba
sed
on
the
inf
or
m
at
ion
in
Table
1,
we
c
onstr
uct
the
pro
je
ct
netw
ork
di
agr
am
us
in
g
C
PM
analy
sis.
This
a
naly
sis
gi
ves
t
he
res
ult
of
EF
of
61
da
ys
as
the
no
rm
al
pro
j
ect
due
date
(
P
n
)
a
nd
a
c
riti
cal
path
c
onsist
of the c
riti
cal
acti
vity
o
f A, D,
G
,
J,
a
s s
how
n i
n
Fi
gure
2.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
5292
-
5303
5296
Figure
2. Net
w
or
k diag
ram
f
or the
nor
m
al
p
r
oj
ect
at
the i
niti
al
3.2.
Ident
if
yin
g th
e Conf
li
ctin
g Objec
tive usi
n
g Nega
tive
T
otal Fl
oat (NT
F)
Seve
ral
pr
e
vious
stu
dies
ha
ve
su
ggest
e
d
th
at
the
op
ti
m
izati
on
pro
blem
so
lvi
ng
of
m
ulti
-
obj
ect
ives
will
inv
ol
ve
a
nu
m
ber
of
c
on
flic
ti
ng
act
ivit
y
[19]
,
[
26
]
.
U
nfor
t
un
at
el
y,
t
he
stu
dy
does
no
t
e
xp
la
i
n
w
hi
ch
one
of
t
he
c
onflic
ti
ng
act
ivit
ie
s.
We
hav
e
de
velop
e
d
a
s
et
proc
edure
to
ide
ntify
the
c
onflic
ti
ng.
We
ha
ve
re
placed
LF
with
P
c
in
th
e
n
et
w
ork
diag
ram
syst
e
m
analy
sis
as
sh
ow
n
in
Fig
ur
e
3
t
o
br
in
g
up
the
diff
e
re
nce
betwe
e
n
LF and EF
. T
hi
s co
ndit
ion
wi
ll
sh
ow N
T
F on
s
om
e activities ind
ic
at
ed
a
s
confli
ct
ing
act
i
viti
es that nee
d t
o
be
sat
isfie
d
to
ac
hieve
the
du
e
date
of
the
ac
cel
erated
pro
j
e
ct
.
As
an
il
lustrati
on
of
the
va
lue
of
NTF
(
on
th
e
act
ivit
ie
s
of
A
,
D,
G,
a
nd
J
w
it
h
NT
F
eac
h
of
-
6)
t
hat
m
ay
occur
due
to
y
of
6
days,
as
sh
ow
n
in
Figur
e
3.
In
the
case
of
P
S
C
for
y
of
1
to
12
days
will
gi
ve
res
ults
to
NTF
an
d
var
i
ous
c
onflic
ti
ng
act
ivit
ie
s,
as
s
how
n
in
Table
2.
The
a
ct
ivit
ie
s
of
A,
D,
G,
a
nd
J
a
r
e
com
bin
ed
ac
ti
viti
es
to
ob
ta
in
the
opti
m
a
l
so
luti
on
a
nd
vi
sible
so
luti
on
of PS
C for y
for 6
da
ys, as s
how
n
i
n
Ta
b
le
3.
This
a
naly
sis
will
con
t
rib
ute
to
underst
an
di
ng
t
he
c
om
bi
nation
of
acce
l
erated
act
ivit
y
(con
flic
ti
ng
act
ivit
ie
s)
on
PSC
us
i
ng
M
OINLP
.
T
he
num
ber
of
c
onflic
ti
ng
obj
e
ct
ives
m
ay
va
ry,
de
pe
nd
i
ng
on
the
netw
ork
m
od
el
and
t
he
num
ber
of
PSC p
la
nned
.
I
n
thi
s
stu
dy,
we
i
den
ti
fy
the
act
ivit
ie
s
t
hat
ha
ve
ex
pe
rience
d
confli
ct
ing
obje
ct
ive.
I
n
the
act
ual
i
m
ple
m
entat
ion
of
the
sche
du
le
,
NT
F
co
ndit
ion
s
possibly
occur.
It
is
pro
bab
ly
cause
d
by
a
delay
in
the
pr
e
deces
so
r
act
ivit
y
or
cause
d
by
the
act
ivit
y
that
has
be
en
im
ple
m
ente
d
la
te
r
than
the
pl
ann
e
d
act
ivit
y
du
rati
on,
espe
ci
al
ly
in
the
crit
ic
al
activity
.
T
his
conditi
on
is
si
m
i
la
r
to
the
case
of
PSC,
w
her
e
the
com
pr
esse
d
durati
on
of
t
he
act
ivit
y
is
conditi
on
e
d
as
a
crit
ic
al
act
ivit
y
carried
out
faster
than
it
s
norm
al
tim
e.
Figure
3.
Pro
j
e
ct
Netw
ork dia
gr
am
w
it
h
m
odific
at
ion
in
the
la
te
st
finish
of
CPM
te
rm
inal
node
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Ne
ga
ti
ve
T
ota
l
Flo
at to Im
pro
ve a Mult
i
-
obje
ct
iv
e In
te
ge
r N
on
-
l
ine
ar
Pr
og
ra
m
ming
…
(
Fachrurr
az
i
)
5297
Unde
r
the
init
ia
l
analy
sis
of
the
netw
ork
diagr
am
,
the
LF
value
in
net
work
dia
gr
am
is
t
he
sam
e
a
s
with
it
s
EF
of
61
days
(t
he
init
ia
l
du
e
dat
e
of
t
he
pro
j
e
ct
),
as
s
how
n
in
Fig
ure
2.
A
fter
c
om
pr
essi
ng
the
pro
j
ect
sche
du
l
e
with
y
(
0
to
12
days
)
,
as
show
n
in
Ta
ble
2,
s
hows
s
om
e
act
ivit
ie
s
hav
ing
t
o
va
ry
NT
F.
It
is
cond
ucted
by
change
L
F
in
the
te
rm
inal
no
de
to
be
Pc
,
for
e
xam
ple
a
s
show
n
i
n
Fi
gure
3.
Ba
se
d
on
t
he
resu
lt
s
of
TF
c
(55
days)
on
t
he
Ta
ble
2,
both
ne
gative
a
nd
posit
ive
va
lue,
in
dicat
es
t
hat
if
t
he
total
float
norm
al
(TF
n
)
of
the
act
ivit
y
is
le
ss
than
th
e
va
lue
of
y
in
PSC
then
the
TF
of
the
act
ivit
y
will
be
neg
at
ive
value
(it
will
exp
e
rience
c
onf
li
ct
ing
obj
ect
iv
e
of
act
ivit
ie
s
functi
on).
W
e
con
cl
ud
e
that
the
great
er
t
he
valu
e
of
y
will
m
or
e
co
nf
li
ct
ing
ac
ti
viti
es
and
it
NTF
value
.
T
he
e
xam
ple
in
Table
2
with
cases
1
t
hro
ugh
12,
ind
ic
a
ti
ng
a
c
ha
ng
e
in
t
he
value
of
Total
Fl
oa
t
le
ading
to
a
n
inc
reasin
gly
la
rg
e
NTF.
NT
F
of
s
om
e
act
i
viti
es
exp
e
riences
a
n
increase
in
both
ne
gative
va
lues
an
d
the
nu
m
ber
of
co
nf
li
ct
ing
act
ivi
ti
es
al
on
g
wit
h
the
increasin
g o
f
y
in
P
SC, as
sho
wn in T
able
2.
Table
2.
T
he
c
onflic
ti
ng
o
bj
e
ct
ive
of P
SC usi
ng
MOI
NL
P
with y
of 0 t
o 1
2 days
Cas
es o
f
Co
m
p
r
es
sio
n
Project Inf
o
r
m
atio
n
Total Flo
at of
Acti
v
ities (T
F
c
)
Co
n
f
lictin
g
Objectiv
e
(Co
n
f
lictin
g
Activi
ty
of
Pr
o
ject
Sch
ed
u
le)
Initial
Du
e date
o
f
Pr
o
ject
(
)
Project
Sch
ed
u
lin
g
Co
m
p
r
ess
io
n
(
y
=
-
)
Du
e date
af
ter
Co
m
p
r
ess
io
n
(
)
A
B
C
D
E
F
G
H
I
J
K
L
Initial
61
0
61
0
9
6
0
8
12
0
9
6
0
6
6
No
Cas
e 1
61
1
60
-
1
8
5
-
1
7
11
-
1
8
5
-
1
5
5
A,
D
,
G
,
J
Cas
e 2
61
2
59
-
2
7
4
-
2
6
10
-
2
7
4
-
2
4
4
A,
D
,
G
,
J
Cas
e 3
61
3
58
-
3
6
3
-
3
5
9
-
3
6
3
-
3
3
3
A,
D
,
G
,
J
Cas
e 4
61
4
57
-
4
5
2
-
4
4
8
-
4
5
2
-
4
2
2
A,
D
,
G
,
J
Cas
e 5
61
5
56
-
5
4
1
-
5
3
7
-
5
4
1
-
5
1
1
A,
D
,
G
,
J
Cas
e 6
61
6
55
-
6
3
0
-
6
2
6
-
6
3
0
-
6
0
0
A,
D
,
G
,
J
Cas
e 7
61
7
54
-
7
2
-
1
-
7
1
5
-
7
2
-
1
-
7
-
1
-
1
A,
C,
D,
G,
I,
J,
K
,
L
Cas
e 8
61
8
53
-
8
1
-
2
-
8
0
4
-
8
1
-
2
-
8
-
2
-
2
A,
C,
D,
G,
I,
J,
K
,
L
Cas
e 9
61
9
52
-
9
0
-
3
-
9
-
1
3
-
9
0
-
3
-
9
-
3
-
3
A,
C,
D,
E,
G,
I
,
J,
K,
L
Cas
e 10
61
10
51
-
10
-
1
-
4
-
10
-
2
2
-
10
-
1
-
4
-
10
-
4
-
4
A,
B,
C,
D
,
E,
G
,
H,
I
,
J,
K,
L
Cas
e 11
61
11
50
-
11
-
2
-
5
-
11
-
3
1
-
11
-
2
-
5
-
11
-
5
-
5
A,
B,
C,
D
,
E,
G
,
H,
I
,
J,
K,
L
Cas
e 12
61
12
49
-
12
-
3
-
6
-
12
-
4
0
-
12
-
3
-
6
-
12
-
6
-
6
A,
B,
C,
D
,
E,
G
,
H,
I
,
J,
K,
L
Me
thod
f
or
ca
lc
ulati
ng
T
otal
Float
of
ea
ch
act
ivit
y
after
set
ti
ng
P
c
(the
du
e
date
of
P
roject
sc
heduli
ng
Com
pr
ession)
us
es t
he
m
at
hem
at
ic
al
f
or
m
ul
at
ion
s as
s
how
n
in
for
m
ula (1
1).
TF
c
=
TF
n
−
y
(11)
Wh
e
re
TF
c
is
the
Total
Float
of
act
ivit
y
after
the
PSC
(To
ta
l
Float
co
uld
be
PTF,
Z
TF,
a
nd
NTF);
TF
n
is
the
init
ia
l
Total
Fl
oat
of
act
ivit
y
(the
Total
Floa
t
of
act
ivit
y
bef
ore
the
PSC);
and
y
is
the
co
m
pr
essed
durat
ion
of
the pr
oj
ect
.
3.3.
Dev
el
op
in
g
the
MO
INLP
b
a
sed on
th
e
ba
s
ic
M
OP
for Pr
oj
ec
t
Sc
heduli
ng
C
omp
ressi
on
In
this
st
ud
y,
we
a
pp
ly
t
he
con
ce
pt
of
N
e
gative
T
otal
F
loat
(
NTF)
to
M
ulti
-
O
bject
iv
e
M
odel
s
,
as
a
const
raint
t
hat
m
us
t
be
sat
is
fied
t
o
s
olv
e
the
m
ini
m
izing
pro
blem
of
th
e
ob
j
ect
ive
functi
on.
De
velo
ping
a
m
ul
ti
-
obj
ect
ive
m
od
el
al
so
re
fer
s
to
the
M
OLP
m
od
el
de
velo
ped
by
De
ckro
[
19
]
.
T
he
dev
el
opm
ent
of
the
m
od
el
w
il
l be
exp
la
ine
d as f
ol
lows
:
1.
The o
b
j
ect
ive
fun
ct
io
n (
T
he
S
olv
e
r
: t
arget
ce
ll
):
Mi
n
Z
(
y
)
=
Mi
n
∑
(
(
f
(
x
n
)
|
x
n
∈
{
0
,
1
,
2
,
3
,
…
,
x
n
max
}
)
m
n
=
1
(12)
Mi
n
Z
(
y
)
=
Mi
n
∑
(
(
37x
A
2
+
x
A
)
+
(
30x
B
3
)
+
(
2x
C
3
)
+
(
x
D
3
+
x
D
2
+
x
D
)
+
(
2x
E
3
−
2
x
E
2
)
+
(
2x
F
5
−
x
F
4
)
+
(
x
G
4
+
x
G
)
+
(
7x
H
2
)
+
(
x
I
6
−
2x
I
5
+
3x
I
)
+
(
5x
J
2
+
5x
J
)
+
(
2x
K
4
)
+
(
5
x
L
2
)
|
x
n
∈
{
0
,
1
,
2
,
3
,
…
,
x
n
max
}
)
(13)
2.
The varia
ble
of the m
od
el
(
Th
e So
l
ver
:
by c
ha
ng
i
ng cell
s),
x
n
:
x
A
;
x
B
;
x
C
;
x
D
;
x
E
;
x
F
;
x
G
;
x
H
;
x
I
;
x
J
;
x
K
;
x
L
(14)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
5292
-
5303
5298
3.
The
c
onstrai
nts
of the
m
od
el
(
The Sol
ver
:
s
ubj
ect
t
o
t
he
c
onstrai
nts)
:
a.
The
c
om
pr
esse
d durati
on fo
r e
ach acti
vity
(
x
n
)
≥ 0,
it
is a c
onstrai
nt of t
he n
on
-
neg
at
ive
.
x
A
;
x
B
;
x
C
;
x
D
;
x
E
;
x
F
;
x
G
;
x
H
;
x
I
;
x
J
;
x
K
;
x
L
≥
0
(15)
This
de
vel
op
e
d
co
ns
trai
nt
is
a
non
-
ne
gative
const
raint
to
e
ns
ure
that
t
he
a
m
ou
nt
of
c
om
pr
essed
dura
ti
on
f
or
each
act
ivit
y
is
gr
eat
e
r
tha
n
or
e
qual
to
ze
ro.
The
c
onstr
ai
nt
is
in
li
ne
with
the
pap
e
r
was
de
velo
pe
d
by
Deckr
o
[
19]
.
b.
The
a
ddit
ion
al
cost fo
r
eac
h
a
ct
ivit
y (
C
n
)
≥ 0,
it
is a constrai
nt
of the
non
-
ne
ga
ti
ve.
C
A
;
C
B
;
C
C
;
C
D
;
C
E
;
C
F
;
C
G
;
C
H
;
C
I
;
C
J
;
C
K
;
C
L
≥
0
(16)
This
de
velo
pe
d
co
ns
trai
nt
is
a
no
n
-
neg
at
ive
const
raint
to
ensure
th
at
the
add
it
io
nal
cost
due
to
the
com
pr
essed
du
rati
on
f
or
each
act
ivit
y
is
gr
e
at
er
tha
n
or
eq
ual
to
zer
o.
Th
e
co
ns
trai
nt
is
in
li
ne
with
t
he
pa
pe
r
was de
velo
ped b
y
Deckr
o
[19]
.
c.
The
c
om
pr
esse
d durati
on fo
r e
ach th
e
acti
viti
es (
x
n
)
≤
(
x
n
max
)
.
x
A
≤
6
;
x
B
≤
6
;
x
C
≤
13
;
x
D
≤
10
;
x
E
≤
12
;
x
F
≤
4
;
x
G
≤
6
;
x
H
≤
13
;
x
I
≤
4
;
x
J
≤
12
;
x
K
≤
8
;
x
L
≤
16
(17)
This c
onstrai
nt
w
il
l ens
ur
e
tha
t t
he
act
ivit
ie
s ar
e
no
t al
lo
we
d ov
e
r
t
he
li
m
itati
on
of the m
axim
u
m
a
m
ou
nt of
crash
i
ng
[
19]
.
d.
The
c
om
pr
esse
d durati
on fo
r e
ach acti
vity
(
x
n
)
= I
ntege
r.
It is
the
inte
ger re
quirem
ent o
f d
urat
ion
reducti
on fo
r
t
he
act
ivit
y. T
his is the
flexi
ble fun
ct
io
n
t
hat c
ou
l
d be
rem
ov
ed
if
it
is not th
e intege
r
case
.
x
A
;
x
B
;
x
C
;
x
D
;
x
E
;
x
F
;
x
G
;
x
H
;
x
I
;
x
J
;
x
K
;
x
L
=
Integer
(18)
e.
The
ea
rlie
st co
m
ple
ti
on
ti
m
e
of project
≥
Pro
je
ct
co
m
pr
essio
n du
e
d
at
e
.
This
co
ns
trai
nt
arises
fr
om
the
rep
la
cem
ent
of
the
LC
(pr
oject
)
in
the
la
st
node
(project
te
rm
inal
no
de
)
to
be
the
P
SC
due
da
te
(
P
c
)
. P
SC
du
e d
at
e is a
tar
ge
t for
pro
j
ect
f
i
nish.
EF
(
Proje
ct
)
≤
P
c
(19)
This
co
ns
trai
nt
assur
es
Ea
rly
Finish
in
t
he
te
rm
inal
no
de
(
EF)
m
us
t
be
le
ss
than
or
e
qual
to
the
du
e
da
te
of
Pr
oject
sc
he
du
l
ing
c
om
pr
essio
n (
P
c
)
[
19
]
f.
Neg
at
ive
T
otal
Float
(N
TF
)
of
eac
h
act
ivit
y
≥
0
(zero).
This
co
ns
trai
nt
is
to
ensu
re
the
NTF
in
s
om
e
act
ivit
ie
s
m
us
t be m
or
e tha
n o
r
e
qu
al
t
o
ze
ro.
Th
is c
onst
rain
t i
s one
of
m
odel
d
evel
op
m
ent in th
is
p
a
per.
NTF
c
≥
0
(20)
3.4.
Implem
entin
g MO
INLP
Mo
del
U
sin
g
The
So
lver
This
stu
dy
i
m
pr
oves
MO
INL
P
by
ad
ding
N
egati
ve
Total
F
loat
(N
TF
)
co
nst
raints,
w
hich
ind
ic
at
es
as
a
value
that
shou
l
d
be
sat
isfi
ed
to
achie
ve
the
ex
pected
P
SC.
The
sat
isf
act
ion
of
act
ivit
ie
s
that
hav
e
NTF
is
achieve
d
by
re
du
ci
ng
t
he
dur
at
ion
of
the
ac
ti
vity
(
).
S
om
e
of
the
act
ivit
y
com
bin
at
ion
s
can
pote
ntial
ly
be
sat
isfie
d
(r
e
du
ced)
to
ac
hiev
e
PSC
ta
rg
et
s.
The
best
c
ombinati
on
of
act
ivit
y
sel
ect
ion
is
the
m
ini
m
u
m
total
cost
of
so
m
e
act
ivit
ie
s
(as
a
m
ini
m
u
m
add
it
ion
al
c
os
t
of
t
he
pro
j
ect
)
bas
ed
on
the
tim
e
-
cost
f
un
ct
io
n
of
t
he
act
ivit
y.
Table
3
an
d
Ta
ble
4
sh
ow
the outp
ut
PSC
for
y
at
6
an
d
7
days (pr
oject
durati
on
re
duct
ion)
us
ing
the
MOI
NLP
m
ode
l sim
ulate
d
by
the
S
olv
e
r. Th
e outp
ut d
e
scri
bes res
ults with tw
o
cat
e
gori
es,
nam
ely:
a.
The o
pti
m
al
so
luti
on
sho
ws
th
e m
ini
m
u
m
v
al
ue of
outp
ut
obj
ect
ive.
b.
The feasi
ble s
ol
ution
is
the
res
ult o
f
the
so
l
ution sp
ace a
nd it
is sti
ll
n
ot a
m
ini
m
u
m
[27]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Ne
ga
ti
ve
T
ota
l
Flo
at to Im
pro
ve a Mult
i
-
obje
ct
iv
e In
te
ge
r N
on
-
l
ine
ar
Pr
og
ra
m
ming
…
(
Fachrurr
az
i
)
5299
Ta
ble
3.
T
he
s
olu
ti
on
for
M
O
IN
L
P fo
r
P
SC
of 6 days
(Pro
je
ct
d
ue
d
at
e
55
d
ay
s)
Activ
ities
No
r
m
al
Du
ration
Initial TF
TF
crashin
g
Op
ti
m
al Solu
tio
n
Feasib
le Solu
tio
n
Mod
e 1
Mod
e 2
Mod
e 3
Mod
e 4
n
D
n
TF
n
TF
c
x
n
C
n
TF
n
x
n
C
n
TF
n
x
n
C
n
TF
n
x
n
C
n
TF
n
A
12
0
-
6
0
0
0
1
38
0
1
38
0
0
0
0
B
12
9
3
0
0
3
0
0
3
0
0
3
0
0
3
C
25
6
0
0
0
0
0
0
0
0
0
0
0
0
0
D
20
0
-
6
2
14
0
4
84
0
5
155
0
6
258
0
E
23
8
2
0
0
4
0
0
4
0
0
3
0
0
2
F
8
12
6
0
0
10
0
0
7
0
0
6
0
0
6
G
11
0
-
6
2
18
0
0
0
0
0
0
0
0
0
0
H
25
9
3
0
0
3
0
0
3
0
0
3
0
0
3
I
8
6
0
0
0
2
0
0
5
0
0
6
0
0
6
J
18
0
-
6
2
30
0
1
10
0
0
0
0
0
0
0
K
15
6
0
0
0
2
0
0
3
0
0
3
0
0
3
L
30
6
0
0
0
0
0
0
0
0
0
0
0
0
0
To
tal
6
62
6
132
6
193
6
258
Du
e date of
PSC
55
55
55
55
An
al
yz
in
g
PS
C
with
y
of
6
days
resu
lt
s
in
4
act
ivit
ie
s
wit
h
NTF,
nam
el
y
A,
D,
G
a
nd
J
act
ivit
y.
All
of
the
N
TF
va
lue
for
the
A,
D,
G
,
and
J
a
ct
ivit
y
is
-
6.
It
exp
la
ins
that
the
four
act
ivi
ti
es
will
exp
er
ie
nce
confli
ct
ing
obje
ct
ive
with
t
he
m
axi
m
u
m
co
m
pr
essing
is
6
days
as
a
t
otal
com
pr
essio
n
(
TC)
of
P
SC
wi
th
y
of
6
days
.
T
his
al
so
a
pp
li
es
to
ot
her
m
od
es
(m
od
e
1
to
m
od
e
4)
with
va
ryi
ng
x
n
,
as
show
n
i
n
T
able
3.
Mi
nim
u
m
conditi
on
of
th
e
add
it
io
nal
co
st
fo
r
P
SC
is
i
n
m
od
e
1
(Cn
=
62
)
as
a
n
optim
al
so
luti
on
with
com
pr
e
ssin
g
act
ivit
y
are
suc
cessf
ully
A
(
0
days),
D
(
2
days),
G
(
2
da
ys),
a
nd
J
(
2
da
ys).
T
he
po
i
nt
in
this
case
(
y
=
6
days),
act
ivit
ie
s
that
do
not
ha
ve
NT
F
will
no
t
occur
com
pr
es
si
ng
on
the
act
ivit
y,
as
x
n
in
Table
3.
It
is
cl
ear
that NT
F
has
a
correlat
io
n
wit
h
the
con
flic
ti
ng
obj
ect
ive
.
Using
the
sam
e
m
et
ho
d
as
in
Figure
3,
a
nal
ysi
s
of
NT
F
in
the
PSC
of
7
days
will
resu
l
t
in
the
NTF
values
va
ryi
ng
as
in
Ta
ble
4.
The
s
ol
ution
i
n
th
os
e
m
od
es
(w
e
only
sho
w
4
m
od
es)
shows
the
c
om
pressi
ng
act
ivit
y
(
x
n
)
fol
lowing
the
N
TF
patte
r
n
wit
h
the
value
of
x
n
are
un
der
a
range
of
the
a
bs
ol
ute
val
ue
of
NTF
(|NTF
|)
.
T
his
c
onditi
on
will
a
lso
ha
ve
t
he
sa
m
e
patte
rn
for
al
l
of
t
he
y
va
lues
(t
he
c
om
pr
essin
g
nu
m
ber
f
or
PSC)
in
this
stud
y.
We
co
nclud
e
that
the
va
lue
of
|NT
F|
sh
ows
the
m
axim
u
m
value
that
can
be
achie
ved
by
com
pr
essin
g
a
ct
ivit
y (
x
n
)
of a
ll
p
la
nn
e
d PSC
and all
m
od
es
pro
du
ce
d (both
optim
al
an
d fe
asi
ble so
l
utio
ns).
Table
4.
T
he
s
olu
ti
on
for
M
O
IN
L
P fo
r
P
SC
of 7 days
(Pro
je
ct
d
ue
d
at
e
54
d
ay
s)
Activ
ities
No
r
m
al
Du
ration
Initial
Total
Flo
at
NTF
Op
ti
m
al Solu
tio
n
Feasib
le Solu
tio
n
Mod
e 1
Mod
e 2
Mod
e 3
Mod
e 4
x
n
C
n
TF
n
x
n
C
n
TF
n
x
n
C
n
TF
n
x
n
C
n
TF
n
A
12
0
-
7
0
0
0
0
0
0
2
150
0
4
596
0
B
12
9
2
0
0
2
0
0
2
0
0
2
0
0
2
C
25
6
-
1
1
2
0
1
2
0
0
0
0
1
2
1
D
20
0
-
7
3
39
0
3
39
0
2
14
0
3
39
0
E
23
8
1
0
0
3
0
0
2
0
0
5
0
0
5
F
8
12
5
0
0
9
0
0
9
0
0
8
0
0
5
G
11
0
-
7
2
18
0
3
84
0
1
2
0
0
0
0
H
25
9
2
0
0
2
0
0
2
0
0
2
0
0
2
I
8
6
-
1
0
0
2
0
0
2
0
0
3
0
0
6
J
18
0
-
7
2
30
0
1
10
0
2
30
0
0
0
0
K
15
6
-
1
0
0
2
0
0
2
0
0
2
0
0
2
L
30
6
-
1
0
0
0
0
0
0
1
5
0
1
5
1
Total
8
89
8
135
8
201
9
642
Du
e date of
PSC
54
54
54
54
4.
DISCU
SSI
ON
4.1.
The i
mpo
r
t
ant r
ole of
Nega
t
ive To
ta
l
Flo
at
(
NTF
)
in
M
OP
This
pa
per
pr
e
sents
the
N
TF
ro
le
to
at
ta
in
the
co
nf
li
ct
in
g
obj
ect
ives
of
a
ct
ivit
y
in
or
de
r
to
achiev
e
the
PSC.
T
he
r
esult
of
the
c
onflic
ti
ng
obj
ec
ti
ve
in
MOI
N
LP
of
a
P
SC
as
sh
ow
n
in
Ta
ble
2
sho
ws
th
ere
is
a
correla
ti
on
of
NTF
val
ue
wit
h
the
m
a
xi
m
u
m
red
uctio
n
of
the
act
ivit
ie
s
f
or
bot
h
op
ti
m
a
l
and
feasible
s
olu
ti
on
of
MOI
NL
P
f
or
PSC
as
s
how
n
in
Ta
ble
3
a
nd
Ta
ble
4
.
As
an
il
lustr
at
ion
for
a
PSC
with
a
val
ue
of
y
of
7
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
5292
-
5303
5300
days
hav
i
ng
a
confli
ct
ing
act
i
vity
are
A,
C
,
D,
G,
I
,
J,
K
,
a
nd
L
with
NT
F
res
pecti
vely
a
re
-
7,
-
1,
-
7,
7,
-
1,
-
7,
-
1,
an
d
-
1
as
sh
ow
n
in
Table
4
.
I
n
Table
4
sh
ows
that
the
PSC
will
be
sat
isfie
d
by
co
m
pr
essing
a
num
be
r
com
bin
at
ion
of
so
m
e
or
al
l
the
pro
j
ect
act
ivit
ie
s
(A
,
C,
D,
G,
I,
J,
K,
an
d
L
act
ivit
ie
s),
w
it
h
a
value
of
ar
e
no
t
e
xceed
i
ng
it
s
NTF
val
ue.
It
is
a
novelty
of
this
r
esea
rch
where
NT
F
co
uld
be
an
autom
at
ic
m
axi
m
u
m
value
of
t
he
ac
ti
vity
du
rati
on
reducti
on,
al
th
ough
we
do
not
sp
eci
fy
a
m
axi
m
u
m
value
of
the
act
ivit
y
dur
at
ion
reducti
on
(
).
T
hu
s
,
t
his
st
ud
y
f
or
m
ulate
s
the
X
n
Var
ia
ble
(t
he
num
ber
of
act
ivit
y
durati
on
r
edu
ct
io
n)
has
a
constrai
nt
as for
m
ula (
21)
.
X
n
≤
Mi
nim
u
m
(
|
NTF
|
or
m
a
x
)
(21)
The
f
or
m
ula
(21)
sho
ws
us
th
at
will
not
be
achieve
d
if
the
NT
F
a
bsolute
le
ss
tha
n
.
It
i
s
i
m
po
rtant t
o u
nd
e
rstan
d
it
re
gardin
g h
ow
t
o de
velo
p
the
th
e m
axi
m
u
m
v
a
lue of the
acti
vi
ty
d
ur
at
io
n re
du
ct
io
n
(
max
)
.
T
he
N
TF
a
bs
ol
ut
c
ould
be
a
co
ntr
ol
f
or
ma
x
of
m
od
el
m
ulti
obj
ect
ive.
T
he
NT
F
c
ould
be
an
autom
at
ic
of
ma
x
in
M
OP
m
od
el
.
This
di
f
fers
f
r
om
the
Dec
kro
[
19
]
a
nd
the
s
a
m
e
oth
er
res
e
arch
w
hich
do
no
t
c
oncer
n
to
ward
the
NT
F
as
a
pa
ram
eter
co
ns
i
der
e
d.
The
neg
at
i
ve
float
co
nce
pt
(NF)
has
al
s
o
bee
n
introd
uced
by
Li
m
’s
pap
er
w
hich
sta
te
s
tha
t
NF
is
the
am
ou
nt
of
dura
ti
on
of
act
ivit
y
that
ca
n
be
re
du
ce
d
without
a
ff
ect
i
ng
the
com
pleti
on
ti
m
e
of
th
e
pro
j
ect
[16]
.
H
ow
e
ve
r,
t
he
Lim
’s
pap
e
r
[
16
]
does
not
e
xp
la
in
about
the
ty
pe
of
float
use
d
in
his
resear
ch.
Me
an
w
hile,
Su
’
s
pa
per
introd
uce
ne
ga
ti
ve
of
inter
f
eren
ce
float
[28]
.
This
is
a
con
trast
to
the
NT
F
(n
e
gative
total
flo
at
)
con
ce
pt
pre
sented
in
this
pa
per,
w
her
e
th
e
NTF
m
us
t
be
per
f
orm
ed
to
cond
uc
t
the
Pr
oject
S
cheduli
ng
Co
m
pr
ession
(
PS
C)
by
cha
ng
i
ng
the
Lat
est
Finish
in
te
rm
inal
node
of
netw
ork
diagr
am
with
val
ue
(
due
date
of
PSC)
,
as
s
ho
wn
i
n
Fig
ure
3.
I
n
this
pa
pe
r,
we
al
so
intr
oduc
e
the im
po
rtant
r
ole in c
onduct
i
ng m
ulti
o
bj
ect
ive to sol
ve
P
S
C pro
blem
.
4.2.
Implem
entin
g the
basic
MO
P Model
as
M
OIN
LP
The
m
od
el
dev
el
oped
in
t
his
stud
y
ref
e
rs
to
the
Deckr
o
m
od
el
[19]
as
a
basic
m
od
el
of
MOP
.
Howe
ver,
this
pro
po
se
d
M
OINLP
m
od
el
ha
s
ad
diti
on
al
c
onstrai
nts
,
as
show
n
i
n
F
orm
ul
a
(
20),
w
hich
has
a
sign
ific
a
nt
ef
fe
ct
on
s
olv
i
ng
m
or
e
com
plex
PSC
pro
blem
s
(it
can
so
l
ve
pro
blem
s
with
hig
h
-
le
vel
va
ria
bles).
This
co
ndit
ion
ind
ic
a
te
s
that
the
f
or
m
ula
(20),
as
N
TF
c
onstrai
nt,
is
si
gnific
ant
to
im
pr
ove
the
basic
m
od
el
.
This is in
li
ne
with r
esea
rch
s
ta
te
s that the f
l
oat
-
path
the
or
y
to
so
lve the tim
e
-
cost trad
e
-
off
pro
blem
, al
though
the
resea
rc
h
use
s
ne
gative
of
inter
fer
e
nce
float
[28
]
.
T
he
pr
i
nciple
of
de
velo
ping
this
MOI
NLP
m
od
el
is
creati
ng
a
n
un
balance
d
c
ondi
ti
on
of
T
otal
F
loat
(which
ne
eds
to
be
sat
is
fied)
by
m
anipu
la
ti
ng
t
he
ne
ga
ti
ve
value
of
the
to
ta
l
float.
W
e
s
ee
an
opport
unit
y
to
m
anipu
la
te
the
Total
Float
by
rep
la
ci
ng
the
La
te
st
Finis
h
(LF)
value
on
t
he
te
rm
inal
no
de
of
the
net
w
ork
diag
ram
with
the
val
ue
(
du
e
date
of
pro
j
ect
c
om
pr
es
sion),
as
show
n
in
Fi
gure
3.
C
ho
os
ing
t
he
rig
ht
ac
ti
vity
,
we
us
e
t
he
S
olv
e
r
A
ppli
cat
ion
as
an
op
ti
m
iz
ation
too
l
t
o
el
i
m
inate
NTF
by
re
du
ci
ng
t
he
durati
on
of
t
he
act
ivit
y.
Ba
sed
on
t
his
co
ndit
ion
,
we
def
i
ne
the
NTF
as
a
basis
to
reduce
the
a
ct
ivit
y
du
rati
on
to
achie
ve
due
date
of
the
pro
j
ect
sche
dule
co
m
pr
essio
n.
Her
e
we
argu
e
that
pro
blem
so
lving
of
MO
P
f
or
PSC
case
by
i
nvol
ving
the
NT
F
is
the
key
point
to
im
pr
ov
e
a
MOP
as
M
O
IN
L
P
m
od
el
.
The
c
oncl
us
i
on
of
thi
s
stud
y
ha
ve
s
how
n
that
the
cl
assic
m
od
el
sti
ll
has
an
im
portant
ro
le
to
so
lve
com
plex
pro
bl
e
m
s w
hic
h
it
c
an
c
om
pen
sat
e f
or
m
or
e s
ophi
sti
cat
ed
MOP t
echn
i
qu
e
s.
Our
pres
um
pti
on,
this
MO
I
NLP
m
o
del
w
il
l
giv
e
the
op
tim
iz
at
ion
pro
cess
faster
th
a
n
the
ea
rlie
r
m
od
el
to
achi
eve
a
n
opti
m
a
l
so
luti
on.
T
hi
s
is
a
f
uture
r
esearch
to
ass
ess
the
ti
m
e
l
eng
t
h
of
t
he
it
rati
on
op
ti
m
iz
ation
proces
s,
f
or
bot
h
of
discrete
m
od
el
or
e
ve
n
dist
rib
ution
durati
on
m
od
el
[
2
9
]
,
to
reac
h
th
e
op
ti
m
al
so
luti
on
of
a
PSC
pro
blem
.
Ba
sed
on
t
hese
res
ults
ind
ic
at
e
that
th
ere
is
a
relat
io
ns
hi
p
betwee
n
NTF
,
as a co
ns
trai
nt o
n
th
e m
od
el
d
evelo
ped
i
n
this pap
e
r,
with both
the
op
ti
m
al
so
luti
on and fea
sible solutio
n.
T
he
so
luti
on
in
dica
te
s
the
im
po
rtance
to
us
e
NT
F
on
the
m
od
el
de
velo
ped
,
es
pecial
ly
in
this
stud
y.
S
om
e
of
them
,
we
ca
n descri
be
h
e
re ar
e:
1.
Each acti
vity
with
NTF val
ue
is the c
onflic
ti
ng
obj
ect
i
ve o
f
MO
INLP
for PSC
case.
2.
The
obso
le
te
va
lue
of
NTF
(|
NTF
|)
is
m
axim
u
m
value
tha
t
can
be
ac
hiev
ed
by
the
com
pr
essi
ng
act
ivit
y
(in
case
the
re i
s no
m
axi
m
u
m
lim
i
t specifie
d f
or
t
he
c
om
pr
essing ac
ti
vity
, se
e
X
n
max
in Tab
le
1).
3.
In
t
he
case
of
PSC,
the
act
ivi
ty
du
rati
on
reducti
on
can
not
decr
ease
but
it
will
increase
or
co
ns
ta
nt
i
n
li
ne
with incre
asi
ng
o
f
P
SC. T
he
exp
la
nation
is
, th
e v
al
ue
of x
n
(
durati
on
re
du
ct
ion
of
acti
vity
)
caused
by of y
(v
al
ue
of the P
SC) is
m
or
e o
r e
qu
al
the
n
x
n
c
ause
d
by y
-
1
(i
t cou
ld
be
com
par
e
d
on
T
a
ble
3
an
d
Ta
ble
4
).
This
PSC
case
has
a
diff
e
rence
fr
om
the
product/
m
at
eria
l
m
ix
case
wh
er
e
it
var
ia
ble
cou
l
d
be
up
an
d
dow
n
in
li
ne w
it
h
the in
crease
it
targ
et
, like
the case
in
[
30
]
.
5.
CONCL
US
I
O
N
The
NT
F
co
uld
be
us
e
d
to
identify
the
co
nf
li
ct
ing
obj
ec
ti
ve
in
the
PSC
case
s
.
The
com
bin
at
ion
of
act
ivit
ie
s
hav
i
ng
N
TF
will
com
pete
for
one
to
eac
h
oth
e
r
to
achie
ve
t
he
m
ini
m
u
m
a
dd
it
io
nal
c
os
t
of
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Ne
ga
ti
ve
T
ota
l
Flo
at to Im
pro
ve a Mult
i
-
obje
ct
iv
e In
te
ge
r N
on
-
l
ine
ar
Pr
og
ra
m
ming
…
(
Fachrurr
az
i
)
5301
pro
j
ect
com
pr
ession.
It
sho
ul
d
be
co
nduct
ed
with
out
ign
ori
ng
the
pr
i
nciples
of
CP
M
and
feasi
bl
e
of
al
l
const
raints
of
M
OINLP.
NT
F
cou
l
d
be
m
a
nipulat
ed
by
re
placi
ng
t
he
Lat
est
Finish
(LF)
value
on
the
te
rm
inal
node
of
the
ne
twork
dia
gr
am
with
t
he
val
ue
(
du
e
date
of
pro
j
ect
com
pr
e
ssion)
t
o
know
w
hich
one
of
the
durati
on
of
act
ivit
y
are
red
uc
ed.
I
n
this
pa
pe
r,
we
inc
reas
e
the
per
f
or
m
an
ce
of
the
MOP
basic
m
od
el
to
so
lve
the
PSC
com
plex
prob
le
m
by
introd
uce
t
he
MOI
NL
P
c
oncept.
I
nvol
vin
g
the
NT
F
i
n
M
OINLP
ha
ve
a
n
i
m
po
rtant
r
ole
and a
key point
to
im
pr
ove a
MOI
NLP
.
Neg
at
ive
T
otal
Float
(N
TF
)
can
be
m
anipu
la
te
d
by
rep
la
ci
ng
t
he
value
of
the
Lat
est
Fini
sh
(LF)
in
a
te
rm
inal
node
of
p
r
oj
ect
network
diag
ram
as
the
du
e
date
of
Pro
j
ect
sc
hedulin
g
C
ompressi
on
(
).
For
t
he
pur
po
se
of
co
nductin
g
the
PS
C
,
t
he
certai
nly
has
a
sm
al
le
r
val
ue
tha
n
the
Earli
er
Finis
h
of
N
or
m
al
Proje
ct
(EF).
T
hese
diff
e
ren
ces
ca
n
gen
e
rate
NT
F
on
s
om
e
or
al
l
of
the
act
ivit
y.
NTF
recove
ry
can
be
do
ne
by
reducin
g
the
a
m
ou
nt
of
dur
at
ion
of
the
act
iv
it
y
on
the
corr
esp
onding
pat
h.
This
is
a
log
ic
t
hin
g
in
P
SC.
Ba
se
on
the
c
onditi
on,
we
de
fine
that
the
Neg
a
ti
ve
Total
Float
(N
TF)
is
th
e
m
axi
m
u
m
value
for
reduc
ing
the
act
ivit
y du
rati
on to
achie
ve d
ue
d
at
e
of the
Proj
ect
sche
duli
ng C
om
pr
ession
ACKN
OWLE
DGME
NTS
The
Au
t
hors
w
ou
l
d
li
ke
t
o
a
ppreciat
e
var
i
o
us
pa
rtie
s
invol
ved
in
t
his
rese
arch,
es
pecial
ly
to
Direct
or
of
Re
instat
e, Staf
f of
CEM
Labo
rator
y
-
Ci
vil E
ng
i
neer
i
ng of
Syi
ah
K
uala
U
niv
e
rsity
.
REFERE
NCE
S
[1]
Fachr
urra
z
i,
S
.
Hus
in,
Munirwans
y
ah
,
and
Hus
ai
ni,
“
The
Subcontra
c
tor
Sele
c
ti
o
n
Prac
ti
c
e
using
ANN
-
Multi
lay
er
,
”
Int.
J. Tec
hno
l.
,
vol.
8
,
no
.
4
,
p
.
7
61,
Jul.
2017
.
[2]
D.
A.
No
y
ce
an
d
A.
S.
Hanna
,
“
Planne
d
sche
du
le
compress
ion
conc
ep
t
fi
le
for
el
e
ct
ri
ca
l
cont
r
a
ct
ors,”
J.
Constr
.
Eng.
Manag
.
,
vo
l.
123
,
no
.
2
,
pp
.
189
–
197,
Jun.
1
997.
[3]
R.
M.
W
.
Hor
ne
r
and
B
.
T.
Talh
ouni,
“
Eff
ec
ts
of
Acc
e
le
r
at
ed
W
orking,
De
lay
s
a
nd
Distrupti
on
[i
e
Disruption]
on
La
bour
Produc
tivit
y
,
”
Cha
rte
r
ed Instit
ute of
Bu
ilding,
1996
.
[4]
C.
-
K.
Chang,
A.
S.
Hanna
,
J.
A.
La
ckn
e
y
,
and
K.
T.
Sulli
van
,
“
Q
uant
if
y
ing
th
e
i
m
pac
t
of
sche
dule
compress
ion
on
la
bor
produc
t
ivi
t
y
for
m
ec
hanica
l
and
shee
t
m
et
a
l
cont
ra
ct
or,”
J.
Constr
.
Eng.
Manag
.
,
vol.
133
,
no.
4,
pp.
287
–
296,
Apr.
2007.
[5]
P.
Bal
le
st
ero
s
-
Pére
z
,
“
Modell
in
g
the
boundar
ies
of
proje
ct
fast
-
tra
ck
ing,”
Aut
o
m.
Constr
.
,
vol.
84,
pp.
231
–
241,
Dec
.
2017.
[6]
A.
Kanda
and
U.
R.
K.
Rao
,
“
A
net
work
flow
proc
edur
e
for
pro
je
c
t
cra
shing
wit
h
pena
lty
nod
es,
”
Eur.
J
.
Oper
.
Res.
,
vol. 16, no. 2, pp. 174
–
182,
Ma
y
1984.
[7]
N.
George
s,
N.
Sem
aa
n,
and
J.
Riz
k,
“
Crash :
an
Autom
at
ed
Tool
for
Schedul
e
Crashing,
”
Int
.
J.
Sci.
Env
iron
.
Technol
.
,
vo
l. 3,
no.
2
,
pp
.
374
–
3
94,
2014
.
[8]
A.
El
Yam
ami,
K.
Mansouri,
M.
Qbadou
,
and
E
.
H.
Ill
ousam
en
,
“
Multi
-
objecti
v
e
IT
project
select
ion
m
ode
l
fo
r
improving
SM
E s
tra
te
g
y
depl
o
ym
ent
,
”
Int. J. El
ec
tr.
Comput.
E
ng
.
,
vol. 81, no.
2,
pp
.
1102
–
111
1,
Apr.
2018.
[9]
Fachr
urra
z
i,
S.
Hus
in,
N.
Malaha
y
ati,
and
Irz
aidi,
“
Ide
ntif
y
ing
ina
c
cur
a
c
y
of
MS
Projec
t
using
s
y
stem
anal
y
s
is,
”
IOP
Conf. Se
r.
Mate
r.
S
ci. Eng
.
,
vol. 352, no. 1,
p.
012036
,
Ma
y
2
018.
[10]
F.
Fachr
urra
zi,
“
The
End
Us
er
Requi
rement
for
Projec
t
Mana
g
e
m
ent
Software
Acc
ura
c
y
,
”
Int
.
J.
Elec
tr
.
Comput.
Eng.
,
vol
.
8
,
no
.
2,
pp
.
1112
–
112
1,
Apr.
2018.
[11]
A.
Faruq,
M.
F.
Nor
Shah,
and
S
.
S.
Abdulla
h,
“
Multi
-
object
ive
Optimiza
ti
o
n
of
PID
Control
le
r
using
Pare
to
-
base
d
Surrogate
Model
ing
Algorit
hm
for
MIM
O
Eva
pora
tor
S
y
stem,
”
Int.
J.
E
le
c
tr.
Comput.
Eng.
,
vol
.
8,
no.
1,
p.
556,
Feb.
2018
[12]
J.
Mote,
D.
L
.
Ol
son,
and
M.
A.
Venka
t
ara
m
a
nan,
“
A
compar
at
iv
e
m
ult
iobjec
ti
ve
progr
amm
ing
stud
y
,
”
Ma
th
.
Comput.
Mode
l.
,
vol. 10, no. 10,
pp.
719
–
729
,
Ja
n.
1988
.
[13]
R.
S.
Jürgen
Branke
,
Kal
y
anmo
y
Deb,
Kaisa
Miet
ti
nen
,
Multi
obj
e
ct
iv
e
Optimiza
t
i
on:
Inte
ra
ct
iv
e
a
nd
Evol
uti
on
a
r
y
Approac
hes
.
Springer
-
Ve
rlag
,
20
08.
[14]
D.
E.
Be
ll
,
R
.
L
.
Kee
ne
y
,
and
H.
Rai
ffa
,
“
Confli
c
ti
ng
objecti
v
es
i
n
dec
isions,
”
Confl
icting
Obje
ct.
Dec
is.
,
pp
.
298
–
322,
1977
[15]
D.
Dubois,
H.
F
arg
ie
r
,
and
V.
Galva
gnon
,
“
On
la
te
st
star
ti
ng
ti
m
es
and
floa
ts
in
ac
t
ivi
t
y
ne
tworks
with
il
l
-
known
dura
ti
ons,
”
Eur.
J.
Op
er.
Re
s.
,
vo
l.
147
,
no
.
2
,
pp
.
266
–
280,
Jun.
2
003.
[16]
A.
Li
m
,
H.
Ma
,
B.
Rodrigue
s,
S.
Te
ck
Ta
n
,
and
F.
Xiao,
“
New
c
once
pts
for
ac
t
iv
ity
flo
at
in
resou
rce
-
constr
ai
n
ed
proje
c
t
m
ana
g
e
m
ent
,
”
Comput.
Oper
.
R
es.
,
vol
.
38,
no
.
6
,
pp
.
91
7
–
930,
Jun.
201
1.
[17]
K.
G. Loc
k
y
e
r a
nd
J.
Gordon,
Pr
oje
c
t
m
ana
g
ement
and
proj
ec
t
ne
twork
techniqu
e
s
,
vol. 7. Pea
rso
n
Educat
ion
,
200
5
[18]
B.
H.
R
ei
ch
,
S.
Y.
W
ee
,
and
W
.
Siew
Yong,
“
Sear
ch
ing
for
Kn
owledge
in
the
PM
BOK
Guide,
”
Proj
.
Manag.
J.
,
vol.
37
,
no
.
2
,
pp
.
11
–
26
,
2006
.
[19]
R.
F.
Dec
kro
an
d
J.
E.
Heb
ert,
“
A
m
ult
ipl
e
obj
ective
progr
amm
ing
fra
m
ework
for
tra
deof
fs
in
p
roj
ec
t
sch
edul
ing
,
”
Eng.
Costs
Prod.
E
con.
,
vo
l. 18,
no.
3
,
pp
.
255
–
2
64,
Jan
.
1990
.
[20]
A.
Franc
is
and
S.
Morin
-
Pepi
n,
“
The
Conc
ep
t
of
Float
C
alc
ula
ti
on
Based
o
n
the
Sit
e
Occ
upat
ion
using
t
he
Chronogra
phical L
ogi
c,”
in
Proc
edi
a
Engi
ne
erin
g
,
vol
.
196
,
pp
.
6
90
–
697,
2017
.
[21]
J.
M.
Hend
erson
and
R
.
E. Qua
nt
,
Micro
ec
onom
i
c
Th
eor
y
:
A Mat
hemati
c
al Approac
h
.
McGraw
-
Hi
ll
,
1980
[22]
A.
Ishiz
aka
and P
.
Nem
er
y
,
Mul
t
i
-
Crit
e
ria Decisi
on
Anal
y
sis: Me
thods a
nd
Softw
are
.
John W
ile
y
&
Sons
,
2013
.
Evaluation Warning : The document was created with Spire.PDF for Python.