V9 DZ W 1%9

Reminder :

f* =
A' Bb For change in b

cBA*
*
c
=
c- For change in c

* *
2
=

zot Cpt
↓
Alternatively ,
use objective function

& Cixi
<
Primal objective function

substitution of c; by bit e) for change in C

While xi are the same (from b*)

Substitution of ✗i by the entries of 6¥ for change in 6

While the Ci are the same
 Initial Final and
Notations:
● B0 = set of Basic Variable in initial Simplex Tableau ordered to form Identity in A
● B = set of Basic Variable in Final tableau ordered to form Identity in A*
● CB= [ B entries in initial row c]
● AB = Columns of B in the initial matrix A
● A*B0 = Columns of B0 in the final matrix A*
● (AB)-1 = A*B0

What if new variables or constraints (see section 5.7) are added?

How would the solution change without having to restart the simplex algorithm all over?

● xn+1 = new variable →cost coefficient cn+1 → add column of coefficients for the constraints A(n+1)
o Original minimize z = c X → minimize new z = c X+ cn+1 xn+1

o Such that

* *(n+1)
and c n+1 = cn+1 – cB A
*(n+1) -1 (n+1)
o New column in FINAL tableau: A = ( A B) A
*
o If c n+1 ≥0 xn+1 =Not a BV xn+1 =0 X* = remains Optimal BFS
o If c *
n+1 <0 proceed with the simplex algorithm enter xn+1 to BV
5. 4 Addition of a New Variable

{
Add ✗n " Objective Function +
Cnt, ✗my 2 =
CX + Cntixnti
"

Mam, µ →
[ AÑ ]
↑
Column for Xntl

A*p◦
'

Recall: A-pb =

captures all simplex operations

* "
A. =
1
A-, AM
] C*nH ≥ 0 Xnt / =
0 Optimal Point is

C*n+
*
[✗ ]
*
=

Cnn C, A 0

\
-

, ,

*
[ n +, <
0 Continue the simplex algorithm
(Enter Xnt Exit ? )
, ,

Furniture Example

Cont below
if
. -
 1.*"
● Furniture Example (Extra Variable F): final f xn"

Initial f
2
f
%
%
Do I

4 k
'

? C*nH
Suppose the furniture company has expanded to also produce bedframes (F). Suppose the requirements
of resources are given below. Find the change (if any) with =15 first, then =17.
Cnt / =/ 5 ( 4=15 Cnt/ =D Cifl>

Use Original Final Tableau (2nd above right without F-column) to find (AB)-1 = A*B0

*(n+1) -1
Then, Compute A = ( A B) A(n+1) *
and Compute c n+1 = cn+1 – cB A*(n+1) = 1/2
● for =15→ c*n+1 = 1/2 ≥0
○ xn+1 =Not in BV xn+1 =0
○ X* = remains Optimal BFS
● for =17 → c*n+1 =-3/2 <0
○ extra pivot to enter F (see
below)

● Can Example (Extra Variable v):

Suppose the aluminum can company could invest
in a fourth can-producing process v: such
producing 14 cases of the Type A, 16 cases of the
Type B, using 1 lb of recycled aluminum, and
costing (a) = $110 or (b) =$90.
Furniture Example

' *
⑨ A-p =
A Bo
"
*
""

A'BA
[ I
① = = =

CB #
""
② C*n+ ,
=
Cnt ,
-

t =
IS on 17
"2 ≥
0
]
/
"

✗ -

[ -19 -12 0] = .
; + ¥
↑
, ,

\ -

% ± 0
Maximizing

#→
③ Conclusion : for 1=15 →
Same final tableau →
Same 6 Same optimal point ¢ value w/✗my =
0

for 11=17 →
Cn*+ ,
= -

? <
0 →
Enter ✗my ,
Exit ?

>
Perform the std simplex algorithm
.
$
review new final tableau
Notice no slack vars felt and
, , z has increased!
 *
b
From section 5.5, what if changes in the constant-term column can bring negative entries into λ ?b
The associated basic solution is NOT feasible; that is, the constant-term column contains negative entries.
With the simplex method as we have developed it, the only way to handle such a problem would be to multiply the
equations with negative constant terms by (—1), add artificial variables if necessary to put the problem into canonical
form, and proceed with the two-stage simplex process!
This section: Dual Simplex Algorithm to better solve problems in this form (negative RHS entries) steps of the
Consider a LP in canonical form with respect to a set of basic variables and such that all cj ≥0.
1- If all bi≥ 0 then the Min value is achieved by the BFS given by the basic variables and RHS.
2- If there exists bi < 0 and aij ≥ 0 for all j (all i-row), NO solution exists.
3- Otherwise, use the Dual simplex Method below (use bi<0 to exit and aij <0 to enter). Example:
Determine the variable to exit and then the variable to enter:
1- Exit Variable: Pivot Row R with most negative b term
In this case, 2nd row, thus exit x5
2- Enter Variable: Column j with Max ratio cj/arj for arj < 0
In this case 10/(-4) = -2.5 & 4/(-2)=-2, pick -2 enter x3
3- Repeat if negative entries on the RHS (two pivots below)
Min is 24 and is attained at the point (2,0,1,0,0)

Can Example (more changes to RHS of constraints –see section 5.5):

New government regulations require an increase of the use of recycled aluminum by 28%, from 600 lb to 768 lb. How
will that increase affects the company's operation, marginal can production costs and using the recycled aluminum?

λ=168 bλ* = (AB)-1bλ = (AB)-1b + (AB)-1

Recall xi=# hours of process i. Modify RHS

with max z* = 150(-30) +200(276) =50,700,
apply the Dual Simplex Algorithm.
New optimal profit of $51,200 by using
Process 3 for 256 hr, an extra 160 cases of
Type A and 1040 cases of Type B are
produced, and the marginal cost (shadow
price) for meeting each of the requirements is zero, while for using the recycled aluminum is increased to
$66.67/lb.
5.6 Dual Simplex Algorithm

What if b changes to by leading to 6¥ not being positive

?

↳ Perform the dual simplex using the last column (instead of the last row !)

① if bj ≥
0 i
>
STOP , we have an optimal solution

µ
②
aij
≥ O tf j(in row i ) >
STOP ,
no solution
if y f, o
③ if aij < 0 >
Dual Simplex Method
Exit variable for that row i
•

Enter variable from the column j

•

,
such that Hai, for aij < 0
is a Max
◦
Use row operation for he entered variable to have 0's in column ,

except at row i. , column j > 1

Furniture Example with extra constraint:
Suppose now that we add the extra constraint that there is demand for no more than 40 bookshelves.
Added a constraint add a BV to the basis add the slack variable to the new constraint: {B;C; s3; s4}.

● Can Example (Extra Constraint):

The optimal
schedule
suggested
(0,75,150). They
want first process
to be used at least
16 hr/month

add a constraint x1≥ 16

Introducing a slack variable x9 and
multiplying by (-1) -x1 +x9 = -16
Add it to the second tableau and apply
the Dual Simplex Algorithm generates
the tableaux to the right.
The optimal solution point calls for using
Process 1 for 16 hr, Process 2 for 57
hr, and Process 3 for 162 hr.
This production schedule uses all three
processes for at least 16 hr and thus is
an optimal operating schedule, with a
profit of $41,990.