0% found this document useful (0 votes)
22 views21 pages

Chapter 4 Simplex Method

The document discusses the simplex method for solving linear programming problems in standard form. It explains how to find an initial basic feasible solution and construct the initial simplex tableau. It also provides some properties and remarks about the simplex tableau when representing a linear program.

Uploaded by

Chau Felix
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
22 views21 pages

Chapter 4 Simplex Method

The document discusses the simplex method for solving linear programming problems in standard form. It explains how to find an initial basic feasible solution and construct the initial simplex tableau. It also provides some properties and remarks about the simplex tableau when representing a linear program.

Uploaded by

Chau Felix
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 21

Chapter 4 Simplex Method

Enumeration Method :

X1 : convert a general LP Dito standard form .

Stg : Find all BS .


and eliminate those which are not feasible .

¥3 : Find those that are optimal among all BFS .

Remark :
Computation can be expensive ( ( Tn ) B. S )

Goal :
Develop systematic way to
study the BFS to optimize LPP .

C Simplex Method ) .
Adjacent BFS '
. Two BFS of a LP in standard form are said to be

Adjani if they have all but one basic variables in common .

EG

12K
it as E 4 2K , tha t 23 = 4

|
.

N, t 222 Ef ki t 222 +24 = 6

K 2230
. . Ki ,
Nz , Jez , 24 30

Ii
T
( 4 6)

{
= O . O .
.

T
Iii: = (O . 3 ,
I ,
O >

I =
( 43.93 .
o . 05

Goal :
Simplex Method find opt Sol

by moving from one BFS to an adjacent


BFS which increases obj fun value in

each
step .
Simplex Method for LP in feasible canonical form

Max 2- = E'I E. g . Max 2- = 32 , t 422

St .
AZ EI ( 835 ) St . 2e , t 222 E5

£ 38 K, -
sea E 4

K . .
22 30
R2
^
The LP can be converted to :

MAX Z = 32 , t 4k

sit

:*::* :.÷ : :
.
'

K . .
K2 ,
K3, Na 70

Find initial BFS ( let all slack variables be B. v )


Step 1 : an .

initial tableau

I
:

K, Kz Kz 24 Z to

in
.

E Je
'
-

3 -4 O O l O F Z -
= o

initial I = ( O, O, 5 . 4)
T
with Z = 0 .

Remark : ( properties of the tableau )

4) Row lls
captures eg't Cl ) i Row 12 )
captures eqt 12 )

( NW O )
(2) Last the
obj fun
tow
captures
(3) B. v in with coefficient
exactly row 1
.

appears one .

(4) The Col associated to the B. v . are all 0


except for the t1 in

the row labeled by the B - V .

of B. v be obtained from the Col


(5) Values . can
rightmost .

(5) Col 2- can be deleted .


In general ,
the tableau looks like :

:/ :/
H,Hz---HnKht,NhtL__.2en+mZ#
Nutt Ali AIL - - -
Ain t O - - -

O O b,

Katz 021 922 . . .

Azn O l - - i
O O .bz

: : : : : i :

NntmAmiAmz---Amn00---/0b#
-
C,
-

Cz - - . -
Ch O O - -
-
O 1 O
\

or . more
compactly

Is|A2.#
- -

#
Z Ks
- →

-T - T
-

C O 1 O
Step check
2 :
optimality

I 1
recall : Z =
32 , -1422

I:t: :÷÷÷÷÷
sea

In general . 2- = -2 dj Kj t E o .

Kj
N . B. V .
B. V .

Where dj =
negatives of the entries in row o .

If some
dj are positive . Z can be T by T the value
of N -
B. v .

Kj from
O .

Hence ,

t.ie?ni:ethe..:::an.n::.::.negameen.
Step 3 : choose one entering variable
N . B. V .

T
In general

kBrµµ3r
,

Kk
choose N B. whose coefficient
-

a .
u .
see
÷

in row O is negative
.
.

÷
In our
example .

( eutennguaiaba.I.fi?.a7g!o:o./5g)
choose a. as the
Step 4 : choose departing I leaving variable

After choosing x , as new B. u .


.
we need to choose either as

or 24 to become nonbasic
to

-1-1010/43
1121005
Z
.

K , K2 2.3 24
-

NOW Nz 5 ki 222
Kay
/
= - -

24=4 -
se , -
sea .

-4 O O l O

keep NEO Kz 5 K,

{
= -

me 4- a .

we choose the largest 2e , such that 2330 . Ka 30 .

Then ,
se , = min { % ,
% Y =
4 .

When N, =
4 , Ks = I 24=0 .
.
-

. 24 becomes nonbasic .

(
departing variable )

N . B. V .

In general ,
T

I
once Nk is chosen as
entering "
:

r '

Br
variable ,
we choose upon as the

he
departing variable
if
Br
% ,
is smallest
positive .
Remark .

If UK is chosen as the entering variable .

7Bi/9ikE÷/p
HK RBI
-

Vk O

ith row Yik Kk t HB ;


pi
: =

KB; =
pi Yi KKK
-

{
if yik
-

- O , ignore it .

② it Yik > O
'
Bi -

Yik Nk 30 Kk
-

-
min /yF÷ I of
yie >

=
BI
Yrk
③ if Yik co .
ignore it .

when Kk = BYyrk ,
KBR =
Br -

brk .
( Rhyne ) =
O .

( departing variable )

.:÷:÷%÷÷i::÷:itm +
a

you :¥÷n÷÷÷÷i÷¥I
.
"
Form
Step 5 :
a new BFS
b
( a new tableau )

to

-140,0€
:/
Initial tableau :
K, Ks Ks 24 Z

'2i
-

I
(* )

3 -4 O O l 0

In the new BFS :

① B. V .
i
. 23 .
K , .
i N . B. V. : K2 ,
Ka .

② K, -_ 4 ,
Ks = I . 22=24--0 2- = 32 , -1422=12 .

To preserve the properties of the tableau ,


we want .

to

←4
K, Kz Kz 24 Z

c** >

O O l 12

Note : we can convert CH ) into C** ) by Gaussian elimination .

be (
pivot operation ! ,


§#ZI
,,o,o#
,

-
I

3
2

-4
I

O
O

O
O

l
5

Remark .

to

#
K, Ks Ks 24 Z
① Col 2- can be deleted .

② Move from 10.05 tan

K#
adjacent BFS with ZT .

My ,

Bts EEC " "* ⇐ "

a,
K
,
Step 6 : Continue the process until opt Sol is reached

.÷÷÷÷÷t
Since vs = -7 co ,
Jez is not
optimal . let 2e , be entering variable .

's
/
32Gt Nz
{
= I 23 = I 3N 2 DO K E

{
-

Kz toe , k at sea 223 -4


=
4 ignores
-

, = so c

when 22=43 ,
23=0 ,
K, = 13/3
(
departing variable )

or
equivalently by ,
min ratio test :

min f Bily I -

, z
yiz >o 's = I 43 .
-
I =
Ys .
=
Bfg ,,

KB , =
Nz is the departing variable .

Do pivot operation at Yu =3 .
we have a new tableau :

b-

0143-4343%2,1104343/13130
-13 : K, ka Ks Ka

O 73 43 43/3
Now all coefficients in row o are
positive Hence , we can not T Z .

I# C ' 43 ,
43 ,
O ,
05 with 2- = 43/3 .

opt
Sol .

R2
^
The simplex Method . I LP in feasible canonical form )

Construct initial tableau and get initial BFS


Steph .
an an .

NIs|A2#
- -

-T - T
c o o j
-

" -

"

check optimality
steps .

j Yoo

if
{ if
all 30 the current BFS is optimal ; stop
rj ,
.

I
rj
so ,
go to
step 3 .

Steps choose the entering variable .

let N B
- - V . Kk with rkco enter the basis .

choose the
departing variable using Min
step "
-
ratio test .

Byy ..
= mm
{ Bibi I .
six > o ) .

KBR is the
departing variable .

yrk and establish tableau


Stg .
pivot at a new .

K j


Yrj

:Il:÷÷÷
r

Return to 2
step .
Lecture 4: The Simplex Method 4-8

Example 4.1.1. Consider the LP problem:

Max z = 3x1 + x2 + 3x3


Subject to 2x1 + x2 + x3  2
x1 + 2x2 + 3x3  5
2x1 + 2x2 + x3  6
x 1 , x2 , x 3 0

By adding slack variables x4 , x5 and x6 , we have the following initial tableau.


Tableau 1:

x1 x2 x3 x4 x5 x6 b Ratio
2
x4 2 1⇤ 1 1 0 0 2 1 = 2⇤
5
x5 1 2 3 0 1 0 5 2 = 2.5
6
x6 2 2 1 0 0 1 6 2 =3
z 3 1 3 0 0 0 0

Initial tableau, current BFS is x = [0, 0, 0, 2, 5, 6]T and z = 0.

We choose x2 as the entering variable to illustrate that any nonbasic variable with neg-
ative coefficient can be chosen as entering variable. The smallest ratio is given by x4 row.
Thus x4 is the leaving variable.

Tableau 2:

x1 x2 x3 x4 x5 x6 b Ratio
2
x2 2 1 1 1 0 0 2 1 =2
1
x5 3 0 1⇤ 2 1 0 1 1 = 1⇤
x6 2 0 1 2 0 1 2
z 1 0 2 1 0 0 2

Current BFS is x = [0, 2, 0, 0, 1, 2]T and z = 2.


Lecture 4: The Simplex Method 4-9

Tableau 3:

x1 x2 x3 x4 x5 x6 b Ratio
1
x2 5⇤ 1 0 3 1 0 1 5
x3 3 0 1 2 1 0 1
x6 5 0 0 4 1 1 3
z 7 0 0 3 2 0 4

Current BFS is x = [0, 1, 1, 0, 0, 3]T and z = 4.

Tableau 4:

x1 x2 x3 x4 x5 x6 b
1 3 1 1
x1 1 5 0 5 5 0 5
3 1 2 8
x3 0 5 1 5 5 0 5
x6 0 1 0 1 0 1 4
7 6 3 27
z 0 5 0 5 5 0 5

Optimal tableau, optimal BFS is x = [1/5, 0, 8/5, 0, 0, 4]T and z ⇤ = 27/5.

We note that the extreme point sequence that the simplex method passes through are
{x4 , x5 , x6 } ! {x2 , x5 , x6 } ! {x2 , x3 , x6 } ! {x1 , x3 , x6 }.
Question ?
:
Optimality condition : all
rj > o

A.tgebraofs.MN .

A LP in standard form :

Max 2- = E' I
sit AI -_ I 1538 )
at )
.

I 38

A- = ( B N ) I =
( III )
AZ -

- I BIB t N Jew =3

Z = E' I Z -

IIIB -
Eisen =
o

BN#
-

KB2wt#

- T -T

CB 0
-
-
car

During the computation ,


the tableau has the
property that cols

under Irs form an


identity matrix . and ri -0 t B. v .
Hi . .

Im*#
→ -

KB NN Khs
#

Iis
- T
O * Yoo
\
During the computation ,
the tableau has the
property that cols

under Irs form an


identity matrix . and ri -0 t B. v .
Hi . .

- -

① ③

for ① : BIB + NEW = I

B- BI,
'
t B- N
'

Jew =
B-
'
I
Im Est B- N
'
Jew = B-
'

ImB-'N|p
→ -

NB Kw Khs
#

Ks 5
- T
O * Yoo
\

Also , Jeb = B- ' I -


B
-
'
N Tew H)

for ② : Z -
EEE ,
- IT In = o

Z -
Est ( B- I'
-
B- N
'
Ind -
Eisen =o

z + C EI B- N '
-
CIT ) Ew = IT B- I '

general tableau :

- -

KB NN Nhs

"
¥⇐..¥ h÷Ei "
tow

row
' -

O
m

Yoo
general form :

- -

:÷:. E÷i
KB NN Nhs

¥⇐.:i÷t:÷ or

since for Jeb .

Im = B- B
'

.
(Im B- N
'
) =
B- ( B
'
N) =
B- A
'
.

TBT B-
'

E' =
'
B -
Est ( O II B- N '
-
Ent )

Est B- B E5 ET
"

=L '
-

.
Is B- N
'
-

= £5 B- '
CB N ) -

C Est INT .
s

E5 B- E
' '
'
=
A -

K rhs

B-in-E.IQ
Est B- A E '
-
'

① let
D= ( Bijan ) =
B-
'
I Est B- I '
= Est p .

of Tj Tej
'
② each Col B- A
'
: = B-

③ each entry in low O :

rj
= Est B- Tej ' -

g-
=
CITYJ -

g- =
Zj
-

Cj .

there
Zj = IoTB- AJ '
=
E5yj
if Kj is basic,

rj = IT B-
'

AJ -

g-
=
Tsing -

g-
=
G-
-

Cj
-

- o .
F.G. I

,,,o
.
K . 22 Us 24 ths

23 I 2 I O S

z,

-3 -4 O O O

=/ ! ? ?) b-
=/ ! ) E-
'
A- ( 3. 4. o .
cost .

For Ti : B ,
=
Iz .
B- A
'
= A B-
'
5=5 .

CTB -
'
A - E' = GTA -
E - co ,
o > A -
E' = -
ET
CFB I - '
= Trib =
co , o > (I) =
o .

,→#
TI : N , 22 23 ka ths

23 O 3 I -
l l

z,

O -7 O 3 12

'

( i)
'

i) (
-

! )
'

(
-

Betas .at ,
-

Be
-

=
-_
o o o

to ill : : :O ) lo ? : ;)
-

B-' a- -
-

. .

I
( ill ? )
'

II )
-

B-' -

- -

( 7 ? I)
" -

EBTB -
'
A - E' =
co , 3 ) , o
-

CB .
4 .
o ,
o )

=
( 3 ,
-3 ,
O ,
3) -
( 3,4 ,
O ,
O) =
( O, -7 ,
O , 3 )

CFB I
(L )
'
(
-

= O, 3 > = 12 .
recall :
optimality condition for Max problem .

if opt Sol
{
all
rj 30 , .

if I
rj co . Not optimal .

proposition I :

thenthecuorentBTrsn.isaukqeopt.gs
For Max if for all N B.
a
problem rj =
Ej Cj 70 v
-
.
.
,

PI ①
.
t feasible Sol I .
let D= I -
I .

Then AT = Ac E - I > = J

(B
( Ffg ) BWI -1 NWT =3
ie .
N ) =

Wiz = -
B
-
'
N WI where wT= In -
Jew =
In 35
ET
( WI; ) Eius Tiwa
'
=
cos Ii's = +

= -
CFB -
'
N WIT E' win = -
C Ist B- w '
-

CF ) WI EO .

E'I = Eiwta > = E' hit E' I ⇐ E'I .

I is an
opt Sol .

② Assume I is also an
opt Sol . then E' I = Eta .

E' T -

-
o To =3 In =
Jew =3 .

Also TB = -

B- ' NWT =3 IB =
IB
2E = I

opt Sol Je is
unique .

#
Proposition 2 .

÷::a::::%:t÷÷¥i::÷÷:
PI .
( similar as pf of proposition 1 ) ( Exercise ! )

The simplex Method .


( Max problem )

Construct initial tableau and get initial BFS


Steph .
an an .

check optimality
Steps .

if
{ if
all 30 the current BFS is optimal ; stop
rj ,
.

I
rj
so ,
go to
step 3 .

Steps choose the entering variable .

let N B - - V . Kk with rkco enter the basis .

choose the
departing variable using Min
step "
-
ratio test .

Byy ..
= mm
{ Bits ) .
six > o ) .

KBR is the
departing variable .

yrk and establish tableau


Stg .
pivot at a new .

K j


Yrj

:l:÷÷÷
r

is :
Return to 2
step .
Question :

In Simplex method ,
we need an initial BFS to Start computation .

If the LP is not in feasible canonical form . how should we


get an initial BFS ?

E.G .
Max Z =
Hit Nz initial tableau :

Z/
Sit .
2N , -1 Nz -

Nz =4 K , Nz Hs rhs

" '
+ Z" =b

Ni 22 30
I 2 O 6
. . 23 .

I -
I O O

( No initial BFS )

It can be solved by M -
method and Two -

phase method .

You might also like