tora moment

x\ + 2x2 + 2x3 — X4 = 25
Xi,X2,X3,X4
TORA provides > 0module for the dual simplex method. From the SOLVE>MODIFY
a tutorial
menu select Solve 1 Algebraic 1 Iterations 1 Dual Simplex . Remember that you need to
show that and X41can
X2 convert = 2 constraints to inequalities.
serve as basic variablesYou
fordoa not need
basic to convert
feasible 1 Ú 2 constraints because
solution.
TORA will do the conversion internally.
Compute the associated A^ and c\. What is your conclusion?
7. Given4.4.2
the problem of maximizing
generalized c-X Simplex
LPP Lab (Dual
simplex to AX 4.4
subject Algorithm)
algorithm < b, Additional
X > 0, with Simplex
b > 0,Algorithms 183
suppose slack variables are added and the simplex algorithm is applied, leading
x1 +(primal)
orThe x2 … 1,simplex
-x1 - algorithm
x … -1. TheChapter
starting3solution is infeasiblenonoptimal.
if at least one
Theofdual
the
to a finite optimal solution. Define 2B and eginusing starts feasible
the basic variables butfrom the
right-hand sides of4.4.1)
simplex (Section the inequalities is than
starts better negative.
optimal and infeasible. What if an LP model
final tableau. Show
Solve the
starts
that —cg5~'
following is an
problems using
both nonoptimal
optimal
andDual
solution
Simplex
infeasible?
point to the dual problem.
Algorithm
Of course we can use artificial variables and arti-
(Hint. Recall ficial
where in the final
constraints tableaua starting
to secure the solution to the
solution. Butdual
this can beisfound.)
really not necessary because the
1.example
key idea 4.4-1
of both the primal and dual simplex methods is that the optimum feasible solu-
tion, when finite, always occurs at a corner point (or a basic solution). This suggests that
5.3 CHANGES IN THE Minimize z = 3x1 + 2x2 + x3
OBJECTIVE
a new simplex algorithm (developed by thisFUNCTION
author) can be developed based on tandem
use oftothe dual simplex and the primal simplex methods. First, use the dual algorithm to
subject
Suppose the simplex method
get rid has been(without
of infeasibility used to worrying
solve theabout
problem of minimizing
optimality). z,
Once feasibility is restored,
3x1 + x2 + x3 Ú 3
z = c X — zo, subject to AX
the primal = b, Xcan>be
simplex 0 with thefind
used to final
thetableau given
optimum. by
Alternatively, we can first apply the
- 3x
primal simplex to secure optimality (without
1 + 3x + x Ú
2 worrying
3 6 about feasibility) and then use the
A* b*
dual simplex to seek feasibility. x + x + x … 3
1 2 3
c* 4 x1, x2, x3 Ú 0
example 4.4-2
and with X* the associated optimal
In the present basic
example, thefeasible solution, but
first two inequalities are that now by
multiplied one- 1orto convert them to
Consider
more of the original the maximization
coefficientsThe
1 … 2 constraints. ofstarting LP
the objectivemodel
tableau is of Problem
function,
thus giventhe components of c,for
4-38(a), repeated
as follows: here areconvenience.
2. does this affect the already computed
changed. How solutionz =
Maximize to 2x
the
3 problem?
From the formulas Basic x x x
intothe previous section, it follows that changes
1 2 3 x 4 Solution
x5 in cx6can induce
subject
changes only in c* and ZQ, and zin fact, -the 3 new - 2 c* can be determined using
-x1 -+1 2x2 -0 2x3 Ú 0 8 0 c* = 0c —
CBA*. If the modified c* remains x4 nonnegative,
-3 -X1 * would remain an0 optimal solution
-3
-x1 -+1 x2 + x3 … 4
1 0
point; if not, more iterations of the
x5 simplex 3 algorithm
-3 -may
1 be0 necessary
1 to
0 complete-6
x 1 1 2x1 -1 x2 +0 4x3 … 010 1 3
the modified problem. However, we could initiate the algorithm on the modified,
6

previously final The

tableau. We illustrate. x1, x2, x3 Ú 0
tableau is optimal because all the reduced costs in the z-row are … 0 1cQ = - 3, cQ = - 2,
1 2
c = - 1, c4 = 0, c5 = 0, c6 = 02. It is also infeasible because at least one of the basic variables
Example 5.3.1.
3. Consider the linear programming problem of
is negative 1x4 = - 3, x5 = - 6, x6 = 32.
According to the dual feasibility condition, x5 1 = - 62 is the leaving variable. The next table
Maximizing
shows how the dual optimality +4x2 +*3
z = llxicondition + 15x4
is used (5.3.1)
to determine the entering variable.
subject to
3xi + X2 + 2x3 + 4x4 < 28 j = 1 j = 2 j = 3

8x1 + 2x2 — X3 +variable

Nonbasic 7x4 < 50 x1 x2 x3
> 0 1cQ j 2
Xi,X2,X3,X4 z-row -3 -2 -1
x5-row, a4j 3 -3 -1
Adding slack variables and then applying
Ratio, @ a @ , the
cQ j
5ja5j 6simplex
0 algorithm
yields
—2
3
the reduced
1
tableaux resolution shown in Table 5.5. The maximum value of z is 106, attained at
Solve the problem after expanding the problem by adding new constraints
the point (0,4,0,6).
The ratios show that x2 is the entering variable.
(i) 4x1 + xThe next tableau is obtained by using the familiar row operations, which give
2 + 4x4 ≥ 29

(ii) x1 + 3x3 + x4Basic

− 4x5 − x6 x=1 − 18 x2 x3 x4 x5 x6 Solution

z -5 0 - 13 0 - 23 0 4
x4 -4 0 - 23 1 - 13 0 -1
1
x2 -1 1 3 0 - 13 0 2
2 1
x6 2 0 3 0 3 1 1