Lecture 2 Simplex methods

basic idea of the is to confine the

search to corner points of the feasible region(

of which
The simplex method
p g (
)in a most intelligent w
f
there are only finitely many ay
keyfor the istomakecomputers
.
The simplex method
i l l i
key for the is to make computers
see corner points;
h b d l h
The simplex method
f simplex algori the can be used to solve an
the must be converted into a problem where all the
thm ,

Before LP
LP
constraints are
maximization
and
objectivefunctionis minimizati
all variables are nonnegati
or
equ
on
ati ve ons .
. The maximization objective function is minimizati or
in this f
on
standard orm is said form to be in a .
.

The
An LP
Standard Fo Convert the rm to the LP LP
not in standard form is The LP
z 4x 3x Maximize = +
in is standard form The LP
z 4x 3x

Maximize = +
1 2
z 4x 3x Maximize
Subject to
= +
1 2
z 4x 3x Ma ximize
Subject to
= +
.
.
1 2
1 2
x x 40
2x

x 60
s
s
+
+
.
.
1
2
1 2
1 2
x x 40
2x x 60
s
s
=






=


+ +
+ +
,
1 2
1 2
x x >
s mmar if aconstraint of an isa constrai
, , ,
In
2
1 2
1 2
1 2
LP i
0 x x 0 s s
s
>
nt con ert summary if a constraint of an is a constrai , In LP i s nt convert
it to an constraint by adding a slack variable to the th
constraintandaddingthesignre
equal
striction
ity
,

i
i
i
0
s
s > constraint and adding the sign restriction .
i
0 s >
an has both and constraints apply the previous
procedurestotheindividual constraints
, If LP s >
procedures to the individual constraints
an below:
.
Consider LP
standar Non d form
M 20 15 +
Standard form
20 15

M
1 2
Max
Subject to
z 20x 15x = +
1 2
z 20x 15x Max
Subject to
= +
1
j
x


1 1
100
x 100
x 100
j
s


s





s
= +
100 +
2
1 2
x 100

50x 35x 6000

s
s
+
2
1 2
2
3
x 100
50x 35x 6000
s
s




+
+
=
= +
1 2
20x 15x 2000
x x 0
>
>
+
-
1 2 4
20x e
s
15x 20
s s e
00
x x 0
= +
>
,
1 2
x x 0 >
, , , , ,
1 2 1 2 3 4
s s s e x x 0 >
2. Preview of the Simplex Algorithm
an with m constraints and n variables
has been converted into the standard form

.
Suppose LP
from of such an is: The LP
M Mi or
1 1 2 2 n n
z c x Max Min
Subject
c x c x
t
= + + +
b
o Subject t

11 1 12 2 1n n 1
21 1 22 2 2n n 2
a x a x a x b
a x a x a x b
o
+ + + =
+ + + =

21 1 22 2 2n n 2
1 1 2 2
a x a x x a + + + b =
m1 1 m2 2 mn
a x a x x a + + +
, . ,
n m
i
b
x 0, i = 1 n >
we define
| | | | | |
| | |
If

11 12 1n 1 1
a a a x b
, , ,
| | |
| | |
| | |
| | |

21 22 2n 2 2
a a a x b
A x b = = =
theconstraintscanbewrittenasasystemof equations
| | |
\ . \ . \ .

m1 m2 mn n m
a a a x b
the constraints can be written as a system of equations
. Ax b =
a basic solution to is obtained by setting
variables equal to and solving for the remaining
i bl


, If n m Ax b
n m 0
>

=
variables
assumes that setting the variables equal to
.
This
m
n 0 m
yields a value for the remaining variables or
equivalently the columns for the remaining variables
li
uniqu
l i d d t
e ,
,
m
m
are linearly independent.
find a basic solution to we choose a set of
variables( ) and set each of
,
the , or nonbasic variables
To A
N
x b n m
BV
=
( )
these variables equal to
we solve for the values of the ( ) variabl
,
.
Then
0
n m m n = es ( )
nonbasic variables basic solutions
( ) that satisfy
lead to different
the , or .
.
basic variables
Different
Ax b BV =
basic so the to the following equations: lution Consider
1 2
x x + 3 =
1 2
number of nonbasic variables . The
2 3
- x x -1
3 2 1 -
+ =
= =
for example { }then { }
canobtainthevaluesfor th basicvariable ese bysetti s ng
, , , , . Setting
We
3 1 2
x NB x x V V B = =
can obtain the values for th basic variable ese by setti s ng
,
We
3
x 0 = that is

.
1 2
2
x x 3
- x - 0 1
+ =
+ =
basic solution and is a
.
, , . Thus
2
1 2 3
x 0
x 2 x 1 x 0 = = =
{ }and { }the becomes basic solution
d
, , If
1 2 3
NBV BV x x x
0 3 2
= =
basic solution
and
{ }and { }the becomes
, .
, , If
1 2 3
2 1 3
NB
x 0 x 3 x 2
x x x BV V
= = =
= =
and
set
, .
Some
1 2 3
x 3 x 0 x -1 = = =
s of variables do not yield a basic solution . m y
the following linear system: Consider
1 2 3
x 2x x 1

+ + =
1 2 3
1 2 3
x 2x x 1
2x x x 4
+ +
+ +
{ }and { }thecorrespondin basic g
.
If
3
x x x NB B V V
=
= = { }and { }the correspondin basic
sol
g
would satisf ut on y i
, ,


If
3 1 2
1 2
x x x
x 2x 1
NB B V V = =
+ =
t
.
Since
1 2
2x 4x 3

+ =
basi his c so system h lution as no solution, there is no y
corresponding to { } , .
1 2
x B x V =
Preview of the Simplex Algorithm
basic solution in which all variables are
is called a ( ) basic feasible soluti
nonnegativ
o
e
n or .
Any
bfs ( )
the basic solutions in the previous exa e mpl , , For
1
x 2
x 1 x
f
=
and arebasicfeasible 0 x 0 x 3 x 2 ,
2
x 1 x = and are basic feasible
solutions the basic solution fails
b ( )
, ,
, , ,
b
but
3 1 2 3
1 2 3
b
0 x 0 x 3 x 2
x 3 x
f
0 x -1
0
= = = =
= = =
to be a ( )
basic feasibl following show that a e
because .
two t The heorems
3
bfs x 0 <
g
is of solution great importance in . LP
easibleregionfor an linear programmingproblem
:
f The
Theorem 1
easible region for any linear programming problem
is a convex set
if h h b i l l i
f
.
The
Al LP if an has an , there must be an
extreme point
o
of
ptim
the
al solution
feasible region optimal that is
,
.
Also LP
Theorem
uniqueextreme any there is a point of the
:
, s

'
For L P P L
2
q
each basic feasib
y p
feasible region corresponding to solution
thereisatleastone correspond
l
ingtoeachextreme
e .
Also bfs there is at least one corresponding to each extreme
point in the fe
, Also bfs
asible region.
relationship between and
l ti tli di i
extremepoints basic feasi
i
ble The
Th E l 1 2 solutions outlined in is seen in
( ) is: in the standard form
, .
with slack variabl es The
Th Example 1
LP
eorem 2
X2
6
0 D
1 2
Maximiz z 4x 3x e
Subj
= +
ect to
5
0
Labor Constraint
Subj
.
1 1 2
x x 40
e
s
s 2x x 60
ct to
+ + =
+ +

3
0
4
0
Feasible Region
B

.
, , ,
2
1
1
2
2
1 2
s
s
2x x 60
x x s 0
+ + =
>
2
0
0
E
Leather Constraint
inequalities show the
.
Two shaded
area

1
0
A C
extrem of the
feasible region are
e points
, , ,
The
B C E . F
X1
10 20 30 40 50
A C
F
BV NBV bfs Corner Point
Table 1
, , , ( )


1 2 1 2 1 2 1 2
s s s s
BV NBV bfs Corne
x x 0 x x 20 E z 140
r Point
= = = = =
, , , , ( )
1 2 2 1 1 2 2 1
x x x 0 x 30 10 C z 120
0
s s s s = = = = =
( ) bf 40 20 0 , , ,
1 2 2 1 1 1 2
x x x 0 s s s x = =

-
-
, ( )
, , , , ( )
2 2
1 2 2 1 1 2 1 1 2
s s
s s s s s
bfs
bfs
40 20 0
x x x 0 x 60 20 0
no
no
= = <
= = = = < ( )
, , , , ( )
1 2 2 1 1
2 1 1 2
2 1 1 2
2 1 1 2
s s s s
f
x x x 0 x 40 20 B z 120 = = = = =

showsthecorrespondenc
, , , , ( )
1 2 1 2 1 2 1 2
Table 1
s s x x x x 0 40 60 F z 0 s s = = = = =
basicfeasiblesolutions ebetweenthe shows the correspondenc Table 1 basic feasible solutions
basicfeasiblesolutions
e between the
to the and the extreme points of the feasible region
tothestandardformof the correspond
.
The
LP
LP basic feasible solutionsto the standard form of the correspond
in a natural fashion to the ex

s
'
The LP
LP treme points.
SimplexAlgorithmProcedurefor maximization The LPs
The Simplex Algorithm For Max LPs
Simplex Algorithm Procedure for maximizati
standa
on
the to
a ( ) fromthe
rd form
standardform


if possible
The
Convert
Obtain
LPs
LP Step 1
Step 2 bf

s a ( ) from the standard form if possible

Obtain
Deter
Step 2
Step
bf
3
s
whether the current is optimal
h i i l d i hi h
mine
If S
bfs
bf 4 the current is not optimal determine which
should be com nonbasic variable basic e a and variabl
which should bas becom icvari
e
able
, If S bfs tep 4
eanonbasicvariable which should bas becom ic variable e a
to find a with a better objective function value
tofindanew
nonbasic variable
.
s Use elementary row operations(ero Step )' bfs 5
bfs
to find a new
with a better objective function value b
s
.
Use
Go
elementary row operations(ero Step ) bfs 5
ack to
thi i l l ith th i i th f
.
I bj ti f ti
Step 3
: this simplex algorithmthe is in the form

, In objective funct n io

1 1 2 2 n n
z c x c x c x 0.

=
. call this format the row version of the objective function We 0
:
1 2 3
z 60x 30x 20x
8 6 48
Examp maximize
bj t
le 2
t s
= + +

1 2 3
1 2 3


8x 6x x 48
4x 2x 1.5x 20
subject t o s
s
+ +
+ +
1 2 3


2x 1.5x 0.

x

5 8 s + +
2
x 5 s
, ,
Standar form
1 2 3
x x x 0. >

1 2 3
1 2
maximi z 60x 30x 20x
8x 6x
ze
subject t

o


= + +
+
3 1
x 48 s = + +
1 2
j
3
1 2
1
2 3
4x 2x 1.5x 20
2x 1 5x 0 5x 8
s
s




=
=
+ + +
+ + +
1 2 3
2
3
2x 1.5x 0.5x 8

s
x
=

+ + +
4
5
0
s = +
, , , , , ,
1 2 3 1 2 3 4
x x x s s s 0. s >
rows togetherinrow andthesignrestrictions
So
Pu
lu
t
tion
ting 1- 4 0
Example 2 - con
rows together in row and the sign restrictions
yields the equations and basic variables

.
Putting 1 4 0
Canonical Form0 BasicVariable Canonical Form 0 Basi

1 2 3
z 60x 30x 20x 0 z 0
8 6 48
c Variable
Row 0
R 1 48


= =
1 2 3
1 2 3
1 1
2
8x 6x x 48 Row 1 s s
Ro
48
4x 2x 1.5x w 2 s




=

= + + +
+ + +
2
s
R 3
20 20
2 1 5 0 5 8 8
= =
3 3
4
1
4
2 3
2
Row 3 s s
Row
2x 1.5x 0.5x 8 8
x 5 4 s 5 s



=
= =
=
+ + +
+
we set If
1 2 3
x = x = x we can solve for the values
{ }and { }
, , , , .
, , , , , , . Thus
1 2 3 4
1 2 3 1 2 3 4
=
BV
0
=
s s s s
s s s NB = x x x V s
each constraint is then in canonical form( s have a coefficient
in one row and zeros in
Since = 1 BV
right- hand side) r ) with a nonnegative (
a canbeobtainedbyinspection
hs all other rows ,
bfs a can be obtained by inspection. bfs
Obt i B i F ibl
Solution
S l ti
- Example 2 - con Step 2
perform the simplex algorithmwe need a basic( , although
.
To
Obtain a Basic Feasible Solution
) variable for
appears in w
not necessarily nonnegative .
Since
row 0
row 0 z ith a coefficient of and does , 1 z pp
not appear in any other we use as the
thisconvention thebasicfeasiblesolutionfor our
basic variable , .
, With
row z
this convention the basic feasible solution for our
initial canonical form has:
{
, With
= V z B s }and { }
1 1 2 3 2 3 4
= x x x NBV s s s { , V z B s }and { }
this initial
, , , , , .
, , , , , , For
1 1 2 3 2 3 4
1 2 3 4
x x x
z = 0 = 48 = 20 = 8 = 5
x = x = x =
NBV
bfs s s s
0
s s s
s
this example indicates a can be used as a
basicvariableif
sla
the of the
ck variable
.
, As
1 2 3
x = x =

rh
x =
s
0
constraintisnonnegative basic variable if the of the rhs constraint isnonnegative.
we have obtained a we need to determine whether it is optimal
:
, . Once
bfs
bfs
Determine if the Current is Optimal Step 3
do this we try to determine if there is any way can be increased
by inc
, To z
nonbasic variable
b i i bl
reasing some from its current value of zero
hil h ldi ll th t th i t l f nonbasic variabl while holding all other at their curr es ent values of
zero
for in yields:
.
Solving row0 z z 60x 30x 20x = + +
non
for in yields:
each


Solving
For
1 2 3
row 0 z z 60x 30x 20x = + +
basic variable
nonbasic variab
we can use the equation above to determine
if increasing a ( le
,
but hold nonbasic ing all other g (
nonbasic
) will increase
any of the w varia i l ble l
vari
g
to . a l

b es
Increasing
0 z
cause an increase in . z
increasing causes the greatest rate of increase in
increases from its current
b i i
value of it will have to become a
bl
.
,
However
If
1
1
zero
x
x
z
basic varia
this reason is
ble.
, For
1
x called the
h th t
entering varia
ti ffi
ble
i ti
.
Ob 0 has the most negative coefficient in . Observe
1
row 0 x
h h ( ) h i i bl
:
bl
Step 4
choose the ( ) to the
with the most negative coefficient
entering variable
in
desiretomake aslargeaspossiblebutaswedo the
nonbasic variable
in a max problem

We
We
row 0.
x desire to make as large as possible but as we do, the
current basi
We
1
x
c variables( ) will change value
increasing maycauseabasicvariabletobecomenegative
, , , .
Thus
1 2 3 4
s s s s
x increasing may cause a basic variable to become negative. ,

Thus
1
x
RATIO
wehave In Since row1 s s = 48 8x 0 x 48 8 = 6 > s we have
we have
wehave
, . ,
, . ,
In Since
In Since
In Since
1 1
2
1 1
1 2 1
row 1 s s
row 2 s s
row 3 s s
= 48 - 8x 0 x 48 8 = 6
= 20 - 4x 0 x 20 4 = 5
= 8 2x 0 x 8 2 = 4


> s
> s
> s we have , . ,
,
In Since
In
3 3 1 1
row 3 s s = 8 - 2x 0 x 8 2 = 4
row

4
> s
we have any will always be
t k ll th b i i bl ti th l t
. , . For
Thi
4 1 4
5 = x s s 0 >
means to keep all the basic variables nonnegative the largest
we can make is { }
,
, , .
This
1
6 5 4 4 m n x i =
The Ratio Te Solution- st Example 2 - con

entering a variable into the basis, compute the ratio:
of of entering variable in row


When
rhs row coefficient g
every constraint in which the entering For
ff
variable has a
positivecoefficient positive coefficient
constraint with the smallest ratio is called the winner
of theratiotest
.
The
of the ratio test
smallest ration is the large enterin st value of the
that will keep
g
v all ari ble a
.
The
thecurrent basicvariables that will keep v all ari ble a the current basic variables
nonnegative
th t i i bl b i i bl i
.
M k 3 the entering variable a basic variable in
since this ( ) was the winner of the ratiotest
( )

constraint
Make
1
x
8 / 2
row 3
row
4 = ( ). 8 / 2 4 =

k i l t b i i bl T
Solu The ti Ratio Test on-
3
Example 2 - con
make a in we use elementary
row operations( ) to make have a coefficient of
in and
b
acoefficient
a
of inall othe
sic va e
r
riabl ,
s
'
To
1
1
1
row x
x
0
3
ero
3

in and a coefficientof in all other 0 row 3 r
procedure is called pivoting on ;and
i ll dth
.
This
ows
row 3 row 3
i

is called the
final result is that replaces as the basicvariable
.
The
3 1
pivot row
s x
for
term in the that involves th
.
The
row 3
pivot row e entering basic
variable is called the . pivot term
Gauss- J ordan method using and
basic shownonthiss va lidemak riable es a
: s
.
'
The
1
ero simplex
tableaus x
Step 5
basic shown on this s va lide mak riable es a .
1
tableaus x
ero 1
z x1 x2 x3 s1 s2 s3 s4 rhs
ero
0 1 -60 -30 -20
1 8 6 1 1 48 1 8 6 1 1 48
2 4 2 1.5 1 20
3 1 0.75 0.25 0.5 4 row 3 divided by 1/2
4 1 1 5
ero 2
1 2 3 1 2 3 4 h
ero 2
z x1 x2 x3 s1 s2 s3 s4 rhs
0 1 15 -5 30 240 60 times row 3 added to row 0
1 8 6 1 1 48
2 4 2 1.5 1 20
3 1 0.75 0.25 0.5 4
4 1 1 5
ero 3
z x1 x2 x3 s1 s2 s3 s4 rhs

0 1 15 -5 30 240 0 1 15 5 30 240
1 -1 1 -4 16 - 8 times row 3 added to row 1
2 4 2 1.5 1 20
3 1 0.75 0.25 0.5 4
4 1 1 5 4 1 1 5
ero 4
z x1 x2 x3 s1 s2 s3 s4 rhs
0 1 15 -5 30 240
1 -1 1 -4 16
2 -1 0.5 1 -2 4 - 4 times row 3 added to row 2
3 1 0.75 0.25 0.5 4
4 1 1 5
The Result
Canonical Form1 Basic Variable
Example 2 - con

3 2 3
Canonical Form 1 Basic Va
z 15x 5x 30 240 z 240
riable
Row 0 s
R

1

+ + = =
4 16 16 Row1
1 3 1
2 3 2
3
2 3
x 4 16 16
x 0.5x 2 4 4
s s s
Row 2 s s s
+ = =
+ + = =


1 2 3 3 1
x 0.75x 0.25x 0.5 4 x 4

Row 3 s
Row 4 x
+ + + = =
2 4 4
5 s s 5 + = =
{ , , , , }and { }
the and
, , , . In
Yielding

1 2 1 3 2 4 3
BV NBV Canonical Form z x x x
z 240 16 4 x 4 5
1 s s s s
s s bfs s
= =
= = = = = the and
procedurefromgoingt
, , , ,
.
Yielding
The
1 2
3
1
2 3
4
z 240 16 4 x 4 5
x
s s bf
0
s
x
s
s
= = = = =
= =

=
oone toabetter adjacent is bfs bfs procedure from going t The o one to a better adjacent is
called an iteration( ) of the simplex algorithm

or sometime pivot .
bfs bfs
to wecandetermineif thecurrent Returning Step 3 bfs
Example 2 - con.
to we can determine if the current
is optimal
from yields:
,
.
Returning
Rearranging
Step 3
row0 canonical form1
bfs
from yields:

Rearranging
2 3 3
row 0 canonical form 1
z 240 15x 5

s x 30 = +
current i o s n The bfs optimal because increasing to
( ) will increase
t
while holding the other to nonbasic variable
3
1 x
0
the value of
either or basic will caused the value of
.
Making
2 3
z
x z s
to
- Reca
decre
ll
a
th
se.
Step 4 e rule for determining the entering variable p g g
is the coefficient with the greatest negative value
is the only variable with a coefficient negative
.
, Since
3 3
x x
row 0
y
should be entered int
g
o e th
,
3 3
basis.
iteration( ) pivot Begin 2
Example 2 - con.
( )
the ratio test using as the entering variable
yieldsthefollowingresults( ):

holding other s to zero
p g
Performing
3
x
N

BV yields the following results( ):
for all vaules of
holding other s to zero
, From
3 1
row 1 0 x
NBV
s > since
1 3
s = 16 x +
,
,
From if
From if
2 3
3 1
row 2 s
row 3
0 x 4 0.5 = 8
x 0 x 4 0.25 = 16
> s
> s
for all vaules of
t k ll th b i i bl ti th
, . From since
Thi
3
3 4
1
4
row 4 s 0 x 5 = s >
means to keep all the basic variables nonnegative the
l
, This
argest we can make is { } , .
3
x 8 16 = min 8
pivot becomes the
use to make a basic variable in
, .
- s, .
'
So
Now
3
row 2 ro
er
w
Step 5 r x ow 2 o
3
use to make a basic variable in - s, .
'
Now
3
ero Step 5 row 2 x
ero 1
z x1 x2 x3 s1 s2 s3 s4 rhs
0 1 15 -5 30 240
1 -1 1 -4 16
2 -2 1 2 -4 8 multiplied row 2 by 2
3 1 0.75 0.25 0.5 4
4 1 1 5
ero 2
z x1 x2 x3 s1 s2 s3 s4 rhs z x1 x2 x3 s1 s2 s3 s4 rhs
0 1 5 10 10 280 add 5 times row 2 to row 0
1 -1 1 -4 16
2 -2 1 2 -4 8
3 1 0 75 0 25 0 5 4 3 1 0.75 0.25 0.5 4
4 1 1 5
ero 3
z x1 x2 x3 s1 s2 s3 s4 rhs
0 1 5 10 10 280
1 -2 1 2 -8 24 add row 2 to row 1
2 -2 1 2 -4 8
3 1 0.75 0.25 0.5 4
4 1 1 5
ero 4
z x1 x2 x3 s1 s2 s3 s4 rhs
0 1 5 10 10 280
1 -2 1 2 -8 24
2 2 1 2 4 8 2 -2 1 2 -4 8
3 1 1.25 -0.5 1.5 2 add -1/4 times row 2 to row 3
4 1 1 5
The Result in Example 2 Canonical Form 2 - con

2 3 2
Canonical Form 2
z 5x 10 10 280 z 280
Basic Variable
Row 0 s s + + + = =
2 3 2
Row 1
1 2 3 1 2
2x 2 8 24 24
2x x 2 4 8 x 8
s s s s
Row 2 s s






+ + = =
+ + = =
2 2 3 3
1
3
2 2 1 3
2x x 2 4 8 x 8
x 1.25x 0.5 1.5 2 x 2
Row 2 s s
Row 3 s s
R 4



+ + = =
+ +

= =
5 5 + Row 4
, In
4 4 2
x 5 5 s s
Canonical Form 2
+ = =
{ , , , , }and { }
the and
, , .
, , , , Yielding

1 4 3 2
1 4
3 2
3
1
1
= z x BV NBV
bfs
x = x
z 28
s s s s
s s 0 24 x 8 x 2 5 = = = = = , , , ,
.
g
1 4
2
3
2 3
1
f z
s s x 0 = = =
Example 2 - con.
for in yields: Solving
2 2 3
z

row 0
z 280 5x 10 10

s s =
can see that increasing (
) will causethevalueof tode
, , or while holding the
other s to
We
2
2
2 3
2 3
z
s s
NBV z
x
0 crease ) will cause the value of to de other s to NBV z 0 crease
solution at the end of is therefore optimal
.
. The iteration 2
following rule can be applie
canonical form
d to determine whether a
optimal s is :


'
The
bfs
optimal is ( for a max pro Acanonical form ) if each
nonbasic variable has a nonnegative coefficient in the
blem
canonical forms .
'
row 0
simplexmethodalgorithmrequiresastarting The bfs
The Big M Meth d o
simplex method algorithm requires a starting
problems have found
l k i bl
starting by using
b th i i bl
.

The
Previous
bfs
bfs
slack variabl ba the as our
an have or constra
sic variabl s
nts
e
i
es .
, If LP > = however a starting
t b dil t
,
bf may not be readily apparent
such a case the method may be used to solve
h
.
, In
bfs
Big M
LP the

1 2
minimize
LP
z 2x 3x = +
.
1 2
0.5x 0.25x 4 subject to + s
.
1 2
x 3x 20 + > .
.
1 2
1 2
x 3x 20
x x 10
0


+ >
+ =
> . ,
1 2
x x 0 >
LP standard form The in
=
1 2
1 2 1
z 2x 3x
0.5x 0.25x 4
Minimize
Subject to

s
+
+ + =
1
2
2
1 2
1
x
j
3x 20
x x 10
e

+ =
+ =
1 2
x x 10

+ =
0 , , , >
1 2 1 2
x x s e
remedythepredicament arecreated artificial variables
order to use the simplex method a is needed , . -
To
In s bf
- remedy the predicament are created
variables will be labeled accord
artificial var
ing t
iables , .

To
The

- o the row in which they
areused are used.
instandardf artificial variab ormwith is les Th LP
Example 3 - con. Artificial Variabl
e
es
instandard f artificial variab orm with is les

Th
1 2
LP
Row 0
R 1
z 2x 3x 0
0 5 0 25
e
=
4
1 2
Row 1 s 0.5x 0.25x + +
2
1
2 1 2
4
x 3x 20 Row 2 e a


=
+ + =
theoptimal solution a artificial v ll ariable m s ust , - In
3 1 2
x x 10 Row a 3 + + =
the optimal solution a artificial v ll ariable m
b
s ust

, In
e set equal to zero
accomplishthis ina aterm isadded
.
To min Ma LP - accomplish this in a a term is added
to the objective functio artificial variables n for each
a theterm isaddedtotheobjective
, ,
.
To
For
i
i
min M
max M
a LP
P
a
a L
-
- a the term is added to the objective
function
, For
i
max - M P a L -
for each
t l b
.
i
M
a
represents some very large number. M -
Example 3 - con.
modified in standard form then becomes



The
1 2 2 3
Row 0
LP
z 2x 3x 0

- M - M a a =
1 1 2
Row 0.5x 0.25x 4

1 s
Row 2
+ + =
1 2 2 2
e x 3x 20 a + + =
theobjectivefunctionthiswaymakesit - Modifying
1 2
2
2
3 1
2
Row x x 3 1 a 0 + + =
the objective function this way makes it
extremely costly for an a
- Modifying
to be positive rtificial variab s le .
optimal solutionshould force . The
2 3
a a 0 = =
Initial Tableau Example 3 - con.

Row z x1 x2 s1 e2 a2 a3 rhs
0 1 00 2 00 3 00 M M 0 00 0 1.00 -2.00 -3.00 -M -M 0.00
1 0.50 0.25 1.00 4.00
2 1 00 3 00 -1 00 1 00 20 00 2 1.00 3.00 -1.00 1.00 20.00
3 1.00 1.00 1.00 10.00

2 3
in Row 0, To make = =0 a a
Row 0+ M*(Row 2)+M*(Row 3)
Pi vot 1 z x1 x2 s1 e2 a2 a3 rhs rati o ero
0 1.00 2M - 2 4M -3 -M 30M Row 0 + M(Row 2) + M(Row 3)
1 0.50 0.25 1.00 4.00 16
2 1.00 3 -1.00 1.00 20.00 6.67
3 1.00 1.00 1.00 10.00 10
ero 1 z x1 x2 s1 e2 a2 a3 rhs ero
0 1.00 2M-2 4M -3 -M 30M
1 0 50 0 25 1 00 4 00 1 0.50 0.25 1.00 4.00
2 0.33 1 -0.33 0.33 6.67 Row 2 di vi ded by 3
3 1.00 1.00 1.00 10.00
ero 2 z x1 x2 s1 e2 a2 a3 rhs ero
0 1.00 (2M-3)/3 (M-3)/3 (3-4M)/3 (60+10M)/3 Row 0 - (4M-3)*(Row 2) ( ) ( ) ( ) ( ) ( ) ( )
1 0.50 0.25 1.00 4.00
2 0.33 1 -0.33 0.33 6.67
3 1.00 1.00 1.00 10.00
ero 3 z x1 x2 s1 e2 a2 a3 rhs ero
0 1 00 (2M 3)/3 (M 3)/3 (3 4M)/3 (60 10M)/3 0 1.00 (2M-3)/3 (M-3)/3 (3-4M)/3 (60+10M)/3
1 0.42 1.00 0.08 -0.08 2.33 Row 1 - 0.25*(Row 2)
2 0.33 1 -0.33 0.33 6.67
3 1.00 1.00 1.00 10.00
ero 4 z x1 x2 s1 e2 a2 a3 rhs ero ero 4 z x1 x2 s1 e2 a2 a3 rhs ero
0 1.00 (2M-3)/3 (M-3)/3 (3-4M)/3 (60+10M)/3
1 0.42 1.00 0.08 -0.08 2.33
2 0.33 1 -0.33 0.33 6.67
3 0.67 0.33 -0.33 1.00 3.33 Row 3 - Row 2
The Solution- con
Pi vot 2 z x1 x2 s1 e2 a2 a3 rhs rati o
0 1.00 (2M-3)/3 (M-3)/3 (3-4M)/3 (60+10M)/3
1 0.42 1.00 0.08 -0.08 2.33 5.60
2 0.33 1 -0.33 0.33 6.67 20.00
3 0.67 0.33 -0.33 1.00 3.33 5.00
ero 1 z x1 x2 s1 e2 a2 a3 rhs ero
0 1.00 (2M-3)/3 (M-3)/3 (3-4M)/3 (60+10M)/3
1 0 42 1 00 0 08 -0 08 2 33 1 0.42 1.00 0.08 0.08 2.33
2 0.33 1 -0.33 0.33 6.67
3 1.00 0.50 -0.50 1.50 5.00 (Row 3)*(3/2)
ero 2 z x1 x2 s1 e2 a2 a3 rhs ero
0 1.00 -0.50 (1-2M)/2 (3-2M)/2 25.00 Row 0 + (3-2M)*(Row 3)/3
1 0 42 1 00 0 08 0 08 2 33 1 0.42 1.00 0.08 -0.08 2.33
2 0.33 1.00 -0.33 0.33 6.67
3 1.00 0.50 -0.50 1.50 5.00
ero 3 z x1 x2 s1 e2 a2 a3 rhs ero
0 1.00 -0.50 (1-2M)/2 (3-2M)/2 25.00
1 1 00 0 13 0 13 0 63 0 25 R 1 (5/12)*R 3) 1 1.00 -0.13 0.13 -0.63 0.25 Row 1 - (5/12)*Row 3)
2 0.33 1.00 -0.33 0.33 6.67
3 1.00 0.50 -0.50 1.50 5.00
ero 4 z x1 x2 s1 e2 a2 a3 rhs ero
0 1.00 -0.50 (1-2M)/2 (3-2M)/2 25.00
Optimal Solution
1 1.00 -0.13 0.13 -0.63 0.25
2 1.00 -0.50 0.50 -0.50 5.00 Row 2 -(1/3)*Row 3
3 1.00 0.50 -0.50 1.50 5.00
p
The summary for the optimal solution
variable with the positive coefficient in row
should enter the basis since this is a problem.
The
min
-
ratio test indicates that hould enter the s The
2
x -
basis in which means the artificial variable , row 2
will leave the basis
toeliminate from and
.
s

'
Use
2
2
row 1 row 3 e x
a
ro - to eliminate from and
ratio test indicates that
wh
s .
should enter the
basis in
Use
The
2
1
row 1 row 3
row 3
e x
x
ro -
ichmeansthenthat will leave a wh basis in , row 3 ich means then that will leave
the basis.
Th R l
3
a
optimal solu is . tion , The
The Result
1 2
z 25 x x 5

= = = -
Tutorial
the simplex method to solve the following Using LP
M i i 3 5
1 2
Maximize
Subject to
z 3x 5x
4
= +
.
.
1
2
Subject to
x 4
2x 12
s
s
2
.
1 2
3x 2x 18
0
s +
,
1 2
x x 0. >
LPs The Simplex Algorithm for Min
the following : Consider LP
2 Mi i i 3
1 2
z 2 Minimize
Subject to
x 3x
4
=
.
1 2
Subject to
x x 4 s +
x x 6 s .
,
1 2
1 2
x x 6
x x 0.
s
>