ALO

INDE6372 Advanced Linear Optimization
INDE6372: LECTURE 8
DEGENERACY IN SIMPLEX
& LECTURE 9: CONVEXITY IN LO
Mary F. McGuire, Ph.D.
Department of Industrial Engineering
University of Houston
Degeneracy
Definition: An LP is degenerate if it has at least one
basic feasible solution in which a basic variable = 0.

Degeneracy occurs when the minimum ratio test produces
a tie.
When this happens, one of the tied basic variables will
have the value of 0 in the next iteration.
Degeneracy
Definition: An LP is degenerate if it has at least one
basic feasible solution in which a basic variable = 0.

Degeneracy occurs when the minimum ratio test produces
a tie.
When this happens, one of the tied basic variables will
have the value of 0 in the next iteration.
Degeneracy is NOT a problem unless it leads to
cycling. This is where the simplex algorithm gets stuck in

a feasible solution, but never reaches an optimal solution.
(degenerate pivot).
Example
Example
Degeneracy Example
max
2x2x
1 ++
2
max 5x5x
1
(1)
(1)
x1x++
x2x6;6;
(2)
(2)
x1x, 1x,2x
0.
2 0.
(4)
(4)
x1x
x 0;
1 2x2 0;
(3)
(3)
Initial
Tableau
Initial
Tableau
Initial
Tableau
Basic
Var. x1x1
Basic
Var.
Z Z
-5 -5
x3 x3
11
x4 x4
11
x2x2
-2-2
11
-1-1
x3x3
00
11
00
x4x4
0 0 00
0 0 66
1 1 00
Jiming Peng Department of Industrial Engineering University of Houston

INDE6372: Lecture 8 Degeneracy in Simplex
Example
Degeneracy Example
Perform the pivot

Basic Var.
Z
x3
x1
x1
0
0
1
x2
-7
1
-1
x3
0
0.5
0
x4
5
-0.5
1
0
3
0

Example
Example
Degeneracy Example
Perform
Perform the
the pivot
pivot
Basic
Basic Var.
Var.
ZZ
xx33
xx1
1
Final tableau
xx1 xx2 xx3

xx4
1
2
3
4
00
-7
00
55 00
-7
00
11
0.5
0.5 -0.5
-0.5 33
11
-1
00
11 00
-1
Basic Var. x1 x2
Z
0 0
x2
0 1
x1
1 0
Final Solution: x1
x3
x4
3.5 1.5 21
0.5 -0.5 3
0.5 0.5
3
= 3, x2 = 3!

A Cycling Example
A Cycling Example
Cycling
Example
Adapted from the textbook, Page 31.
Adapted from the textbook, Page 31.

max
max
10x1 57x2 9x3 24x4
(5)
10x1 57x2 9x3 24x4
(5)
0.5x
5.5x2 2.5x3 + 9x4 0;
1
0.5x
1 5.5x2 2.5x3 + 9x4 0;
(6)(6)
x1x
1;1;
1
(8)(8)
0.5x
1.5x220.5x
+ xx44 0;0;
1
0.5x
0.5x33 +
1 1.5x
(7)(7)
0.
x1x, 1x,2x,2x,3x,3x, 4x40.
(9)(9)
Initial
Tableau
Initial
Tableau
Initial
Tableau
x
x1 1
Z
-10
Z
-10
x5
1
x5x6 1 1
x6 x7 1 1
x7
1
x2 2
57
57
-11
-11-3
-30
0
x33 x44 x55

9 24 0
9 24 0
-5 18 2
-5-1 18
20
2
-10 20 00
0
0
0
x6 x7
x6 x7
0 0 0
0 0 0
0 0 0
20 00 0 0
02 10 1 0

A Cycling Example: Continuation
Cycling Example
Update the pivot

x1
Z
0
x1
1
x6 0
x7
0
x2
-53
-11
8
11
x3
-41
-5
4
5
x4
204
18
-16
-9
x5
20
2
-2
-2
x6
0
0
2
0
x7
0
0
0
1
0
0
0
1


Cycling Example
Update the pivot

Update the pivot
x1 x2 x3
x4
x5
x2 -41
x3 204
x4 20
x5
Z
0x1 -53
-53 -5
-41 18
204 220
x1Z
10 -11
18 -22
x61 0 1 8-11 4-5 -16
-16 -2-2
x
54 -9
7 x6 0 0 118
x7
0
11
5
-9
-2
One more iteration
x1 x 2 x3 x4
x5
Z
0 0 -14.5 98 27/4
x1 1 0
0.5
-4 -3/4
x2
0
1
0.5
-2 -1/4
x7
0 0
-0.5
4
3/4
x6 x7
0x6 0x7
00 00
20 00
02 10
0 1
x6
53/4
11/4
1/4
-11/4
0
00
00
10
1
x7
0
0
0
1
0
0
0
1

10
Cycling Example
One more iteration

x1
Z
29
x3
2
x2 -1
x7
1
x2
0
0
1
0
x3
0
1
0
0
x4
18
-8
2
0
x5
-15
-1.5
0.5
0
x6
93
5.5
-2.5
0
x7
0
0
0
1
0
0
0
1

11

Cycling Example
One more iteration

One more iteration
x1 x2 x3 x 4
x4
Z
29x1 0x2 0x3 18
x3Z
229 00 10 18
-8
xx32 -1 2 10 01 2-8
x
-1
1 0
2
x7 2 1
0 0
0
x7
1
0 0
0
Continuing
x1
x 2 x3 x 4
Z
20
9
0 0
x3
-2
4
1 0
x4
-0.5 0.5 0
1
x7
1
0
0 0
x5
x6
x7
x5
x6 x07
-15
93
-15
93
-1.5
5.5
00
-1.5 5.5 0
0.5
-2.5 0
0.5 -2.5 0
0
0
1
0
0
1
x5
-10.5
0.5
1/4
0
x6
70.5
-4.5
-5/4
0
x7
0
0
0
1
0
00
00
0
1
1
0
0
0
1

12
Cycling Example
One more iteration

x1
x2
Z
-22
93
x5
-4
8
x4 0.5 -1.5
x7
1
0
x3
21
2
-0.5
0
x4
0
0
1
0
x5
0
1
0
0
x6
-24
-9
1
0
x7
0
0
0
1
0
0
0
1

13
Cycling Example: Continuation

AA Cycling
Continuation
Cycling Example
One more
more iteration
iteration
One
xx2
xx3 xx4 xx5 xx6
xx11
2
3
4
5
6
Z
-22
93
21
Z
-22
93
21 00 00 -24
-24
-4
88
22
00 11
-9
xx55
-4
-9
xx4 0.5
0.5 -1.5
-0.5
11 00
11
-1.5
-0.5
4
x7
1
0
0
0 0
0
x7
1
0
0
0 0
0
Continuing
x1
x2
x3
x4 x5 x 6
Z
-10
57
9
24 0 0
x5 0.5 -5.5 -2.5 9
1 0
x6
0.5
-1.5 -0.5
1 0
1
x7
1
0
0
0
0 0
Back to the initial tableau!
xx7
7
00
00
00
1
1
00
00
00
1
1
x7
0
0
0
1
0
0
0
1


14
One Theorem
Cycle Theorem
Theorem 3.1: If the simplex method fails to terminate, then it
must cycle.
Proof: Consider a canonical form of an LP. Suppose we have n
number of variables and m constraints. Then the total number of
possible simplex iteration will be
n
,
m
which is finite. Therefore, if the simplex method loops infinitely, it
must revisit the same dictionary more than once. Thus, the
simplex algorithm must cycle.

Blands Rule
15
Blands Rules (1977)

Bland (1977) proposed the following two rules:
1.
Entering variable selection: If more than one non-basic

variables are positive (or negative) for maximization (or
minimization) problem, choose the one with the lowest index.
Note that the slack (or surplus) variables have a lower priority
than the main variables;
Blands Rule
BlandsBlands
Rule
Rules (1977)
16
Bland (1977) proposed the following two rules:

1.
Bland
(1977)variable
proposedselection:
the following
twothan
rules:one non-basic
Entering
If more
positive
(or negative)
maximization
(or
variables
Entering are
variable
selection:
If more for
than
one non-basic
minimization)
problem,
the for
onemaximization
with the lowest
variables are positive
(orchoose
negative)
(or index.
Note
that the slack
(or surplus)
variables
have
lower priority
minimization)
problem,
choose the
one with
thea lowest
index.
than
main
Notethe
that
the variables;
slack (or surplus) variables have a lower priority
than
the main variables;
2.
Leaving variable selection: If a tie occurs in M.R.T. test,

choose the variable with the lowest index. Note that the slack
(or surplus) variables have a lower priority than the main
variables. If there is no tie in the M.R.T., just choose the
variable with the minimum ratio.
INDE6372
Arevisit
revisitto
to the
the cycling
cycling example
example
17
Advanced Linear Optimization
Cycling
Example
with
Blands
Rules
Blands rule applied

Blands rule applied
x1
-22
x1
Z
Zx5
-22
-4
x5 x4 0.5
-4
xx74 0.5
1
to the following tableau:

to the following tableau:
x2
x3 x 4 x5 x 6
x2
x3 0x4 0x5 -24x6
93
21
93
21
8
2
00 10 -9-24
8
2
-1.5
-0.5
10 01 1-9
-1.5
-0.5
0
0
0 1 00 0 1
x7
1
0
Update the tableau
x1 x2
Z
0 27
x5
0 -4
x1
1 -3
x7 0
3
0
x3
-1
-2
-1
1
0
x4
44
8
2
-2
0
x5
0
1
0
0
x7
0 x7
00
00
10
0
00
00
10
0
x6
20
-1
2
-2
x7
0
0
0
1
0
0
0
1
iming Peng Department of Industrial Engineering University of Houston

x1
Z
-22
x5
-4
x4 0.5
Bolands rule
x7 works!
1
Update
the tableau
the current
tableau
the current tableau
x1
Z
0x1
x5 Z
00
x1x5 10
x
1
x71 0
Jiming Peng
x2
93
8
-1.5
0
x3
x 4 x5 x 6 x 7
21 Advanced
0 Linear
0 Optimization
-24 0
INDE6372
2
0 1
-9
0
-0.5 1 0
1
0
0
0 0
0
1
018
0
0
1
Cycling Example with Blands Rules

x2 x3 x4 x5 x6
x2 -1
x3 x44
x5 0 x6 20x7
4
27
27
-1
-4
-2 448 0 1 20 -1 0
-4
-2
-3
-1 8 2 1 0 -1 2 0
-3
-1
2 0
2 0
3
1
-2 0 -2
x7
0
00
00
0
1
0
0
0
1
x7 0
3
1
-2 0 -2 1 1
the final tableau
x1 x2 x3 x4 x5 x 6 x 7
Z Engineering
0 30University
0 of42
0 Lecture
18 81 Degeneracy
1
Department of Industrial
Houston
INDE6372:
in Simplex
x5
0 2
0
4
1 -5 2 2
x1
1
0 0
0 0
0 1 1
x3 0
3
1 -2 0 -2 1 1
Optimal solution x1 = 1, x2 = 0, x3 = 1, x4 = 0, optimal
objective value 1!
19
Blands Rules (1977)

Using Blands rules, the simplex method terminates in a
finite number of iterations.
Bland, Robert G. (May 1977). "New finite pivoting rules for the simplex
method". Mathematics of Operations Research 2 (2): 103107. doi:10.1287/
moor.2.2.103. JSTOR 3689647. MR 459599.
Introduction:
I
Mathematical
Notation
A mathematical optimization problem has the following form

min
f0 (x)
s.t.
fi (x) bi , i = 1, , m;
x 0.
Let define
X = {x : fi (x) bi , i = 1, , m; x 0}.
Then we can rewrite the problem as
min f (x).
xX
A solution x X is said to be optimal if

f0 (x ) f0 (x), x X .
INDE6372: Lecture 9 Convexity in LO
20
Convexity
A linear function is convex.
In LP, all functions are convex.
LP is a convex problem.
Convex
Non-Convex
21
ng called a conic combination.
k xk
i, i 0 R
22
t i s sum up to 1, it is called an ane combinatio

Definitions
i, iLet R, be the set

i of=real
1 numbers.
i
alled a
combination.
Aconvex
set is called
affine iff for any two points in the set, the
line through them is contained in the set. For any 2 points
in the set, their affine combinations are in the set itself.

i, i 0 R,
i = 1
i
A set is called convex iff any convex

combination of a
2
bination
of istwo
Rset?itself.
subset
also points
contained in
in the
In n-dimensional space, the positive orthant consists of
all vectors with non-negative coordinates.
ints in the set, the line through them is contained
23
Convexity
Convexity
For LP, all the functions are linear (or ane). For a linear function
f (x), it holds
f (x + y ) = f (x) + f (y ),
+ = 1, , 0.
The point x + y is called a convex combination of x and y .

We say a function f is convex if
f (x + y ) f (x) + f (y ),
+ = 1, , 0.
Observation: a linear function is convex. In LP, all the functions

are convex.

24
Example
Examples of convexity
The univariate function f (x) = x 2 is convex.
The absolute function |x| is convex.
The sum of two convex functions is still convex.
More examples?
f(x)=|x|
And, the sum of 2 convex functions is still convex.
Example Set
Convex
Convex Set
Nonconvex Set

25
26
Ane and Convex Set
Convex and Affine Sets

Consider a convex combination
y = x 1 + x 2 ,
+ = 1, , 0.
We can write
y = x 1 + (1 )x 2 = x 2 + (x 1 x 2 ).
A set S is said to be convex if for any two points x 1 , x 2 S, the
convex combination x 1 + (1 )x 2 S for any [0, 1].
A set S is said to be ane if for any two points x 1 , x 2 S, the
data point x 1 + (1 )x 2 S for any .
27
Affine Space & Subspace

An affine space is a geometric structure that generalizes
certain properties of parallel lines in Euclidean space.

EX: a linear subspace of a vector space that has been
translated away from the origin.
An affine space is nothing more than a vector space
whose origin we try to forget about, by adding translations
to the linear maps.
An affine subspace of a vector space V is a subset
closed under affine combinations of vectors in the space.
28
Subspaces: Linear vs affine

In R3, the origin, lines and planes through the origin and
the whole space are linear subspaces.
In R3,the points, lines and planes in general as well as the
whole space are the affine subspaces.
NOTE: A linear subspace is an affine subspace containing
the origin, or, equivalently, a subspace that is closed

under linear combinations.
29
Transformations: Linear vs affine

A linear transformation is a function that preserves all
linear combinations.
An affine transformation is a function that preserves all
affine combinations.
30
Affine transformation
An affine transformation is any transformation that
preserves
collinearity (i.e., all points lying on a line initially still lie on
a line after transformation) and
ratios of distances (e.g., the midpoint of a line segment
remains the midpoint after transformation).
Note: While an affine transformation preserves proportions on lines, it
does not necessarily preserve angles or lengths. Any triangle can be
transformed into any other by an affine transformation, so all triangles
are affine and, in this sense, affine is a generalization of congruent and
similar.
http://mathworld.wolfram.com/AffineTransformation.html
Ane and Convex Set
31
Affine Set is Convex

Theorem
An ane set is convex.
Example: the straight line aT x = b is ane.
The convex hull of a set S is defined by
Hull(S) = {1 x 1 + + k x k |1 + + k = 1, i
Hull(S) is the smallest convex set containing S.
32
Theorem
Convex
Hull
Example: the straight line aT x = b is ane.
An ane set is convex.
The convex hull of a set S is defined by

Hull(S) = {1 x 1 + + k x k |1 + + k = 1, i 0}.
Example
Hull(S) is the smallest convex set containing S.

Convex Hull
Convex Hull
33
Intersection of 2 Convex Sets

Theorem
The intersection of two convex sets S1 S2 is convex.
Question: How about the union of two convex sets?

34
Convex Set
Convex functions and convex sets

Theorem
Suppose S = {x : fi (x) 0; i = 1; ; m}. If all the functions

fi (x) are convex, then S is convex.
Let us define
Si = {x : fi (x) 0};
because fi (x) is a convex function, for any pair x 1 ; x 2 Si , we have

fi ( x 1 + (1 )x 2 ) fi (x 1 ) + (1 )fi (x 2 ) 0;
[0; 1]:
Therefore Si is a convex set. Because S = Si , thus S is convex

too.

35
So how do we generalize LO?

The linear optimization model can be written in standard
form as
where we express the feasible set as the intersection of
an affine subspace
with
, the non-negative orthant (all vectors with nonnegative coordinates)
36
Non-negative orthant
An orthant is the analogue in n-dimensional Euclidean
space of a quadrant in the plane or an octant in three
dimensions.
In R2, there are 4
orthants (quadrants).
I is the nonnegative orthant.
Cone
37
Non-negative orthant Cone

A set S is called a cone if for every x S and 0, we have
x S.
Example: the set X = {x n : x1 , x2 , , xn 0} is a cone.
Such a cone is also called the nonnegative orthant, denoted by n+ .
The set
S = {x 3 : x12 + x22 x3 }
is a cone, the ice-cream cone!
http://www.convexoptimization.com/wikimization/index.php/Positive_semidefinite_cone
Hyperplane and Halfspace

38
Hyperplane and Halfspace

A hyperplane is a set of the form
{x : aT x = b},
where a n , b Re.
The feasible set in standard LP is the intersection of the
nonnegative orthant and some hyperplanes.
The set {x : aT x b} is called halfspace.

Halfspace in R3

39
Global
Example
vs local max / min

40
41
Local
and Global
Minimum
Local
Minimum
To describe a local minimizer of an optimization problem, we
define the neighborhood of a data point:
N(x, r ) = {y : y x r }.
Consider the convex optimization problem
min f (x)
xX
We say x is a local minimal solution to the problem if there exists

a neighborhood with a small r such that
f (x ) f (x),
x N(X , r ) X .
Theorem
If f (x) is convex and the constrained set X is also convex, then a
local minimal solution to the problem is also a global minimal
Consider the convex optimization problem
42
min f (x)
xX
Global
We say x isMinimum
a local minimal solution to the problem if there exists
a neighborhood with a small r such that
f (x ) f (x),
x N(X , r ) X .
Theorem
If f (x) is convex and the constrained set X is also convex, then a
local minimal solution to the problem is also a global minimal
solution.
43
Proof of the Theorem
Global Minimum: Proof of Theorem
Suppose to the contrary that x is not the global minimal solution, then there exists another y X satisfying
f (y ) < f (x ). Now let us consider the convex combination x + (y x ) for [0, 1]. Since X is
convex, we have
x , y X = x + (y x ) X , [0, 1].
Because f (x) is convex, it follows
f (x
+ (y
x )) (1 )f (x ) + f (y ) f (x ),
and the inequality holds strictly if 0 < 1. Recall that for small , we have
(y
x ) = y
x .
For fixed x , y , we can choose a very small parameter such that

y
x r.
This shows that the point x + (y x ) is in the neighborhood of x , but with a smaller objective value. This
violates our basic assumption that x is a local minimum. In this way, we have derived a contradiction. Thus we
can conclude that x must be a global minimal solution.
44
Farkas Lemma
Only 1 of these statements is true:
A vector is in a given convex cone
There exists a (hyper)plane separating the vector from the
cone.
45
Farkas Lemma
Farkas Lemma
Lemma
The system Ax b has no solutions if and only if there is a y such
that
AT y = 0, y 0, b T y < 0.
Proof: consider the primal LP
max
0
Ax b
and its dual

min
bT y
AT y = 0
y 0.

Proof of Farkas Lemma: Continuation
46
Farkas Lemma: Proof

Clearly the dual is feasible. Thus if the primal is feasible, then the
dual is bounded. This implies that if the primal is infeasible, then
the dual is unbounded. It remains to show that the dual is
unbounded if and only if the relations in the lemma hold (recall the
definition of unbounded LP). This finishes the proof of Farkas
Lemma.

Polyhedra and halfspaces
47
48
The separation theorem
Separation Theorem
A polyhedron is the intersection of a finite number of halfspaces
{x n : Ax b, A mn }
Theorem
Let P and P are two disjoint nonempty polyhedra in n . Then
there exist disjoint halfspaces H and H such that P H, P H.

Proof
of the separation
theorem
Separation
Theorem
49
Proof
Let
P = {x n : Ax b, A mn },
b,
A kn }.
P = {x n : Ax
Since these two polyhedra have no intersections, thus the set

b}
P = {x : Ax b, Ax
has no feasible points. From the Farkas lemma, there exists (y , y)
satisfying the following
AT y + AT y = 0,
, b T y + bT y < 0,
(y , y) 0.
From the second inequality, we have either b T y < 0 or bT y < 0

(or both). Without loss of generality, we assume that b T y < 0.
Because P is nonempty, thus AT y = 0.
Continued

50
Proof of the separation theorem: Cont.
Separation Theorem Proof

Define
H = {x : y T Ax b T y }
H = {x : y T Ax bT y}
Clearly these two sets are halfspaces. Moreover, there is no

intersection between them, otherwise, we have
b T y y T Ax bT y = b T y + bT y 0,
which yields a contradiction.
Because
Next we show P H and P H.
P = {x : Ax b}, y 0, therefore, for every x P, we have
y T Ax b T y which implies x H. Thus P H. The case P H
follows similarly.

51
Separation Theorem Example
Example
Note: The separation theorem does not hold for nonconvex sets.
52
Strict Complementarity
Strict Complementary Conditions

Consider the primal
min
cT x
Ax = b, x 0,
and its dual

max
bT y
AT y + s = c, s 0.
Theorem
If both the primal and its dual are feasible, then there exists an
optimal solution x to the primal problem, and optimal solution
(y , s ) to the dual problem such that x + s > 0.
53
Proof of Complementarity
the Strict Complementary Conditions
Strict
Proof
Now let us consider the following problem
min
xj
Ax = b, x 0
c T x z .
Thus the optimal solution to the above problem is 0. From the

strong duality theory, we know its dual
max
bT y z
AT y c ej , x0 0
has optimal value 0. Here ej is the vector with zero elements
except for 1 at its j-th position. If (
y , = 0) is the optimal
solution, then it is easy to see that y + y is also an optimal
solution to the original dual with a slack variable sj > 0.
If > 0, then it must hold b T y = z , thus the point y/ is an
optimal solution to the original dual with a slack variable sj > 0.
ample
54
Strict Complementarity Example

Which algorithm should be used to solve this problem?
Primal or dual simplex?
min
x1 + 2x2
x1 + x2 2;
2x1 x2 1;
x1 , x2 0.
Which algorithm should we use to solve the above problem?

Answer: Dual simplex, since if we write out the standard form for
the above problem, then we have a solution which is feasible for
the dual problem.
What is the solution to the dual problem?
ample
55

Which algorithm should be used to solve this problem?
Primal or dual simplex?
min
x1 + 2x2
x1 + x2 2;
2x1 x2 1;
x1 , x2 0.

ANSWER: Dual Simplex, because standard form
the above problem, then we have a solution which is feasible for
gives a feasible solution for the dual problem.
the dual problem.
What is the solution to the dual problem?
Example
56

min
x1 + 2x2
x1 + x2 2;
Solution to Example in Standard

2x1 x2 Form
1;
x1 , x2 0.

the above Form
problem, then we have a solution which is feasible for
Standard
x1 + 2x2 + 0x3 + 0x4
the dual problem. min
+ xproblem?
What is the solution to thex1dual
2 + x3 = 2;
2x1 + x2 + x4 = 1;
x1 , x2 , x3 , x4 0.

Solution to the primal problem x1 = 0.5, x2 = 0, x3 = 1.5, x4 = 0!

Solution to the dual problem is
Solution to Example in Standard Form
57

min
x1 + 2x2 + 0x3 + 0x4

x1 + x2 + x3 = 2;
2x1 + x2 + x4 = 1;
x1 , x2 , x3 , x4 0.
Solution to the primal problem x1 = 0.5, x2 = 0, x3 = 1.5, x4 = 0!

Solution to the dual problem is
y1 = 0, y2 = 0.5, s1 = 0, s2 = 2.5, s3 = 0, s4 = 0.5.
Check for the above solutions, we do have x + s > 0!

58
Midterm Exam (30% of Final Grade)

Thursday, October 15, 2015 @ 4pm in CBB 124
Closed book (no books, papers, phones, calculators,
computers). Bring your pens and pencils. All paper will be

provided.
60 minutes.
Chapters 1 6, textbook (Operations Research:
Applications and Algorithms (4th edition). W.L. Winston.
Duxbury Press, 2003. ISBN 978-0534380588.)
Lectures 1 13.
Material covered in Assignment 1 and Quizzes 1 and 2.
See UH BlackBoard for Midterm Exam Review Outline.
GOOD LUCK!
LECTURE 11A
TRANSPORTATION, ASSIGNMENT
AND TRANSSHIPMENT PROBLEMS
Outline
The Transportation Problem
Algorithms
Northwest Corner Method
Minimum Cost Method
Vogels Method
Simplex for Transportation Problem
Sensitivity Analysis
Assignment and Transshipment Problems
Transportation Problem
Find the best way
to deliver product to n demand points
from m supply points
given the variable cost of shipping the product
from one supply point to a demand point
Powerco
Powerco
Powerco has three electric power plants that supply the

electric needs of four cities.
The cost of sending 1 million kwh of electricity from a plant
to a city depends on the distance the electricity must travel.
Powerco LP
TA B L E
Shipping Costs, Supply, and Demand for Powerco

To
From
City 1
City 2
City 3
City 4
Plant 1
Plant 2
Plant 3
$8
$9
$14
45
$6
$12
$9
20
$10
$13
$16
30
$9
$7
$5
30
Demand
(million kwh)
Supply
(million kwh)
35
50
40
1 to the four cities cannot exceed 35 million kwh. Each variable with first
resents a shipment of power from plant 1, so we may express this restric
constraint
x11 ! x12 ! x13 ! x14 " 35
Powerco LP
Decision Variables:
How much electricity is sent from each plant to each
city;
Let Xij = Amount of electricity produced at plant i and
sent to city j
EXAMPLE: X14 = Amount of electricity produced at plant 1
and sent to city 4
Objective Function:
Minimize the total cost of shipping from plants to
cities.
Powerco LP: Objective Function

TA B L E
Shipping Costs, Supply, and Demand for Powerco

To
From
City 1
City 2
City 3
City 4
Plant 1
Plant 2
Plant 3
$8
$9
$14
45
$6
$12
$9
20
$10
$13
$16
30
$9
$7
$5
30
Demand
(million kwh)
Supply
(million kwh)
35
50
40
Minimize Z = 8X11+6X12+10X13+9X14
1 to the four cities
cannot
exceed +13X
35 million+7X
kwh. Each variable with first
+9X
+12X
22
24
resents a shipment of21
power from
plant 23
1, so we
may express this restric
+14X31+9X32+16X33+5X34
constraint
x11 ! x12 ! x13 ! x14 " 35
Identify Supply Constraints

Supply Constraints: each supply point has a
limited production capacity.
X11+X12+X13+X14 35
X21+X22+X23+X24 50
X31+X32+X33+X34 40
Identify Demand Constraints
Demand Constraints: The total electricity

received by each city needs to meet its demand.
X11+X21+X31 45
X12+X22+X32 20
X13+X23+X33 30
X14+X24+X34 30
Powerco Constraints
Supply Constraints: each supply point has a
limited production capacity;

X11+X12+X13+X14 35
X21+X22+X23+X24 50
X31+X32+X33+X34 40
Demand Constraints: The total electricity
received by each city needs to meet its demand;
X11+X21+X31 45
X12+X22+X32 20
PLUS non-negativity
X13+X23+X33 30
constraints on all
X14+X24+X34 30
variables
Powerco LO Model
Min Z = 8X11+6X12+10X13+9X14+9X21+12X22+13X23+7X24
+14X31+9X32+16X33+5X34
S.T.: X11+X12+X13+X14 35
(Supply Constraints)
X21+X22+X23+X24 50
X31+X32+X33+X34 40
X11+X21+X31 45
(Demand Constraints)
X12+X22+X32 20
X13+X23+X33 30
X14+X24+X34 30
Xij 0 (i= 1,2,3; j= 1,2,3,4) (Non-negative Constraints)
IGURE
Powerco Network (with optimal soln.)

Supply points
Demand points
City 1
d1 = 45
City 2
d2 = 20
City 3
d3 = 30
City 4
d4 = 30
x11 = 0
s1 = 35
x14 = 0
Graphical
resentation of
Problem and
timal Solution
x12 = 10
Plant 1
x21 = 45
s1 = 50
x13 = 25
x22 = 0
x23 = 5
Plant 2
x24 = 0
x31 = 0
s1 = 40
Plant 3
x32 = 10
x33 = 0
x34 = 30
s.t.
x11 ! x21 ! x31 ! x34 " 45
(Demand constraints)
364
TA B L E
c11
A Transportation Tableau
c12
Supply
c1n
s1
c21
c22
c2n
Powerco
Transportation
Tableau
Shipping Costs, Supply,
and Demand for Powerco
TA B L E
s2
To
From
Plant 1
Plant 2
Plant 3 d1
Demand
Demand
cm1
(million kwh)
City 1
cm2
$8
$9
$14d2
45
City 2
City 3
$6
$12
$9
20
$10
$13
$16
30
City 4
cmn
$9
$7
dn $5
30
Supply
(million kwh)
sm 35
50
40
Optimal
solution City 1
3
City 2
City 3
City 4
Supply
exceed 35 million kwh. Each variable with first subscript 1 repTransportation Tableau 1 to the four cities cannot
8
6
9
10
for Powerco resents a shipment of power from plant 1, so we may express this restriction by the LP
10
35
25
Plant 1
constraint
TA B L E
9
Plant 2
12
13
x ! x14 " 35
x11 ! x12 !
5 13
45
50
In a similar fashion, we can find constraints that reflect plant 2s and plant 3s capacities.
14
9
16
5
Basic
Because power is supplied by the power plants, each is a supply point.
Analogously, a
Plant 3
40
10
30
constraint that ensures that the total quantity shipped from a plant doesvariables
not exceed plant
capacity is a supply
constraint.
The LP30formulation30of Powercos problem contains the
45
20
Demand
following three supply constraints:
CHAPTER
!x
!x
!x
" 35
11 Assignment,
12 and 13
14 Problems
7 Transportation,
Transshipment
(Plant 1 supply constraint)
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
Powerco Transportation Tableau

s2
Row Plant 1: 10 + 25= 35

Row Plant 2: 45+5
= 50 cm2
cm1
Row Plant 3: 10 + 30 = 40
cmn
sm
d1 = 45
d2
Col City 1: 45
Col City
2: 10 + 10 = 20
Demand
Col City 3: 30 = 30
Optimal
solution
3
TA B L E
Transportation Tableau
for Powerco
dn
City 1
City 2
6
10
Plant 1
9
Plant 2
CHAPTER
9
35
13
7
50
5
9
16
10
45
Supply
25
45
Plant 3
City 4
10
12
14
Demand
City 3
20
5
30
30
7 Transportation, Assignment, and Transshipment Problems
30
40
Basic
variables
Balanced Transportation Problem

If total supply = total demand, the problem is
said to be a balanced transportation problem:
i =m
s
i =1
j =n
j =1
In this case, all constraints MUST be binding.

Powerco is a balanced transportation problem.
Basic feasible solutions are relatively simple to
find, so try to formulate a transportation problem

as balanced.
require 30 units. The company has two warehouses.

Warehouse 1 has 40 units available, and warehouse 2 has 30
units available. The costs of shipping 1 unit from warehouse
to customer are shown in Table 7. There is a penalty for each
unmet customer unit of demand: With customer 1, a penalty
cost of $90 is incurred; with customer 2, $80; and with
customer 3, $110. Formulate a balanced transportation
problem to minimize the sum of shortage and shipping costs.
Exercise 1 (text, p. 371)
2 Referring to Problem 1, suppose that extra units could

A company supplies goods to be
3 customers,
each
require
30
purchased andwho
shipped
to either
warehouse
for a total
units.
cost of $100 per unit and that all customer demand must be
met. Warehouse
Formulate a balanced
problem to
The company has 2 warehouses.
1 has 40transportation
units
minimize the sum of purchasing and shipping costs.
available, and warehouse 2 has 30 units available. The costs of

3 to
A customer
shoe company
forecasts
the in
following
during
shipping 1 unit from warehouse
are
shown
Table demands
7.
the next six months: month 1200; month 2260; month
There is a penalty for each unmet
customer
unit ofmonth
demand:
With
3240;
month 4340;
5190;
month 6150. It
costs is
$7incurred;
to produce with
a paircustomer
of shoes with
labor
customer 1, a penalty cost of $90
2,regular-time
$80;
and $11 with overtime labor (OT). During each month,
and with customer 3, $110. (RT)
regular production is limited to 200 pairs of shoes, and
Formulate a balanced transportation problem to minimize the sum of
shortage and shipping costs.
TA B L E
To
From
Warehouse 1
Warehouse 2
Customer 1
Customer 2
Customer 3
$15
$10
$35
$50
$25
$40
Exercise 1, answer.
Supply
Demand
General Transportation Problem

A set of m supply points from which goods are
shipped. Supply point i can supply at most si units.

A set of n demand points to which the goods are
shipped. Demand point j must receive at least di
units of the goods.
Shipping one unit of the good from supply point i
to demand point j incurs a variable cost of cij.
Powerco example, m # 3, s1 # 35, s2 # 50, and s3 # 40.

i units.
d3 #most
30, sand
d4In#the30.
A set of n demand points to which the good is shipped. Demand point j must receive
3 Each
unit
produced
at supply
i and shipped
point
in
at least
dj units
of the shipped
good. point
In the Powerco
example, to
n #demand
4, d1 # 45,
d2 #j 20
cost of
cij30,
. Inand
thed4 Powerco
example, c12 # 6.
d3 #
# 30.
2
Let
Each unit produced

at supply point i and shipped
to demand point j incurs a variable
General
Transportation
Problem
cost of c . In the Powerco example, c # 6.
3
12
ij
Let
xij # number of units shipped from supply point i to demand p
xij # formulation
number of unitsof
shipped
from supply point
i to demand
then the general
a transportation
problem
is point j
then the generali#m
formulation
of a transportation problem is
j#n
cijxij
!
!
mini#1
! j#1
!c x
min
i#m j#n
ij ij
j!n
s.t.
xij " si
!
j!1
(i ! 1, 2, . . . , m)
(Supply constraints)
(1
i!m
7
s.t. Transportation,
dj ( j !and1,Transshipment
2, . . . , n) Problems
(Demand constraints)
! xij #Assignment,
CHAPTER
CHAPTER
i#1 j#1
i!1
xij # 0 (i ! 1, 2, . . . , m; j ! 1, 2, . . . , n)
f a problem has the constraints given in (1) and is a maximization problem, then it is stil
a transportation problem (see Problem 7 at the end of this section). If
i!m
j!n
! si ! ! dj
Total Supply < Total Demand?

There is NO feasible solution.
One or more of the demands will be unmet, and
there will likely be penalty costs.
See Exercise 1 (previous).
Total Supply > Total Demand?

Solution: Balance the problem by adding a dummy
demand point.
Since shipments to the dummy demand point are
not real, they have a cost of zero.
Solving Transportation Problems
If a set of values for the xijs satisfies all but one of the
constraints of a balanced transportation problem,
then the values for the xijs will automatically satisfy the
other constraint.
So can omit one constraint (usually the first supply
constraint.)
In a balanced transportation problem with m supply
points and n demand points, the cells corresponding to
a set of m + n 1 variables contain no loop if and only
if the m + n 1 variables yield a basic solution.
Solving Transportation Problems

Methods to find a Basic Feasible Solution (BFS):
1.
2.
Minimum Cost Method
3.
Vogels Method
Begin in the upper left

(northwest) corner of the
transportation tableau
Set x11 as large as
possible and min (s1,
d1).

5
6
2
Demand
Supply
Set x11=3 (meaning demand of demand point 1 is

satisfied by supply point 1).
5
6
2
3
2
6
2
Demand
Supply
Check the east and south cells. Still can go east

(meaning supply point 1 still has capacity to fulfill
some demand).
3
2
6
2
X
6
Demand
2
X
Supply
Next, go south.
3
X
6
2
3
2
Demand
Supply
Continue
3
2
3
X
2
1
2
2
3
3
X
X
2
Demand
Supply
Northwest Corner Method BFS

Finally, we will have the following bfs, which is:
x11=3, x12=2, x22=3, x23=2, x24=1, x34=2
2
3
Demand
X
2
Supply
Northwest Corner Method critique

It does not use shipping costs.
Therefore, it can give an initial bfs with a very
high total shipping cost.

Determining an optimal solution may require
several pivots.
Exercise 2
Use the Northwest Corner Method to find a BFS for
Exercise 1.
Supply
Demand
Exercise 2, answer.
Supply
Demand
Minimum Cost Method

Uses shipping costs to find BFS with lower total
costs.
Goal: get optimal solution in fewer pivots than
Northwest Corner method.
Minimum Cost Method Approach

Find the decision variable with the smallest
shipping cost (xij) and give it the largest possible

value min (si , dj)
Continue similar to the Northwest Corner Method:
Cross out row i and column j.
Reduce the supply or demand of the noncrossed-out
row or column by the value of xij.

Then choose the cell with the minimum cost of shipping
from the cells that do not lie in a crossed-out row or
column
Repeat.
Minimum Cost Method

Step 1: Select the cell with minimum cost.
12
10
15
Step 2: Cross-out column 2

2
8
3
12
15
Step 3: Find the new cell with minimum

shipping cost and cross-out row 2
2
8
3
10
15

shipping cost and cross-out row 1
2
8
3
15

shipping cost and cross-out column 1
2
8
3
5
X
10

shipping cost and cross-out column 3
2
8
3
4
X
Step 7: Finally assign 6 to last cell. The bfs is

found as: X11=5, X21=2, X22=8, X31=5, X33=4
and X34=6
2
8
3
4
X
6
X
Exercise 3
Use the Minimum Cost Method to find a BFS for Exercise
1.
Supply
Demand
Vogels Method
Compute a penalty for each row and column.
Penalty = the difference between the two smallest
shipping costs in the row or column.

Identify the row or column with the largest penalty.
Find the first basic variable which has the smallest
shipping cost in that row or column.
Then assign the highest possible value to that variable,
and cross-out the row or column as in the previous
methods.
Compute new penalties and repeat the procedure.
An example for Vogels Method

Step 1: Compute the penalties.
Demand
Column Penalty
15
80
78
15
15-6=9
80-7=73
78-8=70
Supply
Row Penalty
10
7-6=1
15
78-15=63
Step 2: Identify the largest penalty and assign the

highest possible value to the variable.
80
78
5
15
Demand
Column Penalty
15
15-6=9
78-8=70
Supply
Row Penalty
8-6=2
15
78-15=63

7
5
15
Demand
Column Penalty
80
78
15
15-6=9
Supply
Row Penalty
15

6
0
7
5
8
5
15
80
78
Demand
15
Column Penalty
Supply
Row Penalty
15
Step 5: Finally the bfs is found as X11=0, X12=5, X13=5,

and X21=15
6
0
7
5
8
5
15
80
78
15
Demand
Column Penalty
Supply
Row Penalty
Next class: Simplex for TP
Readings
Chapter 7 in Text:
Operations Research: Applications and Algorithms (4th
edition). W.L. Winston. Duxbury Press, 2003. ISBN
978-0534380588.

1
LECTURE 11B
Outline
Algorithms
Minimum Cost Method
Vogels Method
Find the best way
Simplex for the Transportation Problem
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
Powerco
Transportation
Tableau
TA B L E
s2
To
From
Plant 1
Plant 2
Plant 3 d1
Demand
Demand
cm1
(million kwh)
City 1
cm2
$8
$9
$14d2
45
City 2
City 3
$6
$12
$9
20
$10
$13
$16
30
City 4
cmn
$9
$7
dn $5
30
Supply
(million kwh)
sm 35
50
40
Optimal
solution City 1
3
City 2
City 3
City 4
Supply
8
6
9
10
10
35
25
Plant 1
constraint
TA B L E
9
Plant 2
12
13
x ! x14 " 35
x11 ! x12 !
5 13
45
50
14
9
16
5
Basic
Analogously, a
Plant 3
40
10
30
not exceed plant
constraint.
45
20
Demand
CHAPTER
!x
!x
!x
" 35
11 Assignment,
12 and 13
14 Problems
7 Transportation,
Transshipment
Performing the Pivot

Based on the transportation tableau, the following steps
should be performed.
Step 1. Select the variable to enter the basis. (Use a
criterion such as the Northwest Corner Method)
Step 2. Find the unique loop that includes this entering
variable and some of the basic variables.
Step 3. Count only the cells in the loop, starting with 0
(even) at the entering variable. Label them as even cells
or odd cells on the loop away from the entering variable
based on the count.
Definition: Loop
An ordered sequence of at least four different cells is called
a loop if
1. Any two consecutive cells lie in either the same row or
same column
2. No three consecutive cells lie in the same row or
column
3. The last cell in the sequence has a row or column in
common with the first cell in the sequence
Performing the pivot

Step 4.
The variable to leave the basis will be the variable with
the minimum value, , in the odd cells.
Pivot
Odd cell odd cell .
Even cell even cell + .
The variables that are not in the loop remain unchanged.
Performing the pivot

Step 4 Notes.
If =0, the entering variable =0 , and an odd variable that
has a current value of 0 will leave the basis. In this case a
degenerate bfs existed before and will result after the
pivot.
If more than one odd cell in the loop equals , you may
arbitrarily choose one of these odd cells to leave the
basis; again a degenerate bfs will result.
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
Powerco
Transportation
Tableau
TA B L E
s2
To
From
Plant 1
Plant 2
Plant 3 d1
Demand
Demand
cm1
(million kwh)
City 1
cm2
$8
$9
$14d2
45
City 2
City 3
$6
$12
$9
20
$10
$13
$16
30
City 4
cmn
$9
$7
dn $5
30
Supply
(million kwh)
sm 35
50
40
Starting
Tableau City 1
3
City 2
City 3
City 4
Supply
8
6
9
10
10
35
25
Plant 1
constraint
TA B L E
9
Plant 2
12
13
x ! x14 " 35
x11 ! x12 !
5 13
45
50
14
9
16
5
Because power is supplied by the power plants, each is a supply point. Analogously, a
Plant 3
40
10
30
constraint that ensures that the total quantity shipped from a plant does not exceed plant
constraint.
45
20
Demand
CHAPTER
!x
!x
!x
" 35
11 Assignment,
12 and 13
14 Problems
7 Transportation,
Transshipment
364
cm1
cm2
cmn
sm
d1
d2
dn
Powerco: Northwest Corner Method

Demand
Starting
Tableau
3
TA B L E
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
9
35
13
7
50
5
9
16
10
45
Supply
25
45
Plant 3
CHAPTER
City 4
10
12
14
Demand
City 3
20
5
30
30
What is a BFS for this, using NWC method?
30
40
Powerco: Start with NWC
Basic Variables
Powerco: Select enter variable
Basic Variables
What if x14 enters the basis?
Powerco: Evaluate for leaving variable
Powerco: Evaluate for leaving variable
Min ODD cell: x23
Powerco: Select leaving variable
Min ODD cell: x23

LEAVES BASIS
Powerco: Select leaving variable
Min ODD cell: x23

LEAVES BASIS
Powerco: = value of leaving variable
Min ODD cell: x23

LEAVES BASIS.
= 20
Powerco: Modify cells with
+20
ENTER
-20
+20
-20
LEAVE
+20
-20
Powerco: Modify cells with

35-20
10+20
45
0+20
20
20
20-20
(nonbasic)
35
50
10+20
30-20
30
30
40
Powerco: BFS #2 after pivot

35-20
15
10+20
30
45
0+20
20
20
20
20
20-20
0
(nonbasic)
non-basic
35
50
10+20
30
30-20
10
30
30
40
The solution is FEASIBLE.

There is NO LOOP in the solution, so it is also a BFS.
Transportation Pivot Notes
Since each row has as many +s as s, the

new solution will satisfy each supply and demand
constraint.
By choosing the smallest odd variable to leave
the basis, all variables will remain nonnegative.
Each pivot involves only additions or subtractions.
If all the supplies and demands for a
transportation problem are integers, then the
transportation problem will have an optimal
solution in which all the variables are integers.
In this case, the LP Divisibility Assumption

can be ignored.
Exercise 1.
Use the Transportation Simplex to find a BFS for the
warehouse problem from the previous lecture (text, p.
371). Start with the BFS from the NWC Method:
Supply
Demand
out rows and columns we choose the variable to enter

the basis to be the "best" variable in the row or
column with the largest penalty. Of course in a max
problem, "best" variable means the variable with
largest objective function coefficient
Exercise
1,
answer.
SECTION 7.3
1. We begin with the bfs obtained in Section 7.2

v's
15
35
25
u's
15
35
25
0
30
10
40
10
50
40
15
20
10
30
90
80
110
85
20
20
30
30
30
_
Since c32=40 we enter x32 into basis. The loop involving
x32 and some of the basic variables is (3, 2) - (2, 2) (2, 3) - (3, 3). x33 exits yielding the following bfs
v's
15
35
25
10
15
0 15
4025
20
10
30 40
10
90
80
110
10
50
40
20
20
20
10
30
30
30
30
30
85
15
50
35
90
80
110
_
85
20 The loop
20 involving
Since c32=40 we enter x32 into basis.
30the basic 30
x32 and some of
variables 30
is (3, 2) - (2, 2) (2, 3)
_ - (3, 3). x33 exits yielding the following bfs
Exercise 1, answer.
Since c32=40 we enter x32 into basis. The loop involving

x32 and some of the basic variables is (3, 2) - (2, 2) (2, 3) - (3, 3). x33 exits yielding the following bfs
v's
15
35
25
u's
15
0
30
v's
15 10
15
u's
0 45
35
10
35 50
90
15
30
30
10
20
10
25
13
30
2540
40
30
80
35
11025
30 50
30 40
2040
_
0
30
30
Since c21=20 we enter x21 into the basis. The loop
involving x21 and some of the basic variables is
(2,1)-(1,1)-(1,2)-(2,2). After
x22 exits we obtain the following bfs:
15
v's
u's
15
35
45
13
90
45
80
110
20
30
20
30
30
_
Since c21=20 we enter x21 into the basis. The loop
involving x21 and some of the basic variables is
(2,1)-(1,1)-(1,2)-(2,2). After
x22 exits we obtain the following bfs:
Exercise 1, answer.
v's
15
35
45
u's
15
0
30
35
10
40
10
-5
25
50
40
30
90
45
80
30
110
20
30
20
30
30
_
Now c13=20 so we enter x13. The relevant loop is (1, 3) (2, 3) - (2, 1) - (1, 1). After x11 exits we obtain the
following bfs:
v's
-5
35
25
u's
15
35
25
0
10
30
40
10
50
40
15
30
0
30
u's
15
0
30
35
10
40
10
-5
25
50
40
30
90 answer.
80
Exercise
1,
45
20
30
30
30
110
20
30
_
following bfs:
v's
-5
35
25
u's
15
35
25
0
10
30
40
10
50
40
15
30
0
30
90
80
110
45
20
20
30
30
30
This is an optimal tableau. Thus 10 units should be
sent from Warehouse 1 to Customer 2, 30 units from
Warehouse 1 to Customer 3, 30 units from Warehouse 2 to
customer 1. 20 units of Customer 2's demand will be
unsatisfied.
-5
30
90
45
80
30
110
20
30
20
30
30
_
following bfs:
v's
-5
35
25
u's
15
35
25
0
10
30
40
10
50
40
15
30
0
30
90
80
110
45
20
20
30
30
30
Exercise 1, answer.
This is an optimal tableau. Thus 10 units should be

sent from Warehouse 1 to Customer 2, 30 units from
Warehouse 1 to Customer 3, 30 units from Warehouse 2 to
customer 1. 20 units of Customer 2's demand will be
unsatisfied.
Transportation Problem NBV Pricing

Let BV = the set of basic variables in a BFS.
Assume that the first supply constraint in the original LP
has been dropped.
Then the coefficient of the variable Xij (call it ij) in the
tableaus row 0 ) is given by
ij = cBV B-1aij cij
Where cij is the objective function coefficient for xij and aij is
the column for Xij in the original LP minimization problem.
This is a minimization problem.

Note that the first constraint has been dropped.
If all the ijs are 0, then the current bfs is optimal
Otherwise, we enter into the basis the variable with
the most positive ij.
Calculate cBVB-1. It will have m+n-1 elements:

cBVB-1= [u2 u3um v1 v2vn]
u2 u3um = m-1 supply constraints
v1 v2vn = n demand constraints

Each basic variable Xij must have ij=0 in any tableau.
Thus for each of the m+n-1 variables in BV:
cBV B-1aij cij=0
cBVB"1aij " cij ! 0
(4)
For a transportation problem, the equations in (4) are very easy to solve. To illustrate the
olution of (4), we find cBVB"1 for (5), by applying the northwest corner method bfs to
he Powerco problem.
Powerco BFS via NWC Method

11
10
35
35
9
10
12
20
13
50
20
14
16
10
45
20
5
30
30
40
30
For each of the m+n-1 variables in BV:

cBV B-1aij cij=0
HAPTER
(5)
Powerco NBV Pricing

For each of the m+n-1 variables in BV:
cBV B-1aij cij=0
Powerco NWC bfs
BV={X11, X21, X22, X23, X33, X34}.
0
11= [u2 u3 v1 v2 v3 v4] 0

1
Supply | Demand 0

0

0
- 8 = v 1- 8 = 0
1
0

1

0
0

0
1
0

0

1
0

0
Powerco NBV Pricing

21= [u2 u3 v1 v2 v3 v4]
22= [u2 u3 v1 v2 v3 v4]
23= [u2 u3 v1 v2 v3 v4]
1
0

0

0
1

0
-9 = u2+v1-9=0
-12 = u2+v2-12=0
-13 = u2+v3-13=0
Powerco NBV Pricing
33= [u2 u3 v1 v2 v3 v4]
34= [u2 u3 v1 v2 v3 v4]
0
1

0

0
1

0
0
1

0

0
1

0
-16 = u3+v3-16=0
-5 = u3+v4-5=0
Powerco NBV Pricing

For each basic variable Xij (i 1), the previous
equations reduce to
ui+vj=cij
Where ui is supply and vj is demand.
If u1=0, solve following system of m+n equations.
Powerco NBV Pricing

u1=0
u1+v1=8
u2+v1=9
u2+v2=12
u2+v3=13
u3+v3=16
u3+v4=5
By solving the system above we obtain:
v1=8, u2=1, v2=11,v3=12, u3=4, v4=1
Powerco NBV Pricing

For each nonbasic variable, compute
ij = ui+vj cij
Result:
12 = 0+11 6 = 5
14 = 0+1 9 = -8
31 = 4+8 14 = -2
13 = 0+12 10 = 2
24 = 1+1 7 = -5
32 = 4+11 9 = 6
Since 32 = Max ij >0, X32 enters the basis.

Each unit of X32 entered into the basis will decrease
Powercos cost by $6.
cBVB"1aij " cij ! 0
(4)
For a transportation problem, the equations in (4) are very easy to solve. To illustrate the
olution of (4), we find cBVB"1 for (5), by applying the northwest corner method bfs to
he Powerco problem.
Initial Powerco BFS from NWC Method

8
10
35
35
9
10
12
20
7
50
20
14
9
ENTER
45
13
20
16
10
5
30
30
40
30
From previous calculations, select X32 to enter the

basis.
HAPTER
(5)
Powerco NBV Pricing

35
10
35
20
ENTER
45
20
20
50
10
30
30
30
40
Identify the loop with X32 and some of the basic

variables. [ (3,2), (3,3), (2,3), (2,2)]
EVEN ODD
EVEN ODD
Powerco NBV Pricing

35
35
ODD
10
20
ENTER
45
20
20
10
50
30
40
ODD
30
30
Identify the loop with X32 and some of the basic

variables. [ (3,2), (3,3), (2,3), (2,2)]
EVEN ODD
EVEN ODD
MIN (ODD) = 10
PIVOT
13
TA B L E
35
Loop Involving
Entering Variable x32
13
10
35
35
Powerco NBV Pricing
14
9
The resulting
bfs will
be:
12
10
13
20
50
20
16
10
30
40
X11=35, X32=10, X21=10, X22=10, X23=30 and
X34=30
TA B L E
36
x32 Has Entered the Basis,

and x12 Enters Next
88
45
30
20
vvj j "
=
88
11
11
8
uuii ="00
77
10
35
35
10
12
10
!2
2
13
50
16
10
45
30
14
CHAPTER
12
12
9
11
30
5
30
20
30
40
30
Powerco NBV Pricing

The uis and vjs for the new bfs were obtained by
solving
u1=0
u2+v2=12
ui
u3+v4=5
8
6
10
9
0
u1+v1=8
9
12
13
7
1
14
9
16
5
-2
u2+v3=13
8
11
12
7
u2+v1=9
vj
u3+v2=9
Final Solution
Based on the reduced cost obtained in the previous slide,
we can see that X13 should enter the basis. We then

follow a similar process until an optimal solution is found.
The optimal solution for Powerco is X12=10, X13=25,
X21=45, X23=5, X32=10 and X34=30.
As a result of this solution the objective function value
becomes:
Z=6(10)+10(25)+9(45)+13(5)+9(10)+5(30)=$1020
Simplex for TP in Tableau
We now illustrate how the Simplex for TP works in the tableau form.
Take the Powerco problem for example.
S.1: Use a method to find a BFS (NWC method is used in the example)
(35, 8,0)
(10,9,0)
(20,12,0)
(20,13,0)
(10,16,0)
(30, 5,0)
S.2: Calculate the values of the dual variables (start with u1=0)
(35, 8,0)
(10,9,0)
8
0
(20,12,0)
11
(20,13,0)
(10,16,0)
(30, 5,0)
12
Simplex for TP in Tableau: Cont.
S.3: Fill the cells for non-basic variables where Xij=0, and the reduced
cost is defined by -Cij+Ui+Vj.
(35, 8,0)
(0,6,5)
(0,10,2)
(0,9,-8)
(10,9,0)
(20,12,0)
(20,13,0)
(0,7,-5)
(0,14,-2)
(0,9,6)
(10,16,0)
(30, 5,0)
11
12
S.4: Choose a variable (X32) to enter the basis. Find a loop and determine
how to update the variables (=10) in the loop.
(35, 8,0)
(0,6,5)
(0,10,2)
(0,9,-8)
(10,9,0)
(20,12,0)
(20,13,0)
(0,7,-5)
(0,14,-2)
(0,9,6)
(10,16,0)
(30, 5,0)
11
12

After updating the loop, we obtain a new BFS (X32 entered the basis, and
X33 left).
We then repeat steps 2,3,4 and obtain the following tableau.
Based on the reduced cost, X12 should enter the basis. We thus get a loop
consisting of four cells X11, X12, X22, X21.
(35, 8,0)
(0,6,5)
(0,10,2)
(0,9,-2)
(10,9,0)
(10,12,0)
(30,13,0)
(0,7,1)
(0,14,-8)
(10,9,0)
(0,16,-6)
(30, 5,0)
-2
11
12

After updating all the variables in the loop, we get a new BFS (X12
entered the basis and X22 left).
Then we repeat steps 2,3,4 and obtain the following tableau.
Based on the new reduced cost, we see that X13 should enter the basis.
We find a loop and determine the value of =25.
(25, 8,0)
(10,6,0)
(0,10,2)
(0,9,-7)
(20,9,0)
(0,12,-5)
(30,13,0)
(0,7,-4)
(0,14,-7)
(10,9,0)
(0,16,-1)
(30, 5,0)
12

Based on the new BFS , we calculate the dual variables Ui and Vj, and the
reduced cost for NBVs
Then we recalculate the dual variables, and update the reduced cost.
From the following tableau, all the reduced costs are non-positive. Thus
the current solution (X12=10, X13=25, X21=45, X23=5, X32=10 and X34=30)
is optimal.
(0, 8,-2)
(10,6,0)
(25,10,0)
(0,9,-7)
(45,9,0)
(0,12,-4)
(5,13,0)
(0,7,-2)
(0,14,-5)
(10,9,0)
(0,16,-3)
(30, 5,0)
10
Final Solution
The optimal solution for Powerco is X12=10, X13=25,
X21=45, X23=5, X32=10 and X34=30.

As a result of this solution the objective function value
becomes:
Z=6(10)+10(25)+9(45)+13(5)+9(10)+5(30)=$1020
Next Session
Readings
Chapter 7 in Text:
978-0534380588.
Group Project: December 3, 2015 - 15 %

All Short Proposals have been reviewed. Only those
needing rewrites were notified. Resubmissions should be

in by now.
Status Report 1 Due: October 29, 2015 4 pm *
* (Printed copy in class; pdf copies by email to instructor & TA:
mmcguire2@uh.edu | mmcguir2@central.uh.edu &

aida.khayatian@gmail.com)
Please identify the Group Manager in Status Reports. The
Manager is responsible to compile and submit the groups

status reports and results.

1
LECTURE 11C
Outline
Algorithms
Minimum Cost Method
Vogels Method
Find the best way
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
Powerco
Transportation
Tableau
TA B L E
s2
To
From
Plant 1
Plant 2
Plant 3 d1
Demand
Demand
cm1
(million kwh)
City 1
cm2
$8
$9
$14d2
45
City 2
City 3
$6
$12
$9
20
$10
$13
$16
30
City 4
cmn
$9
$7
dn $5
30
Supply
(million kwh)
sm 35
50
40
Optimal
solution City 1
3
City 2
City 3
City 4
Supply
8
6
9
10
10
35
25
Plant 1
constraint
TA B L E
9
Plant 2
12
13
x ! x14 " 35
x11 ! x12 !
5 13
45
50
14
9
16
5
Basic
Analogously, a
Plant 3
40
10
30
not exceed plant
constraint.
45
20
Demand
CHAPTER
!x
!x
!x
" 35
11 Assignment,
12 and 13
14 Problems
7 Transportation,
Transshipment
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
Powerco Final Solution
c2n
s2
cm1
cmn
The optimal solution
forcm2Powerco is X12=10,
X13=25, X21=45,
sm
X23=5, X32=10 and X34=30.
d1
d2
dn
Objective
Demandfunction value:
Z=6(10)+10(25)+9(45)+13(5)+9(10)+5(30)=$1020
Optimal
solution
3
TA B L E
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
CHAPTER
9
35
13
7
50
5
9
16
10
45
Supply
25
45
Plant 3
City 4
10
12
14
Demand
City 3
20
5
30
30
30
40
Basic
variables
TP Sensitivity Analysis
Consider
1. Changing the objective function coefficient of a
non-basic variable.
2. Changing the objective function coefficient of a
basic variable.
3. Increasing a single supply by and a single
demand by .
1. Changing the objective function coefficient

of a nonbasic variable.
Changing the objective function coefficient of a
nonbasic variable Xij will leave the right hand

side of the optimal tableau unchanged.
The current basis will still be feasible.

of a nonbasic variable.
Changing the objective function coefficient of a
nonbasic variable Xij will leave the right hand

side of the optimal tableau unchanged. Thus the
current basis will still be feasible.
Since cBVB-1 is unchanged, the uis and vjs remain
unchanged.
In row 0, only the coefficient of the nonbasic Xij
will change.
Thus as long as the coefficient of Xij in the optimal
row 0 < 0, the current basis remains optimal as
well as feasible.
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
s
1. Powerco: Changing the objective
function
coefficient of a nonbasic variable.
2
c cost
Reduced
m1
cm2
cmn
sm
For what range of values of the cost of shipping 1

d1
d2
dn
million
Demand kwh of electricity from plant 1 to city 1 will
the current basis remain optimal?
Optimal
solution
3
TA B L E
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
CHAPTER
9
35
13
7
50
5
9
16
10
45
Supply
25
45
Plant 3
City 4
10
12
14
Demand
City 3
20
5
30
30
30
40
Basic
variables
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
s
function
2
Let c11 ccm111 + =cm28 +
cmn
11 = u1 + v1 c11 = 0 + 6 (8 + )= -2 - sm
d1
d2
Constraint
(-2 - )
0; -2
dn
Demand
c11 8 + (-2) 6
Optimal
solution
3
TA B L E
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
Demand
CHAPTER
35
7
50
5
9
20
16
5
30
30
13
10
45
Supply
25
45
Plant 3
City 4
10
12
14
City 3
30
40
Basic
variables
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
s
function
2
Let c11 ccm111 + =cm28 +
cmn
11 = u1 + v1 c11 = 0 + 6 (8 + )= -2 - sm
d1
d2
Constraint
(-2 - )
0; -2
dn
Demand
c11 8 + (-2) 6 CURRENT BASIS STILL OPTIMAL

Optimal
solution
TA B L
E 3
for Powerco
C11 6
City 1
City 2
6
10
Plant 1
9
Plant 2
Demand
CHAPTER
35
7
50
5
9
20
16
5
30
30
13
10
45
Supply
25
45
Plant 3
City 4
10
12
14
City 3
30
40
Basic
variables

of a basic variable.
Since cBVB-1 is changed, the coefficient of
each nonbasic variable in row 0 may change.

To determine whether the current basis
remains optimal, find the new uis and vjs
Use the new uis and vjs to price out all
nonbasic variables.
The current basis remains optimal as long as
all nonbasic variables price out < 0.
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
s
function
coefficient of a basic variable.
2
c
For whatc range of
values of thec cost of shipping 1
m1
m2
mn
million kwh of electricity from plant 1s to city 3 will

d
d
d
the Demand
current
basis
remain optimal?
m
Optimal
solution
3
TA B L E
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
Demand
CHAPTER
35
7
50
5
9
20
16
5
30
30
13
10
45
Supply
25
45
Plant 3
City 4
10
12
14
City 3
30
40
Basic
variables
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
s
function
2
c = 10 +
Let c13 c c13 +
m1
cmn
m2
sm
13 = u1 + v3 = 10 +
d1
d2
dn
Demandsolve the equations for u , v

Next,
i
j
Optimal
solution
3
TA B L E
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
Demand
CHAPTER
35
7
50
5
9
20
16
5
30
30
13
10
45
Supply
25
45
Plant 3
City 4
10
12
14
City 3
30
40
Basic
variables
ermine for the Powerco problem the range of values of the cost
from plant 1 to city 3 for which the current basis remains opti
we change
c13 from 10
to 10 $ the
!. Then
the equation
2. Powerco:
Changing
objective
function
!c13 " 0 c
0 to u1 coefficient
$ v3 " 10 of
$ a!.basic
Thus, variable.
to find the uis and vjs, we must
ions:
u2 $ u1 " 0
u3 $ v2 " 90 $ !
u2 $ v1 " 9
u1 $ v2 " 6
u2 $ v3 " 13
u1 $ v3 " 10 $ !
u3 $ v4 " 50 $ !
u1 $ v3 " 10 $ !
vers topics that may be omitted with no loss of continuity.
ansportation, Assignment, and Transshipment Problems
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
s
function
2
c = 10 +
Let c13 c c13 +
m1
cmn
m2
sm
13 = u1 + v3 = 10 +
d1
d2
dn
Demand
Solution:
u1=0, v2=6, v3=10+, v1=6+, u2= 3-,
u3=3, v4=2
Optimal
solution
3
TA B L E
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
Demand
CHAPTER
35
7
50
5
9
20
16
5
30
30
13
10
45
Supply
25
45
Plant 3
City 4
10
12
14
City 3
30
40
Basic
variables
2. Powerco: Changing the objective function

Solution: u1=0, v2=6, v3=10+, v1=6+, u2= 3-,
u3=3, v4=2
Next, price out reduced costs for each nonbasic
variable.
The current basis will remain optimal as long as
each nonbasic variable has a coefficient <0 in row
0.
9
Plant 2
12
45
7
50
5
14
Plant 3
13
9
10
16
5
30
40
2. Powerco: Changing the objective function

Demand
30
30
coefficient of45a basic20 variable.
Solution:
u1=0,wevobtain
=6+,
Solving
these equations,
0, v2 ! 6, v3 v
!110
" #, v1 u
!26="3-,
#, u2 ! 3 $
2=6,u1v!3=10+,
#, u3 ! 3, and v4 ! 2.
uWe
v4=2out each nonbasic variable. The current basis will remain optimal as long
3=3,
now price
as each nonbasic variable has a nonpositive coefficient in row 0.
c!11 ! u1 " v1 $ 8 ! # $ 2 % 0
c!14 ! u1 " v4 $ 9 ! $7
c!22 ! u2 " v2 $ 12 ! $3 $ # % 0
c!24 ! u2 " v4 $ 7 ! $2 $ # % 0
c!31 ! u3 " v1 $ 14 ! $5 " # % 0
for # % 2$
for # & $3
for # & $2
for # % 5$
c!33 ! u3 " v3 $ 16 ! # $ 3 % 0
for # % 3$
Thus, the current basis remains optimal for $2 % # % 2, or 8 ! 10 $ 2 % c13 % 10 "
2 ! 12.
Increasing Both Supply si and Demand dj by !

Observe that this change maintains a balanced transportation problem. Because the uis
and vjs may be thought of as the negative of each constraints shadow prices, we know
364
as each nonbasic variable has a nonpositive coefficient in row 0.

TA B L E
c11
c12
c!11 ! u1 " v1 $ 8 ! # $ 2 % 0
c!14 ! u1 " v4 $ 9 ! $7
c
c
c!2221 ! u2 " v222$ 12 ! $3 $ # % 0
c1n
Supply
for # % 2$
s1
c2n
for # & $3
s
2. Powerco:
Changing
the
objective
function
c! ! u " v $ 7 ! $2 $ # % 0
for # & $2
c! !of
u a
" vbasic
$ 14 ! $5
"#%0
for # % 5$
coefficient
variable.
24
31
c!33 ! u3 " v3 $ 16 ! # $ 3 % 0
for # % 3$
cm1
cm2
cmn
Thus, the current basis remains optimal for $2 % # % 2, or 8 !
10 $ 2 % c13 % 10 "
sm
2 ! 12.
d1
d2
dn
Demand
Optimal
solution
withdcurrent
basis for range:
Increasing Both
Supply
si andstays
Demand
j by !
8 c13 12
Observe
this changeCitymaintains
aCitybalanced
transportation
problem.
Because
the uis
Optimal
solution
3 that
1
2
City 3
City
4
Supply
andTableau
vjs may be thought of as8 the negative
of each constraints
shadow
prices, we know
Transportation
6
9
10
for Powerco
from
(37')Plant
of1 Chapter 6 that if the current
basis 25
remains optimal,
10
35
TA B L E
New 9z-value ! old

z-value "13#ui " #vj
12
50 (new cost)
45 plant 1s supply and city
5 2s demand by 1 unit, then
Plant 2if we increase
For example,
! 1,020 " 1(0) " 1(6) ! $1,026.
14
9
16
5
Basic
We may
also
find
the
new
values
of
the
decision
variables
as
follows:
Plant 3
40
10
30
variables
1 If xij is a basic variable in the optimal solution, then increase xij by #.
Demand
45
20
30
30
If xij is a nonbasic variable in the optimal solution, then find the loop involving xij and
some of the basic variables. Find an odd cell in the loop that is in row i. Increase the value
C H A P T E R 7 Transportation, Assignment, and Transshipment Problems
of this odd
cell by # and go around the loop, alternately increasing and then decreasing
2
3. Increasing Supply Si and Demand Dj by

Changing both supply and demand by the same amount
will maintain the balance of the transportation problem.

uis and vjs may be thought of as the negative of each
constraints shadow price.
If the current basis remains optimal
New Z value = old Z value + ui + vj
Powerco: Increase plant 1s supply & city 2s demand by 1
unit, then
New cost=1020+1(0)+1(6)=$1026
3. Increasing Supply Si and Demand Dj by

Finding new values of the decision variables:
1. If Xij is a basic variable in the optimal solution,
increase Xij by .
2. If Xij is a nonbasic variable in the optimal
solution, find the loop involving Xij and some of
the basic variables.
Find an odd cell in the loop that is in row i.
Increase the value of this odd cell by and go around the loop,
alternately increasing & decreasing current basic variables in the

loop by .
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n
3. Powerco: Increasing Supply Ssi and

Demand Dj by
2
Basic variable:
X12 cm2
cm1
cmn
sm
S1 S1 + 2
1
D2 D2d+
2
d2
dn
Demand
Original
TA B L
E 3
Optimal
solution
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
CHAPTER
9
35
13
7
50
5
9
16
10
45
Supply
25
45
Plant 3
City 4
10
12
14
Demand
City 3
20
5
30
30
30
40
Basic
variables
3. Powerco: Increasing Supply Si and

Demand Dj by
Basic variable: X12
S1 S1 + 2
D2 D2 + 2
Z = 1020 + 2u1 + 2v2 = 1032
TA B L E
40
New
al Tableau for PowercoOptimal

If
! 35 " 2 ! 37 and
d2 ! 20 " 2 ! 22
Plant 1
solution
vj !
City 1
City 2
City 3
City 4
10
8
uui =
! 00
6
12
33
45
TA B L E
41
al Tableau for Powerco If

! 35 " 1 ! 36 and
vj !
13
7
50
5
9
16
10
Demand
37
12
14
Plant 3
25
9
Plant 2
10
Supply
5
30
40
45
22
30
30
City 1
City 2
City 3
City 4
10
Supply
Basic
variables
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n

Demand Dj by
2
Nonbasiccm1
variable:cm2X11
cmn
sm
S1 S1 + 1
1
D1 D1d+
1
d2
dn
Demand
Find Loop with X11

Original
TA B L
E 3
Optimal
solution
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
CHAPTER
9
35
13
7
50
5
9
16
10
45
Supply
25
45
Plant 3
City 4
10
12
14
Demand
City 3
20
5
30
30
30
40
Basic
variables
364
TA B L E
c11
c12
Supply
c1n
s1
c21
c22
c2n

Demand Dj by
2
Have Loopcm1with X11cm2
cmn
sm
Row 1 odd cell in Loop is X13

d1
d2
Calculate
new optimal
solution: dn
Demand
X13 X13+ 1; X21 X21+1; X23 X23 1
Original
TA B L
E 3
Optimal
solution
for Powerco
City 1
City 2
6
10
Plant 1
9
Plant 2
45
Supply
9
35
X13
13
7
50
5
9
16
10
Plant 3
CHAPTER
25
45
City 4
10
12
14
Demand
City 3
20
5
30
30
30
40
Basic
variables
TA B L E
40
Tableau for Powerco If

35 " 2 ! 37 and
! 20 " 2 ! 22
vj !
City 1
City 2
City 3
City 4
10
Supply
8
6
10
9
3. Powerco:
Increasing
Supply
S
and
i
u =0
12
25
37
Demand Dj by
9
12
13
7
Plant 1
ui ! 0
33
Plant 2
Calculate
new45optimal solution 5with current basis:50
5
X13 X13+ 1; X21 14X21+1; X239 X23 116
Plant 3
10
30
40
Demand Z = 1020+ 45
30
New
u1 + v1 =221026 (note:
typo30in text)
TA B L E
41New
Tableau for Powerco IfOptimal solution

vj !
35 " 1 ! 36 and
! 45 " 1 ! 46
Plant 1
City 1
City 2
City 3
City 4
10
uuii =!00
6
10
33
Demand
36
12
46
13
7
50
4
14
Plant 3
26
9
Plant 2
10
Supply
33
16
10
46
5
30
20
30
40
Basic
variables
30
basic variables. The loop is (1, 1)(1, 3)(2, 3)(2, 1). The odd cell in the loop and row
1 is x13. Thus, the new optimal solution will be obtained by increasing both x13 and x21
le.
ch the
ch the
he new
he new
medical
to the
e given
Exercise: Sensitivity Analysis

a The companys goal is to minimize the cost of meet Two plants
supply demands.
three customers
with medical
ing customers
Find two optimal
bfs for supplies.
this
transportation
problem. from the plants to the
The unit
costs of shipping
b Suppose
that
customer
2s demand
by one
customers,
along
with
the supplies
andincreased
demands,
are
unit. By how much would costs increase?
given below:
TA B L E
42
To
From
Plant 1
Plant 2
Demand
Customer 1
Customer 2
Customer 3
Supply
$55
$10
10
$65
$15
10
$80
$25
10
35
50
signment, and Transshipment Problems
le.
ch the
ch the
he new
he new
medical
to the
e given
Exercise: Sensitivity Analysis

a The companys goal is to minimize the cost of meet A. The
goal is toFind
minimize
the cost
of meeting
ingcompanys
customers demands.
two optimal
bfs for
this
transportation
problem.
customers
demands.
Find two optimal bfs for this
b Supposeproblem.
that customer 2s demand increased by one
transportation
B. Suppose that customer 2s demand increased by one

TA B L E
42
To
From
Plant 1
Plant 2
Demand
Customer 1
Customer 2
Customer 3
Supply
$55
$10
10
$65
$15
10
$80
$25
10
35
50
signment, and Transshipment Problems
Exercise: Sensitivity Analysis Answer

A. The companys goal is to minimize the cost of meeting
customers demands. Find two
18 optimal bfs for this
2(3) - 2(10) =problem.
994.
transportation
5a. Using the transportation simplex we found the

following optimal tableau (z = 950)
OPTIMAL SOLUTION 1
55
10
65
80
10
10
15
10
10
25
10
10
20
20
10
10
The degeneracy of this solution allows us to pivot in

x13 and find a second optimal solution:
OPTIMAL SOLUTION 2
5a. Using the transportation simplex we found the

following optimal tableau (z = 950)
OPTIMAL SOLUTION 1
Exercise: Sensitivity
Analysis
Answer
55
65
80
0
10
10
20
A. The companys goal

minimize25the cost0of meeting
10 is to 15
customers demands. Find

bfs for this
10 two optimal
10
20
10
10
10
10
transportation problem.
The degeneracy of this solution allows us to pivot in

OPTIMAL SOLUTION 2
55
65
10
80
0
10
15
10
10
0
10
25
10
10
20
20
10
10
If we increase customer 2's demand by ! OPTIMAL

SOLUTION 1 becomes

OPTIMAL SOLUTION 2
55
65
80
10
0
10 Answer
20
Exercise: Sensitivity
Analysis
10
15
25
0
10
10
20
B. Suppose that customer

2s demand
by one
10
10
10 increased
10

SOLUTION 1 becomes
55
10
65
80
!
10
10-!
15
10
10+!
10
25
20
0
10
20
10
10-!
Setting ! = 1 yields new optimal solution x11 = 10, x12

=1, x14 = 9, x22 = x23 = 10, z = 1015.
If we try to increase customer 2's demand by !, then
OPTIMAL SOLUTION 2 becomes
55
65
80

SOLUTION 1 becomes
55
10
65
80
0
10-!
20
10
15
25
0
Exercise:
Sensitivity
Analysis
Answer
10
10
20
10
10+!
10
10-!
B. Suppose that customer 2s demand increased by one

Setting
! = 1much
yields
new costs
optimal
solution
unit.
By how
would
increase?
x11 = 10, x12
=1, x14 = 9, x22 = x23 = 10, z = 1015.
If we try to increase customer 2's demand by !, then

OPTIMAL SOLUTION 2 becomes
55
65
10
80
!
10
10
10-!
15
10+!
10+!
25
10-!
10
20
0
20
10-!
19
For !>0, this tableau is
no longer optimal (x12 must
enter the basis!). This is because the problem is no
longer degenerate, and pivoting in x12 will now decrease
z.
SECTION 7.5
Assignment Problems
An assignment problem is a balanced
transportation problem in which all supplies and

demands are equal to 1.
The Cost Matrix = cost of assigning each supply
point to each demand point.
All supplies and demands are integer.
Since RHS = 1, Xij = 0 or 1.
An assignment problem are a special case of
transportation problem for which the transportation
simplex is often very inefficient. Solutions have a
high degree of degeneracy.
Instead, alternate algorithms are used.
Machineco: Assignment Problem

Machineco needs to assign 4 jobs to 4 machines.
Each machine does 1 job (no switching).
Setup time for each machine varies, depending
on the job.
Minimize the total setup time.
Machineco: Assignment Problem

Setup times
(Also called the cost matrix)
Time (Hours)
Job1
Job2
Job3
Job4
Machine 1
14
Machine 2
12
Machine 3
Machine 4
10
Machineco: Assignment Problem Model

min Z = 14 X 11 + 5 X 12 + 8 X 13 + 7 X 14 + 2 X 21 + 12 X 22 + 6 X 23 + 5 X 24
+ 7 X 31 + 8 X 32 + 3 X 33 + 9 X 34 + 2 X 41 + X 42 + 6 X 43 + 10 X 44
s.t. X 11 + X 12 + X 13 + X 14 = 1
X 21 + X 22 + X 23 + X 24 = 1
X 31 + X 32 + X 33 + X 34 = 1
X 41 + X 42 + X 43 + X 44 = 1
X 11 + X 21 + X 31 + X 41 = 1
X 12 + X 22 + X 32 + X 42 = 1
X 13 + X 23 + X 33 + X 43 = 1
X 14 + X 24 + X 34 + X 44 = 1
Xij = 0orXij = 1
Machineco: Assignment Problem Model

Xij=1 if machine i is assigned to meet the
demands of job j
Xij=0 if machine i is not assigned to meet the
demands of job j
Assignment Problem: Hungarian Method

Step1.
Find a bfs.
Find the minimum element in each row of the m x m cost
matrix.
Construct a new matrix by subtracting from each cost the
minimum cost in its row.
For this new matrix, find the minimum cost in each
column.
Construct a new matrix (reduced cost matrix) by
subtracting from each cost the minimum cost in its
column.
Assignment Problem: Hungarian Method

Step2. Draw the minimum number of lines (horizontal and/or
vertical) that are needed to cover all the zeros in the reduced
cost matrix.
If m lines are required, an optimal solution is available
among the covered zeros in the matrix. If fewer than m lines
are required, proceed to step 3.
Step3. Find the smallest nonzero element (call its value k)
in the reduced cost matrix that is uncovered by the lines drawn
in step 2.
Now subtract k from each uncovered element of the reduced
cost matrix and add k to each element that is covered by two
lines.
Return to Step 2.
Assignment Problem: Machineco Hungarian Method

Time (Hours)
Job1
Job2
Job3
Job4
Machine 1
14
Machine 2
12
Machine 3
Machine 4
10
Cost matrix: 4 x 4
(m x m)
1. Row min
14
12
10
1a. Row min
14
12
10
1a. Subtract row i minimum from each cell in row i .

9
10
1b. Column min
1b. Subtract column j minimum from each cell in

column j.

9
10
# lines = 3 < 4, so
continue.
2. Draw lines to
cover all the zeroes.
10
2. Draw lines to
cover all the zeroes.
Smallest, uncovered
nonzero element
k=1
3a. Subtract k=1 from each uncovered element;

add k=1 to each twice covered element.
10=9+1
5=4+1

Job j
TA B L E
49
Four Lines Required; Optimal

Solution Is Available
10
Machine i
3b. 4 lines now required to cover all zeroes.

Thus,
we havesolution
found the
optimal assignment x12 ! 1, x24 ! 1, x33 ! 1, and x41 ! 1.
optimal
exists.
Of course, this agrees with the result obtained by the transportation simplex.
Intuitive Justification of the Hungarian Method
To give an intuitive explanation of why the Hungarian algorithm works, we need to discuss the following result: If a constant is added to each cost in a row (or column) of a

Job j
TA B L E
49

10
X12 = 1
Machine i
X33 = 1
4
5
3c.
X33 = 1, since it is the only covered 0 in col 3.
Thus,
we have found the optimal assignment x12 ! 1, x24 ! 1, x33 ! 1, and x41 ! 1.
X12 = 1,
it is
thethe
only
covered
0 in
Of course,
thissince
agrees
with
result
obtained
by col
the 2.
transportation simplex.

Job j
TA B L E
49

10
X12 = 1
0
X24 = 1
Machine i
0
X33 = 1
4
5
3c.
Row 1 & Col 2 complete.
Thus, we have found the optimal assignment x12 ! 1, x24 ! 1, x33 ! 1, and x41 ! 1.
Col 4 covered
= X24the
. result obtained by the transportation simplex.
Of course,
this agrees0 with
Set X24 = 1.

Job j
TA B L E
49

10
X12 = 1
0
X24 = 1
Machine i
0
X33 = 1
4
5
3d.
X24 = 1.
Thus,
we have found the optimal assignment x12 ! 1, x24 ! 1, x33 ! 1, and x41 ! 1.
Row 2this
& Col
4 complete.
Of course,
agrees
with the result obtained by the transportation simplex.

Job j
TA B L E
49

10
X12 = 1
0
X24 = 1
Machine i
X41 = 1
0
X33 = 1
4
5
3e.
Select X41 = 1.
Thus, we have
found the optimal assignment x12 ! 1, x24 ! 1, x33 ! 1, and x41 ! 1.
Of course, this agrees with the result obtained by the transportation simplex.

Job j
TA B L E
49

10
X12 = 1
0
X24 = 1
Machine i
X41 = 1
0
X33 = 1
4
5
3f. OPTIMAL ASSIGNMENT of machine i to job j.

X12 = 1.
Thus, we have found the optimal assignment x12 ! 1, x24 ! 1, x33 ! 1, and x41 ! 1.
X24course,
= 1. this agrees with the result obtained by the transportation simplex.
Of
X33 = 1.
X41 = 1.
h 1..10). Then edit line 3 to indicate that there are 8 jobs. Finally, in line 12,
pe the 80 entries of your cost matrix, following COST! and you would
oll!
iscussion of the Machineco example, it is unnecessary to force the ASSIGN(I,J) to

his will happen automatically!
Exercise: Hungarian Method
5 employees are available to perform 4 jobs.

The time it takes each person to perform each job is given below.
Determine the assignment of employees to jobs that minimizes the total
time required to perform the four jobs.

Hints: use a cost of M to rule out forbidden assignments and add a
dummy job (Job 5) to balance the problem. Assigning a person to the
rm four jobs.dummy
The job Thas
A BaL0Ecost.
50
ch job is given in
Time (hours)
employees to jobs
Person
Job 1
Job 2
Job 3
Job 4
perform the four
r a relay team for

swim 100 meters
or freestyle. Doc
he times given in
1
2
3
4
5
22
18
26
16
21
18
20
22
30
27
28
25
18
22
28
14
28
Note: Dashes indicate person cannot do that particular job.
balance the problem. Assigning a person to the

dummy job has a 0 cost.
Exercise: Hungarian Method - Answer

1
22
Job
2
18
3
30
4
18
5
0
Row Min
0
18
27
22
26
20
28
28
16
22
14
21
25
28
1
2
Person
3
4
5
Column Min
16
18
25
14
Since all Row minima are 0 we proceed to the column

minima. The reduced cost matrix is
Job
1
1
6
2
0
3
5
4
4
5
0
balance the problem. Assigning a person to the

dummy job has a 0 cost.

1
22
Job
2
18
3
30
4
18
5
0
Row Min
0
18
27
22
26
20
28
28
16
22
14
21
25
28
1
2
Person
3
4
5
Column Min
16
18
25
14

Job
1
1
6
2
0
3
5
4
4
5
0
5
Column Min
16
18
25
14


Job
Person
1
6
2
0
3
5
4
4
5
0
10
14
4
20
14
Only 4 lines (we have used Row 1, Row 4, Col. 3 and

Reduced
Matrix,all
after0's
subtracting
Col. 5) are
neededCost
to cover
in this column
matrix.minima.
The
Only 4 lines
are needed
to we
cover
all 0's in
smallest uncovered
element
is 2 so
subtract
2 this
frommatrix.
all uncovered
and uncovered
add 2 to all
twice
covered
Thecosts
smallest
element
is 2
so we subtract 2
costs. The resulting
matrix
is
from all uncovered costs and add 2 to all twice covered
Job
costs.
1
2
3
4
5
1
6
0
7
4
2
Person
12
12
20
5
14

Col. 5) are needed to cover all 0's in this matrix. The
smallest uncovered element is 2 so we subtract 2 from
all uncovered costs and add 2 to all twice covered
costs. The resulting matrix is
Job
1
2
3
4
5
1
6
0
7
4
2
Person
12
12
5 lines are needed to cover the 0's so an optimal

Resulting Reduced Cost Matrix
solution is available: x12=1, x21=1, x35=1, x44=1, x53=1.
53lines
nowassigned
required atojob
cover
optimal
Thus Person
is not
andalla zeroes.
total time
of
18+18+14+25=75
timeexists.
units is required to complete all
solution
the jobs.
2. Applying the Hungarian Method yields
Free Breast Fly
Back
3
3
0
2
Hall
0
6
1
1
Spitz
0
3
4
6
Montgomery
20
5
14

Col. 5) are needed to cover all 0's in this matrix. The
smallest uncovered element is 2 so we subtract 2 from
all uncovered costs and add 2 to all twice covered
costs. The resulting matrix is
Dummy Job 5
Job
1
2
3
4
5
1
6
7
4
2
X0 = 1

12
Person
X210= 1
12
X350= 1
X530= 1
0= 1
X44
12
5 lines are needed to cover the 0's so an optimal

Resulting Reduced Cost Matrix
solution is available: x12=1, x21=1, x35=1, x44=1, x53=1.
53lines
nowassigned
required atojob
cover
optimal
Thus Person
is not
andalla zeroes.
total time
of
18+18+14+25=75
timeexists.
units is required to complete all
solution
the jobs. Optimal soln: x =1, x =1, x =1, x =1, x =1.
12
21
35
Person 3 is not assigned a job
2. Applying the Hungarian Method yields

Total
time Breast
= 18+18+14+25=75
Free
Fly
Back
complete
all the
3
3 jobs.
0
2
Hall
0
6
1
1
Spitz
0
3
4
6
Montgomery
44
53
time units is required to
Transshipment Problem
Transportation Problem setup: ONLY DIRECT
supply point demand point.

Transshipment Problem: allows INTERMEDIATE
transshipment points.
Transshipment Problem: can be formulated as a
Transportation Problem.
Convert Transshipment Problem to TP

Step 1. If necessary, add a dummy demand point
to balance the problem. This will have a supply of

0 & a demand = the problems excess supply.
Shipments to the dummy and from a point to itself
= 0.
Let s= total available supply.

Step 2a.
Construct a transportation tableau:
Set up a row in the tableau for each supply point
& for each transshipment point.

Set up a column for each demand point & for
each transshipment point.

Step 2b.
Set up each supply point = original supply.
Set up demand point = original demand.

Step 2c.
Let s = total available supply.
For each transshipment point:
supply = (points original supply) + s
demand =(points original demand) + s
For a transshipment point that is a net supplier:
net outflow = points original supply

For a transshipment point that is a net demander:
net inflow = points original demand
Total amount through transshipment points s.
by solving a transportation problem. Given a transshipment problem, we create a balanced

transportation problem by the following procedure (assume that total supply exceeds total demand):
If necessary, add a dummy demand point (with a supply of 0 and a demand equal
to the problems excess supply) to balance the problem. Shipments to the dummy and from
a point to itself will, of course, have a zero shipping cost. Let s ! total available supply.
Step 1
Transshipment: Widgetco
Construct a transportation tableau as follows: A row in the tableau will be needed

for each supply point and transshipment point, and a column will be needed for each demand point and transshipment point. Each supply point will have a supply equal to its
original supply, and each demand point will have a demand equal to its original demand.
Let s ! total available supply. Then each transshipment point will have a supply equal to
(points original supply) " s and a demand equal to (points original demand) " s. This
ensures that any transshipment point that is a net supplier will have a net outflow equal
to the points original supply, and, similarly, a net demander will have a net inflow equal
to the points original demand. Although we dont know how much will be shipped
through each transshipment point, we can be sure that the total amount will not exceed s.
This explains why we add s to the supply and demand at each transshipment point. By
adding the same amounts to the supply and demand, we ensure that the net outflow at each
transshipment point will be correct, and we also maintain a balanced transportation
tableau.
Step 2
Widgetco manufactures widgets at two factories:
Memphis & Denver.

Memphis factory produces 150 widgets/day
Denver factory produces 200 widgets/day
Widgets are shipped by air to customers in LA and
Boston 130 widgets/day to each location.
Select the air routing to minimize total cost.
FIGURE
Memphis
New York
Los
Angeles
Denver
Chicago
Boston
A Transshipment
Problem
7. 6 Transshipment Problems
401

Step 1

Memphis factory
150 widgets/day
Letproduces
s ! total available
supply. Then each transshipment point will have a supply equal to
(points original
supply)
" s and a demand equal to (points original demand) " s. This
Denver factory produces
200
widgets/day
Total supply s =ensures
350 that any transshipment point that is a net supplier will have a net outflow equal
Widgets are shipped
by air original
to customers
LA andwe
Boston
130how
widgets/day
each
to the points
demand.inAlthough
dont know
much will betoshipped
location.
Total demand dThis
= 260
explains why we add s to the supply and demand at each transshipment point. By
adding
the d
same=amounts
to the supply and demand, we ensure that the net outflow at each
Dummy Demand
Point:
s-d = 90
dp
To convert to TP, add s=350 to each transshipment points supply & demand.
tableau.
Step 2
Ignore shipments to dummy & from a point to itself.
FIGURE
Memphis
New York
Los
Angeles
Denver
Chicago
Boston
A Transshipment
Problem
401
by solvinghas
a transportation
Given a transshipment
we create
a balanced
York and Chicago)
a net outflowproblem.
of 0; whatever
flows into theproblem,
transshipment
point
problem
by A
thegraphical
followingrepresentation
procedure (assume
total solution
supply exceeds
tomust leavetransportation
the transshipment
point.
of thethat
optimal
to
tal demand):
the Widgetco
example is given in Figure 10.
SupposeStep
that1 weIf modify
theadd
Widgetco
example
allow
necessary,
a dummy
demandand
point
(withshipments
a supply ofbetween
0 and a Memdemand equal
phis and Denver.
This
would
make
Memphis
and
Denver
transshipment
points
and
would
add columns
for Memphis
and of
Denver
the Table
tableau.cost.
TheLet
Memphis
row
in the supply.
a point
to itself will,
course,tohave
a zero59
shipping
s ! total
available
tableau would now have a supply of 150 " 350 ! 500, and the Denver row would have
Step 2 Construct a transportation tableau as follows: A row in the tableau will be needed
for each supply point and transshipment point, and a column will be needed for each demandN.Y.point and transshipment
point.
Each supply
point will
have a supply
equal to its
Chicago
L.A.
Boston
Dummy
Supply
8
13
25
28
0
20
130
150
Memphis
15
12
26
25
0
70how much will
200 be shipped
to the points original demand. Although 130
we dont know
Denver
0
6
16
17
0
This
explains
why
we
add
s
to
the
supply
and
demand
at
each
transshipment
220
130
350 point. By
N.Y.
6
0
14
16
0
tableau.
350
350
TA B L E
59
Representation of
as Balanced
Chicago
Demand
Memphis
FIGUR
E 9
A10Transshipment
Optimal Solution
Problem
Denver
to Widgetco
350
Memphis
130
350
New York
130
New York
130
Los
Angeles
Chicago
Boston
FIGURE
Denver
Chicago
130
90
Los
Angeles
Boston
130
402
CHAPTER

401
York and Chicago) has a net outflow of 0; whatever flows into the transshipment point
Denver. This
make representation
Memphis andofDenver
transshipment
must leavephis
the and
transshipment
point.would
A graphical
the optimal
solution to points and would
the Widgetco
given
in Figureand
10. Denver to the Table 59 tableau. The Memphis row in the
add example
columnsisfor
Memphis
Suppose
that we
modify
thehave
Widgetco
example
and "
allow
between
Mem- row would have
tableau
would
now
a supply
of 150
350shipments
! 500, and
the Denver
phis and Denver. This would make Memphis and Denver transshipment points and would
add columns for Memphis and Denver to the Table 59 tableau. The Memphis row in the
N.Y.
Chicago
L.A.
Boston
Dummy
Supply
T A B L E 59
Representation of
8
13
25
28
0
Problem MemphisN.Y.
130Chicago
150
L.A.
Boston
Dummy
Supply20
T A B Transshipment
L E 59
Representation of as Balanced
8
13
25
28
0
15
12
26 20
25 150
0
Transshipment
Problem Memphis
Transportation
Problem
130
as Balanced
70
130
200
Denver
15
12
26
25
0
130
0
6
16 70
17 200
0
Denver
220
130
350
N.Y.
220
N.Y.
Chicago
Chicago
Demand
Memphis
FE
I G10
URE
FIGUR
402
Demand
350
Memphis
130
10
Optimal Solution
Optimal Solution
Denver
to Widgetco
to Widgetco
402
CHAPTER
Denver
6
6
16
130
350
14
14
16
16
350
350
130
0
350
350
350
130
130
130
130
New
130 York
Los
Angeles
Chicago
Chicago
Boston
130
130
New York
350
350

CHAPTER
17
130
90
Los
Angeles
Boston
90
must leave the transshipment point. A graphical representation of the optimal solution to
the Widgetco example is given in Figure 10.
Suppose that we modify the Widgetco example and allow shipments between Memphis and Denver. This would make Memphis and Denver transshipment points and would
TA B L E
59
N.Y.
Chicago
L.A.
Boston
Dummy
Supply
Representation of
8
13
25
28
0
20 tableau and
130example, this procedure yields the transportation
150its optiFor Memphis
the Widgetco
as Balanced
mal solution given in Table
59. Because
s ! (total
supply) !25150 " 2000! 350 and (total
15
12
26
demand) ! 130 " 130 ! 260, the dummy demand point

has a demand
of 350200
# 260 !
70
130
90. TheDenver
other supplies and demands in the transportation tableau are obtained by adding
0
6 supply and
16 demand. 17
0
s ! 350 to each transshipment
points
220 solution to the transportation
130
350
In interpreting
the
problem created from a transshipment
N.Y.
problem, we simply ignore the shipments to the dummy and from a point to itself. From
6
0
14
16
0
Table 59, we find that Widgetco
should
produce
130
widgets
at
Memphis,
ship
350
350them to
Chicago
New York, and transship them from New York to Los Angeles. The 130 widgets produced
at Denver should be350
shipped directly
to Boston.
outflow from
350
130 The net
130
90 each city is
Demand
Memphis
FIGURE
10
Optimal Solution
to Widgetco
402
Denver
Memphis:
220 " 130 " 20 " 220 ! 150
Denver:
220 " 130 " 70 " 220 ! 200
Los
130
130
N.Y.: New York220 " 130 # 130
# 220 ! 0
Angeles
Chicago:
350 # 350 # 130 # 220 ! 0
Chicago 350 # 350 #130
Boston
L.A.:
# 220 ! 0
Boston: 130 350 # 350 #130 # 220 ! 0
Dummy:
#20 # 70 # 130 # 220 ! #90
ddp = s-d = 90
C H A P T net
E R 7outflow
Transportation,
Assignment, and
Problems that each transshipment point (New
A negative
represents
anTransshipment
inflow. Observe
Transportation Problem Limitations

TP generally handles only one product supplied
and demanded at multiple locations unless

special conditions are met.
Difficult to generalize the technique for situations
with nonlinear & concave objective functions, e.g.:
Economy of scale, when the per-unit cost of transportation on a
link decreases with volume.

Fixed + variable transportation. For example, truck rental =
fixed charge + mileage charge.
Readings
Chapter 7 in Text:
978-0534380588.

All Short Proposals have been reviewed. Only those
needing rewrites were notified. Resubmissions should be

in by now.
Status Report 1 Due: October 29, 2015 4 pm *



1
LECTURE 11D
& DEALING WITH PROJECT DATA
Outline
Algorithms
Minimum Cost Method
Vogels Method
Find the best way
Transportation Problem setup: ONLY DIRECT
supply point demand point.

Transshipment Problem: allows INTERMEDIATE
transshipment points.
Transshipment Problem: can be formulated as a
Transportation Problem.

Step 1. If necessary, add a dummy demand point
to balance the problem. This will have a supply of

0 & a demand = the problems excess supply.
Shipments to the dummy and from a point to itself
= 0.
Let s= total available supply.

Step 2a.
Construct a transportation tableau:
Set up a row in the tableau for each supply point
& for each transshipment point.

Set up a column for each demand point & for
each transshipment point.

Step 2b.
Set up each supply point = original supply.
Set up demand point = original demand.

Step 2c.
Let s = total available supply.
For each transshipment point:
supply = (points original supply) + s
demand =(points original demand) + s
For a transshipment point that is a net supplier:
net outflow = points original supply

For a transshipment point that is a net demander:
net inflow = points original demand
Total amount through transshipment points s.

Step 1

to the points original demand. Although we dont know how much will be shipped
This explains why we add s to the supply and demand at each transshipment point. By
tableau.
Step 2
Widgetco manufactures widgets at two factories:
Memphis & Denver.

Memphis factory produces 150 widgets/day
Denver factory produces 200 widgets/day
Widgets are shipped by air to customers in LA and
Boston 130 widgets/day to each location.
Select the air routing to minimize total cost.
FIGURE
Memphis
New York
Los
Angeles
Denver
Chicago
Boston
A Transshipment
Problem
401

Step 1

Memphis factory
150 widgets/day
Letproduces
s ! total available
supply. Then each transshipment point will have a supply equal to
(points original
supply)
" s and a demand equal to (points original demand) " s. This
Denver factory produces
200
widgets/day
Total supply s =ensures
350 that any transshipment point that is a net supplier will have a net outflow equal
Widgets are shipped
by air original
to customers
LA andwe
Boston
130how
widgets/day
each
to the points
demand.inAlthough
dont know
much will betoshipped
location.
Total demand dThis
= 260
explains why we add s to the supply and demand at each transshipment point. By
adding
the d
same=amounts
to the supply and demand, we ensure that the net outflow at each
Dummy Demand
Point:
s-d = 90
dp
To convert to TP, add s=350 to each transshipment points supply & demand.
tableau.
Step 2
Ignore shipments to dummy & from a point to itself.
FIGURE
Memphis
New York
Los
Angeles
Denver
Chicago
Boston
A Transshipment
Problem
401
by solvinghas
a transportation
Given a transshipment
we create
a balanced
York and Chicago)
a net outflowproblem.
of 0; whatever
flows into theproblem,
transshipment
point
problem
by A
thegraphical
followingrepresentation
procedure (assume
total solution
supply exceeds
tomust leavetransportation
the transshipment
point.
of thethat
optimal
to
tal demand):
the Widgetco
example is given in Figure 10.
SupposeStep
that1 weIf modify
theadd
Widgetco
example
allow
necessary,
a dummy
demandand
point
(withshipments
a supply ofbetween
0 and a Memdemand equal
phis and Denver.
This
would
make
Memphis
and
Denver
transshipment
points
and
would
add columns
for Memphis
and of
Denver
the Table
tableau.cost.
TheLet
Memphis
row
in the supply.
a point
to itself will,
course,tohave
a zero59
shipping
s ! total
available
Step 2 Construct a transportation tableau as follows: A row in the tableau will be needed
for each supply point and transshipment point, and a column will be needed for each demandN.Y.point and transshipment
point.
Each supply
point will
have a supply
equal to its
Chicago
L.A.
Boston
Dummy
Supply
8
13
25
28
0
20
130
150
Memphis
15
12
26
25
0
70how much will
200 be shipped
to the points original demand. Although 130
we dont know
Denver
0
6
16
17
0
This
explains
why
we
add
s
to
the
supply
and
demand
at
each
transshipment
220
130
350 point. By
N.Y.
6
0
14
16
0
tableau.
350
350
TA B L E
59
Representation of
as Balanced
Chicago
Demand
Memphis
FIGUR
E 9
A10Transshipment
Optimal Solution
Problem
Denver
to Widgetco
350
Memphis
130
350
New York
130
New York
130
Los
Angeles
Chicago
Boston
FIGURE
Denver
Chicago
130
90
Los
Angeles
Boston
130
402
CHAPTER

401
Denver. This
make representation
Memphis andofDenver
transshipment
must leavephis
the and
transshipment
point.would
A graphical
the optimal
solution to points and would
the Widgetco
given
in Figureand
10. Denver to the Table 59 tableau. The Memphis row in the
add example
columnsisfor
Memphis
Suppose
that we
modify
thehave
Widgetco
example
and "
allow
between
Mem- row would have
tableau
would
now
a supply
of 150
350shipments
! 500, and
the Denver
phis and Denver. This would make Memphis and Denver transshipment points and would
N.Y.
Chicago
L.A.
Boston
Dummy
Supply
T A B L E 59
Representation of
8
13
25
28
0
Problem MemphisN.Y.
130Chicago
150
L.A.
Boston
Dummy
Supply20
T A B Transshipment
L E 59
Representation of as Balanced
8
13
25
28
0
15
12
26 20
25 150
0
Transshipment
Problem Memphis
Transportation
Problem
130
as Balanced
70
130
200
Denver
15
12
26
25
0
130
0
6
16 70
17 200
0
Denver
220
130
350
N.Y.
220
N.Y.
Chicago
Chicago
Demand
Memphis
FE
I G10
URE
FIGUR
402
Demand
350
Memphis
130
10
Optimal Solution
Optimal Solution
Denver
to Widgetco
to Widgetco
402
CHAPTER
Denver
6
6
16
130
350
14
14
16
16
350
350
130
0
350
350
350
130
130
130
130
New
130 York
Los
Angeles
Chicago
Chicago
Boston
130
130
New York
350
350

CHAPTER
17
130
90
Los
Angeles
Boston
90
must leave the transshipment point. A graphical representation of the optimal solution to
the Widgetco example is given in Figure 10.
Suppose that we modify the Widgetco example and allow shipments between Memphis and Denver. This would make Memphis and Denver transshipment points and would
TA B L E
59
N.Y.
Chicago
L.A.
Boston
Dummy
Supply
Representation of
8
13
25
28
0
20 tableau and
130example, this procedure yields the transportation
150its optiFor Memphis
the Widgetco
as Balanced
mal solution given in Table
59. Because
s ! (total
supply) !25150 " 2000! 350 and (total
15
12
26
demand) ! 130 " 130 ! 260, the dummy demand point

has a demand
of 350200
# 260 !
70
130
90. TheDenver
other supplies and demands in the transportation tableau are obtained by adding
0
6 supply and
16 demand. 17
0
s ! 350 to each transshipment
points
220 solution to the transportation
130
350
In interpreting
the
problem created from a transshipment
N.Y.
problem, we simply ignore the shipments to the dummy and from a point to itself. From
6
0
14
16
0
Table 59, we find that Widgetco
should
produce
130
widgets
at
Memphis,
ship
350
350them to
Chicago
New York, and transship them from New York to Los Angeles. The 130 widgets produced
at Denver should be350
shipped directly
to Boston.
outflow from
350
130 The net
130
90 each city is
Demand
Memphis
FIGURE
10
Optimal Solution
to Widgetco
402
Denver
Memphis:
220 " 130 " 20 " 220 ! 150
Denver:
220 " 130 " 70 " 220 ! 200
Los
130
130
N.Y.: New York220 " 130 # 130
# 220 ! 0
Angeles
Chicago:
350 # 350 # 130 # 220 ! 0
Chicago 350 # 350 #130
Boston
L.A.:
# 220 ! 0
Boston: 130 350 # 350 #130 # 220 ! 0
Dummy:
#20 # 70 # 130 # 220 ! #90
ddp = s-d = 90
C H A P T net
E R 7outflow
Transportation,
Assignment, and
Problems that each transshipment point (New
A negative
represents
anTransshipment
inflow. Observe
Transshipment: Exercise 1
Sunco Oil produces oil at two wells.
Well 1 produces 150,000 barrels/day.
Oil can be shipped directly from the wells to Suncos customers in
Los Angeles and New York.

Alternatively, Sunco could transport oil to the ports of Mobile and
Galveston & then ship it by tanker to New York or Los Angeles.
Los Angeles requires 160,000 barrels per day.
New York requires 140,000 barrels per day.
The costs of shipping 1,000 barrels between two points are shown in
the next table.
Formulate a transshipment model (and equivalent transportation
model) that could be used to minimize the transport costs in meeting
the oil demands of Los Angeles and New York.
ONLY CREATE THE TABLEAU DO NOT SOLVE!
00 ! 350 " 550. The new Memphis column would have a demand of 0 !
and the new Denver column would have a demand of 0 ! 350 " 350. Fie that shipments between demand points L.A. and Boston were allowed. This
L.A. and Boston transshipment points and add rows for L.A. and Boston.
or both the L.A. and Boston rows would be 0 ! 350 " 350. The demand for
. and Boston columns would now be 130 ! 350 " 480.
Transshipment: Exercise 1
Sunco Oil produces oil at two wells.

Los Angeles requires 160,000 barrels per day.
New York requires 140,000 barrels per day.
nd Detroit and
has shipTvia
A B Mobile
L E 61 or Galveston or direct.
Can
supplies cars to
To ($)
st of shipping a car
From
Well 1
Well 2
Mobile
Galveston
N.Y.
means that a
oduce as many as
Well 1
0
10
13
25
many as 2,900 cars.
Well 2
0
15
12
26
ampa must receive
Mobile
0
6
16
Galveston
6
0
14
ation problem that
g costs incurred in
N.Y.
0
mpa.
L.A.
15
shipments between
shipments between
cost of $5.
Note: Dashes indicate shipments that are not allowed.
L.A.
28
25
17
16
15
0
Transshipment: Exercise 1 Answer

(Supplies and demands are in thousands)
Total SupplyTotal Demand=350300=50
Dummy demand pt. has demand
of 50.
29
The balanced transportation tableau is:

Mobile
10
Galv.
13
NY
25
LA
28
Dummy
0
Well 1
150
15
12
26
25
Well 2
200
0
16
17
14
16
15
Mobile
0+350=350
Galv.
0+350=350
NY
0+350=350
M
LA
Demands
Supplies
15
0
0+350
350
350
140
+350
160
+350
50
Assignment #2: Transshipment

2.1 Solve Exercise 1, using the balanced transportation
tableau from the previous slide.

2.2 Redo Exercise 1. Assume that before being shipped to Los
Angeles or New York, all oil produced at the wells must be
refined at either Galveston or Mobile. To refine 1,000 barrels of
oil costs $12 at Mobile and $10 at Galveston. Assuming that
both Mobile and Galveston have infinite refinery capacity,
formulate AND SOLVE a transshipment and balanced
transportation model to minimize the daily cost of transporting
and refining the oil requirements of Los Angeles and New York.
Assignment #2 Due Thursday November 5, 2015 4 pm *

You may work solo or in pairs. Remember academic honesty!
Transportation Problem Limitations

TP generally handles only one product supplied
and demanded at multiple locations unless

special conditions are met.
Difficult to generalize the technique for situations
with nonlinear & concave objective functions, e.g.:
Economy of scale, when the per-unit cost of transportation on a
link decreases with volume.

Fixed + variable transportation. For example, truck rental =
fixed charge + mileage charge.
Dealing with Project Data
Real Life Data Preparation

WHO
WHAT
WHEN
WHERE
WHY
HOW HOW MUCH?
Real Life Data Preparation
Notes on Real Life Data Preparation

WHO will make a decision based on the analysis
WHAT
INPUT: What data can be supplied; data ranges & format
INPUT: What data types/units (integer/fraction, etc.)
OUTPUT: What data and information will result
WHEN ..Are data over time? One sample?
WHERE where are data from? How good is source?
WHY ..is the decision based on this data important? To
whom?
HOW will data be supplied/stored/processed
(spreadsheet, database, pdf) INPUT & OUTPUT
HOW MUCH? Data, budget,
1. Collect your data

If at beginning of project, set the parameters for new
(prospective) data collection
Which cases should be included? Excluded?

For quantitative, how good is the measuring device? Data is only
as good as the least sensitive measurement device

Make assumptions clear from the start
If using existing (retrospective) data, document the
assumptions for data collection

Look for patterns in your data
Look for data ranges
Watch for missing data
KEEP BACKUPS OF ALL DATA, FROM ORIGINAL TO
FINAL. KEEP TIME-STAMPED AUDIT TRAIL.
2. Select your model(s)

Which model is best to help support the decision that can
be made?
If there are limited models to choose from (such as using
LP models), is there some way to frame the question or
organize the data so that an existing model can be used?
KEEP BACKUPS OF ALL MODELS, FROM ORIGINAL
TO FINAL. KEEP TIME-STAMPED AUDIT TRAIL.
3. Prepare the Data

A. Clean the data
B. Create necessary variables
C. Format the data
3A. Clean the data

Find & eliminate raw data errors such as:
Impossible or otherwise incorrect values for specific
variables
Cases in the data that met exclusion criteria and shouldnt
be in the study
Duplicate cases
Missing data and outliers (identify or replace)
Skip-pattern or logic breakdowns
3B. Create New Variables

Select and/or create the variables that will support the
decision
Dimensionality Reduction aggregate data as needed
to simplify the model. EX. Instead of daily quantities,
group together data from every month or quarter for
analysis.
This may require some preliminary statistical analysis. For
more than 300 points, can do mean or average; for
fewer, use median (half above & half below).
Other examples:
Creating change scores & indices from scales, combining too-
small-to-use categories of nominal variables, centering predictors.
3C. Format the Variables

Set all missing data codes so missing data are treated as
such( ie, NaN or M in some LP models)

Format date variables as dates, numerical variables as
numbers, etc.
Clearly label all variables and categorical values.
4. Prototype the model with the data

Run a limited model with a subset of the data
Try alternate ways of framing the decision
Finalize the model assumptions and the full data set
parameters.
5. Run Final Model with Complete Data

Run model
Do sensitivity analyses
6. Interpret Model for Decision Maker

Describe the results in terms that the decision maker
understands
1 picture = 1,000 words!
Start with Executive Summary (100-200 words)
Then review Who, What, Whenetc. and how the final
model and data supported the resulting descision.
Using Excel to Clean and Prepare Data

for Analysis
See pdf in Lecture 20 Folder, UH Blackboard.
RJ Weiss & RJ Townsend, Using Excel to Clean and Prepare Data
for Analysis. The Industrial-Organizational Psychologist, 42(3),
2005, 89-96.
Describes a phased approach to data preparation:
1. Create the data file.
2. Clean the data.
3. Process the data.
4. Create an analysis-ready copy of the data.
5. Document the data.
Readings
Chapter 7 in Text:
978-0534380588.

Status Report 1 Due TODAY: October 29, 2015 4 pm *

UH Grad School Research Day 10/30/15
Date: Friday, October 30th
Time: 10:45 a.m. 6:00 p.m.

Location: Student Center (SC) Ballroom, Theatre, and
Houston Room
RSVP at http://www.uh.edu/graduate-school/events/
research-day/
LECTURE 12A
NETWORK MODELS
1
.
Outline
Basic introduction to network models/applications
Graph, nodes, arcs
Shortest path problem, maximum flow problem
Algorithms
Applications in production control and project
management
2
.
What is a Network?
In general, a network is a collection of items and the
relationships among those items.

The relationships may be 1-1, 1- many, or many-to-many.
The relationships may be 1-directional, multidirectional, or
non-directional.
The relationships may have a weight or probability
associated with them.
The items may have a weight or probability associated
with them.
Although networks are usually visualized as graphs in 2D,
they can be constructed in nD mathematically.
3
.
Networks are Everywhere

Physical Man made Networks
Road Networks
Railway Networks
Airline Traffic Networks
Electrical Networks, e.g., the power grid
4
.

Abstract Networks
organizational charts
precedence
relationships in
projects
social networks
www.bonkersworld.net/
organizational-charts/

Networks in Nature
6
.
Overview
Networks and graphs are powerful modeling tools
Most OR models have networks or graphs as a major
aspect of the data representation of items and their

relationships
Data representation influences how one thinks about how
to solve problems
EX. a row/column spreadsheet encourages thinking in 2D
Major purposes of data representation

efficiency in algorithms
ease of use
7
.
Network Analysis & LP

Several Major Basic Classes of Network Problems
How to recognize and formulate them? What are the features they
can be used to model? What are their limitations?
All can be formulated as an LP
Several important ad-hoc algorithms and their rationales will be
provided
Most efficient transportation of goods, information etc.
through a network
Transportation problem (already discussed)
Transshipment problem
8
.

1. Shortest path problems
2. Maximum flow problems
3. CPM-PERT project scheduling models
4. Minimum cost network flow problems
5. Minimum spanning tree problems
9
.

LP: What is the most efficient way to travel from one point
to another in a physical distance network? OR

LP: What is the optimal way to go from one point to
another in an abstract network? EX: a cost network
representing production, inventory, etc.?
Network Analysis: find the shortest path
Limitations: algorithm can sometimes breakdown when side
constraints are added, but can reformulate in LP
10
.
Basic Definitions
A graph or network is defined by two sets of symbols:
Nodes: A set of points or vertices (call it V) are called
nodes of a graph or network.
Nodes
1
Arcs: An arc consists of an ordered pair of vertices and
represents a possible direction of motion that may occur

between vertices.
Arc
1
11
.
Basic Definitions
Chain: A sequence of arcs such that every arc has exactly
one vertex in common with the previous arc is called a

chain.
Common vertex
between two arcs
1
12
.
Basic Definitions
Path: A path is a chain in which the terminal node of each
arc is identical to the initial node of next arc.

For example in the figure below (1,2)-(2,3)-(4,3) is a chain
but not a path; (1,2)-(2,3)-(3,4) is a chain and a path,
which represents a way to travel from node 1 to node 4.
1
13
.
Network Analysis:
Dijkstras Shortest Path Algorithm
In each iteration, the shortest path from the origin to one
of the rest of the nodes is found.

One new solved node is found in each iteration.
More than one such path and node may be found in one iteration
when there is a tie. There may also exist multiple shortest paths
from the origin to some nodes.
The algorithm stops when the shortest path to the
destination is found.
Assumption: all arc lengths > 0.
14
.

For EACH iteration
Let S be the current set of solved nodes (the set of nodes whose
shortest paths from the origin been found).

Let N be the set of all nodes, and N S be the set of unsolved
nodes.
15
.

1. Consider only the neighbors of S that can be
reachable via one direct link or arc.

2. For each of these neighboring nodes, find the shortest
path from the origin via only S and the corresponding
distance from the origin.
3. In general, there exist multiple such neighboring nodes.
Select the one with the shortest distance from the origin.
Call this new node solved node and add to S.
4. Repeat iteration over all nodes.
16
.
Dijkstras Shortest Path Algorithm Details

Goal of the nth iteration: Find the nth nearest node to the
origin.
Repeat for n = 1, 2, until the nth nearest node is the
destination.
17
.

Input for the nth Iteration:
(n 1) nearest nodes to the origin (solved for at the
previous iterations).
Their shortest path and distance from the origin.
These nodes plus the origin will be called solved nodes
S; the others are unsolved nodes.
18
.

Candidates for the nth nearest node:
Each solved node that is directly connected by a link to
one or more unsolved nodes provides one candidate.
Select the unsolved node with the shortest connecting
link.
Ties provide additional candidate nodes.
19
.

Calculation of nth nearest node:
For each solved node and its candidate, add the distance
between them and the distance of the shortest path from
the origin to this solved node.
The candidate with the smallest total distance is the nth
nearest node. Ties provide additional candidates.
20
.
EX. Road System for Seervada Park

Cars are not allowed into the park
There is a narrow winding road system for trams and for
jeeps driven by the park rangers

The road system is shown without curves in the next slide
Location O is the entrance into the park
Other letters designate the locations of the ranger stations
The scenic wonder is at location T
The numbers give the distance of these winding roads in miles
The park management wishes to determine which route
from the park entrance O to station T has the shortest

total distance for the operation of the trams.
21
.
7
A
5
D
2
4
1
3
4
E
4
C
22
.

Iteration Zero: Shortest distance from node O to Node O.
S = {O}.
Iteration 1:
Step 1: Neighboring Nodes = {A, B, C}
Step 2: Shortest path from O to neighboring nodes that traverse
through the current set of solved nodes S. Min {2,4,5} = 2

(corresponding to node A).
Step 3: The shortest path from O to A has been found with a
distance of 2. S = {O, A}
23
.
A
Solved Nodes
O
C
24
.

Iteration 2:
Step 1: Neighboring nodes = {B, C, D}
Step 2: Min (Min (2 + 2, 5), 4, (2 + 7)) = 4.
Step 3: Shortest path from B and C has been found. S = {O, A, B,
C}
7
A
Current Solved Nodes
(2)
(0)
2
B
C
O
4
25
.

Iteration 3:
Step 1: Neighboring nodes = {D, E}. Only AD, BD, BE, and CE
Step 2: Min(Min(2 + 7, 4+4), Min(4 + 3, 4+4)) = 7
Step 3: The shortest path to E has been found S = {O, A, B, C, E}
Current Solved
Nodes
A
(2)
B
O
(4)
E
(4)
26
.

Iteration 4
Step 1: Include (only) Nodes D and T. Include only arcs AD, BD,
ED, & ET
Step 2: Min((min(2+7, 4+4, 7+1), (7+7))) = 8
Step 3: Shortest path from node O to Node D has been found. S =
{O, A, B, C, D, E}
7
(2)
O
(4)
27
Current solved nodes

(4)
D
4
1
E
(7)
C
.

Iteration 5
Step 1: Include only node T and include arcs DT and ET
Step 2: Min(8+5, 7+7) = 13 (no other competing nodes)
Step 3: The shortest path from the origin to T, the destination, has
been found, with a distance of 13
A (2)
Current solved nodes
(8)
O
28
(0)
(4)
T
7
(4)
(7)
.

Final Solution
Incidentally, we have also found the nth nearest node from
the origin sequentially
(2)
(8)
D
4
O
(0)
B
4
T
1
(4)
(13)
E
(7)
29
(4)

n
Solved Nodes
Directly
Connected to
Unsolved Nodes
Closest
Connected
Unsolved
Node
Total
Distance
Involved
Nth
Nearest
Node
Minimum
Distance
Last
Connection
OA
2, 3
O
A
C
B
4
2+2=4
C
B
4
4
OC
AB
A
B
C
D
E
E
2+7=9
4+3=7
4+4=8
BE
A
B
E
D
D
D
2+7=9
4+4=8
7+1=8
D
D
8
8
BD
ED
D
E
T
T
8+5=13
7+7=14
13
DT
6
30
Dijkstras
Algorithm:
All
Shortest
Paths

To find the shortest path from a node to all other nodes, we
divide the problem into several stages, where in stage i, the
shortest path consists of only a path involves at most i arcs.
For example, consider the seervada park example. In stage 1,
we have a vector of shortest distance
[0
The red elements are permanent

31
.
Dijkstras Algorithm
The formula for distance in stage i+1
d(s,t, i+1)=minv [ d(s,v,i)+d(v,t)]
Stage 2: Shortest distance of paths with at most 2
arcs in the path
[0
Stage 3: paths with at most 3 arcs

[0
7 14 ]
32
.
Dijkstras Algorithm
The formula for distance in stage i+1
d(s,t, i+1)=minv [ d(s,v,i)+d(v,t)]
Final Stage 4: paths with at most 4 arcs
[0
7 13 ]
33
.
4 Use Dijkstras algorithm to find the shortest path

node 1 to node 4 in Figure 5. Why does Dijkstras algor
fail to obtain the correct answer?
Exercise: Dijkstras Algorithm

Find the
path4from node 1 to node 5.
F Ishortest
GURE
Network for Problem 2

2
1
8
12
2
5
4
6
418
5
10
CHAPTER
34
.
8 Network Models
4 Use Dijkstras algorithm to find the shortest path

node 1 to node 4 in Figure 5. Why does Dijkstras algor
Exercise Answer: Dijkstras Algorithm

The shortest
F I Gpath
U R from
E 4node 1 to node 5:
125,Network
length 14.
for Problem 2
2
1
8
12
2
5
4
6
418
5
10
CHAPTER
35
.
8 Network Models
4 Use Dijkstras
algorithm
to find
the shortest
path from
Node
Temporary Label
(* denotes
next
assigned
permanent
node 1 to node 4 in Figure 5. Why does Dijkstras algorithm
label)
3
min{7} = 7*
5
min{14,6+10} = 14
Now labels are [0* 2* 7* 6* 14]. Since there is no node
joining the newest permanently
F I G U R E 4 labelled node (node 3) to node
5, we may give make node 5 a permanent label. We obtain [0*
Network for Problem 2
2* 7* 6* 14*]. Since c25 = 14 - 2 and c12 = 2 - 0 we find the
shortest path from node 1 to
14).
2 5 to be 1-2-5(length
12
2
1
Exercise as Transshipment Problem

1
418
Node 1
Node 2
Node 3
Node 4
10
3.
8 Network Models
Node 5
Supply
M
1
CHAPTER
Node 2
2
Node 3
8
Node 4
12
10
Demand
36
.
4. We begin by giving node 1 a permanent label [0* 2 1 !].

Next node 3 obtains a permanent label: [0* 2 1*!]. Node 4 now
5 Suppo
annual op
shown in
determine
of owning
Network Analysis:
Maximum Flow Problem
Description
All flow through a directed and connected network originates
at one node (source) and terminates at one another node
(sink).
All the remaining nodes are transshipment nodes.
Flow through an arc is allowed only in the direction indicated
by the arrowhead
The maximum flow = the capacity of that arc.
At the source, all arcs point away from the node.
At the sink, all arcs point into the node
Objective: maximize the total amount of flow from the
source to the sink
(measured as the amount leaving the source or the amount
entering the sink)
37
.
Maximum Flow Problem

Typical applications:
Maximize the flow through a companys distribution network from
its factories to its customers

Maximize the flow through a companys supply network from its
vendors to its factories
Maximize the flow of oil through a system of pipelines
Maximize the flow of water through a system of aqueducts
Maximize the flow of vehicles through a transportation network
38
.
Maximum Flow Algorithm

Some Terminology
The residual network shows the remaining arc capacities for
assigning additional flows after some flows have been assigned to
the arcs
The residual capacity for assigning some flow from node B to node O
The residual capacity for flow from node O to Node B

39
.

An augmenting path is a directed path from the source to the sink
in the residual network such that every arc on this path has strictly
positive residual capacity
The residual capacity of the augmenting path is the minimum of
these residual capacities (the amount of flow that can feasibly be
added to the entire path)
Basic idea
Repeatedly select some augmenting path and add a flow equal to
its residual capacity to that path in the original network.
This process continues until there are no more augmenting paths,
so that the flow from the source to the sink cannot be increased
further
40
.

The Augmenting Path Algorithm
Assume that the arc capacities are either integers or rational
numbers (numbers that can be written as fractions a/b)
1. identify an augmenting path by finding some directed
path from the source to the sink in the residual network

such that every arc on this path has strictly positive
residual capacity. If no such path exists, the net flows
already assigned constitute an optimal flow pattern.
2. Identify the residual capacity c* of this augmenting
path by finding the minimum of the residual capacities of
the arcs on this path. Increase the flow in this path by c*
41
.
EX. Maximum Flow for Seervada Park

During the peak season the park management of the Seervada
park would like to determine how to route the various tram trips
from the park entrance (Station O) to the scenic (Station T) to
maximize the number of trips per day.
Each tram will return by the same route it took on the outgoing
trip so the analysis focuses on outgoing trips only.
To avoid unduly disturbing the ecology and wildlife of the
region, strict upper limits have been imposed on the number of
outgoing trips allowed per day in the outbound direction on
each individual road.
For each road the direction of travel for outgoing trips is
indicated by an arrow in the next slide.
The number at the base of the arrow gives the upper limit on
the number of outgoing trips allowed per day.
42
.

Consider the problem of sending as many units from node O to
node T for the following network (current flow, capacity):
(0,3)
A
(0,9)
(0,5)
(0,1)
(0,4)
T
(0,1)
(0,7)
(0,6)
B
(0,5)
(0,4)
(0,2)
(0,4)
C
43
.

Iteration 1: Select the augmenting path OBET, with
residual capacity of min{7, 5, 6} = 5. Assign a flow of 5.
(0,3)
A
(0,9)
(0,5)
(0,1)
(0,4)
T
(0,1)
(5,7)
(5,6)
B
(5,5)
(0,4)
(0,2)
(0,4)
44
C
.

Iteration 2: Assign a flow of 3 to the augmenting path
OADT. The resulting residual network is

(3,3)
(3,9)
(3,5)
(0,1)
(0,4)
T
(0,1)
(5,7)
(5,6)
B
(5,5)
(0,4)
(0,2)
(0,4)
45
C
.

OABDT. The resulting residual network is

(3,3)
A
(4,9)
(4,5)
(1,1)
(1,4)
T
(0,1)
(5,7)
(5,6)
B
(5,5)
(0,4)
(0,2)
(0,4)
46
C
.

OBDT. The resulting residual network is

(3,3)
A
(6,9)
(4,5)
(1,1)
(3,4)
T
(0,1)
(7,7)
(5,6)
B
(5,5)
(0,4)
(0,2)
(0,4)
47
C
.

OCEDT. The resulting residual network is

(3,3)
(7,9)
(4,5)
(1,1)
(3,4)
T
(1,1)
(7,7)
(5,6)
B
(5,5)
(1,4)
(0,2)
(1,4)
48
C
.

OCET. The resulting residual network is

(3,3)
A
(7,9)
(4,5)
(1,1)
(3,4)
T
(1,1)
(7,7)
(6,6)
B
(5,5)
(2,4)
(0,2)
(2,4)
49
C
.

There are no more flow augmenting paths, so the current
flow pattern is optimal.

3
A
1
13
O
D
3
1
6
7
5
2
2
C
50
.
13
Network Analysis & LP to be continued

51
.
Assignment #2: Transshipment

2.1 Solve Exercise 1, using the balanced transportation
tableau from the previous slide.

2.2 Redo Exercise 1. Assume that before being shipped to Los
Angeles or New York, all oil produced at the wells must be
refined at either Galveston or Mobile. To refine 1,000 barrels of
oil costs $12 at Mobile and $10 at Galveston. Assuming that
both Mobile and Galveston have infinite refinery capacity,
formulate AND SOLVE a transshipment and balanced
transportation model to minimize the daily cost of transporting
and refining the oil requirements of Los Angeles and New York.
Assignment #2 Due Thursday November 5, 2015 4 pm *

You may work solo or in pairs. Remember academic honesty!
52
.
LECTURE 12B
NETWORK MODELS
1
.
Network Analysis & LP continued

2
.
Network
Flows: Theory
Networks
Basic elements:
N
A
3
Nodes (let m denote number of them).

Directed Arcs
subset of all possible arcs: {(i, j) : i, j N , i = j}.

arcs are directed: (i, j) = (j, i).
R. J. Vanderbei (2008) Linear Programming: Foundations and Extensions, 3rd Ed., Kluwer. ISBN: 978-0-387-74387-5.
Network
Flow
Network Flow
DataData
bi, i N , supply at node i
cij , (i, j) A, cost of shipping 1 unit along arc (i, j).

Note: demands are recorded as negative supplies.
4
Network Flow Problem
Network Flow Model

Decision Variables:
xij , (i, j) A,
quantity to ship along arc (i, j).
Objective:
minimize
cij xij
(i,j)A
Network Flow ProblemCont.
Network Flow Model

Constraints:
Mass conservation (aka flow balance):

inflow(k) outflow(k) = demand(k) = supply(k),
xik
xkj
=
bk ,
kN
i:
(i, k) A
j:
(k, j) A
Nonnegativity:
6
kN
xij 0,
(i, j) A
Network
Flow Model: Matrix Notation
Matrix Notation
minimize cT x
subject to Ax = b
x 0
where
cT =
1 1
1
A =
11
1
1
1 1
15
16
Notes
1
1
1 1
1
1 1 1
18
1 1 1
0
19
16
33
b=
0
36
A is called node-arc incidence matrix.

A is large and sparse.
Dual Problem
Network Flow Model: Dual Problem

maximize bT y
subject to AT y + z = c
z0
In network notation:
maximize iN biyi
subject to yj yi + zij = cij
zij 0
8
(i, j) A
(i, j) A
Complementarity Relations
Network Flow Model: Complementarity

The primal variables must be nonnegative.
Therefore the associated dual constraints are inequalities.
The dual slack variables are complementary to the primal variables:

xij zij = 0,
(i, j) A
The primal constraints are equalities.
Therefore they have no slack variables.
The corresponding dual variables, the yis, are free variables.

No complementarity conditions apply to them.
9
Definition: Subnetwork
Network Definitions
Network
10
Subnetwork
Connected vs. Disconnected
Network Definitions
Connected
11
Disconnected
Cyclic vs. Acyclic
Network Definitions
Cyclic
12
Acyclic
Trees
Network Definitions
Tree = Connected + Acyclic
13
Not Trees
anning Trees
Network Definitions
Spanning TreeA tree touching every node
e Solution
14
xij = 0
for (i, j) Tree Arcs
Spanning Trees
Network Definitions
Spanning TreeA tree touching every node
Tree Solution
xij = 0
for (i, j) Tree Arcs
Note: Tree solutions are easy to computestart at the leaves and work inward...
15
Maximum Flow Network Model

Each arc has a capacity that limits the quantity of a
product that may be shipped through the arc.

Objective: transport the maximum amount of flow from a
starting point (source) to a terminal point (sink).
16
.
Max-Flow Min-Cut Theorem

In a flow network, the maximum amount of flow passing
from the source to the sink is equal to the minimum

capacity that, when removed in a specific way from the
network, causes the situation that no flow can pass from
the source to the sink.
It is a special case of the duality theorem for linear
programs and can be used to derive Menger's theorem
and the KnigEgervry theorem.
17
.
Max-Flow Min-Cut Theorem

A cut is defined as any set of directed arcs containing at
least one arc from every directed path from the source to
the sink.
For any particular cut, the cut value is the sum of the arc
capacities of the arcs of the cut.
The theorem states that, for any network with a single
source and sink, the maximum feasible flow from the
source to the sink equals the minimum cut value for all
cuts of the network.
18
.
Maximum Flow Problem: Sunco

Sunco Oil wants to ship the maximum possible amount of
oil (per hour) via pipeline from node so to node si as shown
in the figure below.
Arc
Capacity
a0
(1)3
so
(2)2
(1)3
(1)4
(2)2
(1)1
si
(so,1)
(so,2)
(1,2)
(1,3)
(3,si)
(2,si)
The arcs represent pipelines of different diameters.

(xij) bij = (possible flow in arcij) maximum flow in arcij
19
.

Arc
Capacity
a0
(1)3
so
(2)2
(1)3
(1)4
(2)2
si
(1)1
Node so is the source node

Node si is called the sink node
20
.
(so,1)
(so,2)
(1,2)
(1,3)
(3,si)
(2,si)

Arc
Capacity
a0
(1)3
so
(2)2
(1)3
(1)4
(2)2
si
(1)1
Node so is the source node

Node si is called the sink node
ADD artificial arc a0 from source to sink
21
.
(so,1)
(so,2)
(1,2)
(1,3)
(3,si)
(2,si)
Max-Flow Min-Cut Theorem Example

This network has a value of flow of 7.
The vertex in white and the vertices in grey form the subsets S
and T of an s-t cut, whose cut-set contains the dashed edges.

Since the capacity of the s-t cut is 7, which equals to the value
of flow, the max-flow min-cut theorem tells us that the value of
flow and the capacity of the s-t cut are both optimal in this
network.
22
.

Objective:
Maximize X0 = total amount of oil entering the sink
Subject to constraints:
1. 0 flow through each arc arc capacity
2. Flow into node i = Flow out of node i (conservation-offlow constraint)
23
.

Let Xij = Millions of barrels of oil per hour that will pass
through arc(i,j) of pipeline.
For a flow to be feasible:
1. 0 flow through each arc arc capacity
2. Flow into node i = Flow out of node i (conservation-offlow constraint)
Assumption: No flow loss
artificial arc a0 satisfies the conservation-of-flow
constraint for the source & sink
Let X0 = total amount of oil entering the sink
24
.
s.t.
xso,1 ! 2
xso,2 ! 3
x12 ! 3
(Arc capacity constraints)
x ! 2 Flow Problem: Sunco

Maximum
2,si
x13 ! 4
x3,si !
x0 "
xso,1 "
xso,2 # x12 "
x13 # x12 "
x3,si # x2,si "
xij $
1
xso,1 # xso,2
x12 # x13
x2,si
x3,si
x0
0
(Node
(Node
(Node
(Node
(Node
so flow constraint)
1 flow constraint)
2 flow constraint)
3 flow constraint)
si flow constraint)
optimal solution to this LP is z " 3, xso,1 " 2, x13 " 1, x12 " 1, xso,2 " 1, x3,si "
i " 2, x0 " 3. Thus, the maximum possible flow of oil from node so to si is 3 milarrels per hour, with 1 million barrels each sent via the following paths: so12si,
3si, and so2si.
25
.
continued

One optimal solution:
z=3, xso,1 =2, x13 =1, x12 =1, xso,2 =1, x3,si = 1, x2,si =2, x0 =
3.
Max flow is 3 million barrels per hour
1 million barrels on each of 3 paths: so12si, so13
si, and so2si.
a0
(1)3
so
(2)2
(1)3
(1)4
26
.
2
3
(2)2
(1)1
si
We assume that no oil gets lost while being pumped through the network, so at each
node, a feasible flow must satify (2), the conservation-of-flow constraint. The introduction
of the artificial arc a0 allows us to write the conservation-of-flow constraint for the source
and sink.
If we let x0 be the flow through the artificial arc, then conservation of flow implies that
x0 " total amount of oil entering the sink. Thus, Suncos goal is to maximize x0 subject
to (1) and (2):
max z " x0
s.t.
xso,1 ! 2
(Arc capacity constraints)
xso,2 ! 3
x12 ! 3
x2,si ! 2
x13 ! 4
x3,si ! 1
27
x0 "
xso,1 "
xso,2 # x12 "
x13 # x12 "
x3,si # x2,si "
xso,1 # xso,2
x12 # x13
x2,si
x3,si
.
x0
(Node
(Node
(Node
(Node
(Node
so flow constraint)
1 flow constraint)
2 flow constraint)
3 flow constraint)
continued
si flow constraint)
Exercise: Maximum Flow Problem

The Hatfields, Montagues, McCoys, and Capulets are
going on their annual family picnic.

Four cars are available to transport the families to the
picnic.
The cars can carry the following number of people: car 1,
four; car 2, three; car 3, three; and car 4, four.
There are four people in each family, and no car can carry
more than two people from any one family.
Draw the maximum-flow network graph that shows the
problem of transporting the maximum possible number of
people to the picnic.
28
.
Exercise Answer: Maximum Flow

Problem
8.
29
See figure.
9. Step
4
units
flow alon
1-3-si
Step
2
units
flow alon
2-4-si
Step
1 unit of flow along so-2-4-3-si

.
Step 4-Add 1 unit along so-2-1-3-si
Cannot label si. Maximum flow = 8 Arcs in cut are
Network Analysis & LP continued

30
.
CPM/PERT
CPM: Critical Path Method
Used to determine length of time of a project
Based on known duration of time of each activity
Used to determine slack in project activities
PERT: Project Evaluation and Review Technique
Used to estimate the probability that a project will be completed by
a specific deadline
Used when duration of time of each activity is uncertain
Used in project management
See http://www.pmi.org for PMP certification & PMBOK Guide
(global standard)
31
.
CPM/PERT
PERT chart for a project with five milestones (10 through 50) and
six activities (A through F).
The project has two critical paths: activities B and C, or A, D, and F
giving a minimum project time of 7 months with fast tracking.
Activity E is sub-critical, and has a float of 1 month.
32
.
Wikipedia.org
CPM/PERT: Gantt Chart
33
http://www.vertex42.com/ExcelTemplates/excel-gantt-chart.html
.
CPM/PERT: Applications
Scheduling construction projects such as office buildings,
highways and swimming pools

Developing countdown and hold procedure for the
launching of space crafts
Installing new computer systems
Designing and marketing new products
Completing corporate mergers
Building ships
ALL ARE PROJECTS, with a beginning and an end,
based on specified termination criteria.
34
.
CPM/PERT: Project Planning, Scheduling

and Control
Planning: organized approach to accomplish the goal of
minimizing elapsed time of project

defines objectives and tasks; represents tasks interactions on a
network; estimates time and resources
Scheduling: a time-phased commitment of resources

identifies critical tasks which, if delayed, will delay the projects
completion time.
Control: means of monitoring and revising the progress of
a project
35
.
CPM: Network Representation

Tasks (or activities) are represented by directed arcs
Each task has a duration denoted by tj
Node 0 represents the start and node n denotes the finish of the
project
Precedence relations are shown by arcs
specify what other tasks must be completed before the task in
question can begin.
A path is a sequence of linked tasks going from beginning
to end
Critical path is the longest path
36
.
CPM: Basic technique

First, construct a project model with:
A list of all activities required to complete the project
(typically categorized within a work breakdown structure
or WBS)
The time (duration) that each activity will take to

complete
The dependencies between the activities and,
Logical end points such as milestones or deliverable

items.
continued
37
.

Then, use CPM to calculate the longest path of planned
activities to logical end points or to the end of the project,
and the earliest and latest that each activity can start and
finish without making the project longer.
continued
38
.
This process determines which activities are "critical" (i.e.,

on the longest path) and which have "total float" (i.e., can
be delayed without making the project longer).
continued
39
.
In project management, a critical path is the sequence of

project network activities which add up to the longest
overall duration, regardless if that longest duration has float
or not. This determines the shortest time possible to
complete the project. There can be 'total float or
slack (unused time) within the critical path.
40
.
CPM: Project Network

For each activity there is a set of activities (called the
predecessors of the activity) that must be completed

before the activity begins.
A project network is used to represent the precedence
relationships between activities.
Activities are represented by directed arcs
Nodes represent events: completion of a set of activities
(Activity on arc (AOA) type of network).
1
41
Activity A must be completed before activity B starts

.
CPM: Project Network Rules

Node 1 represents the start of the project. An arc should
lead from node 1 to represent each activity that has no

predecessors.
Include a finish node representing project completion.
Number the nodes in order:
Beginning of activity node number < completion of activity node
number.
An activity is represented by at most one arc

Two nodes can be connected by at most one arc.
To avoid violating the above 2 rules, use a dummy
activity that takes zero time.

42
.
CPM: Formulating the Problem

Input Data:
Precedence relationships and durations
Decision Variable:
ESi: Earliest starting times for each of the tasks, ti
Objective:
Minimize the elapsed time of the project
where node n is the last node in the graph
43
.
CPM: Constraints
If tj is the earliest starting time of a task, ESi is the earliest
starting time of an immediate predecessor and ti is the
duration of the immediate predecessor, then
ESj = ESi + ti for every arc (i, j)
44
.
CPM: Critical Path Definitions

Earliest Start Time (ES) is the earliest time a task can
feasibly start
Earliest Finish Time (EF) is the earliest time a task can
feasibly end
Latest Start Time (LS) is the latest time a task can feasibly
start, without delaying the project at all.
Latest Finish Time (LF) is the latest time a task can
feasibly end, without delaying the project at all
45
.

Forward Pass: used to find ES & EF
Go through the jobs in order
Start each job at the earliest time while satisfying the
precedence constraints
Find the earliest start and finish times
EFi = ESi + ti
Earliest start time for an activity leaving a particular node
is equal to the maximum of the earliest finish times for all
activities entering the node.
46
.

Backward Pass: used to find LS & LF
Set the finish time
Look at tasks in reverse order
Lay out tasks one at a time on a Gantt chart starting at the
finish and working backwards to the start

Start the task at its latest starting time
LSi = LFi - ti
Latest finish time for an activity entering a particular
node is equal to the minimum of the latest start times for
all activities leaving the node.
47
.
CPM: Critical Path

Theorem: The minimum project duration is the length of
the longest path in the schedule. The longest path is

called the critical path.
Look for tasks whose earliest start time and latest start
time are the same. These tasks are critical, and are on a
critical path.
48
.
CPM: Critical Path

Critical paths are found by identifying those tasks where
ES=LS (equivalently, EF=LF)

No flexibility in scheduling tasks on the critical path
The total length (makespan) of the critical path schedule
equals the LF of the final task
Slack is the difference between LS and ES, or LF and EF.
An activity with a slack of zero is on the critical path
49
.
CPM: Widgetco
Widgetco is about to introduce a new product. Below find
the data for CPM.

Activity
Predecessors
A:train workers
B:purchase raw materials
C:produce product 1
A, B
D:produce product 2
A, B
E:test product 2
D
F:assemble products 1&2
C, E
50
.
Duration(days)
6
9
8
7
10
12
CPM: Widgetco Project Diagram

C8
F 12
3
A6
5
D7
Dummy
1
B9
E 10
Node 1 = starting node

Node 6 = finish node
51
.
CPM: Widgetco Forward Pass (ES,EF)
(0,6)
(9,17)
C8
(26,38)
F 12
A6
5
D7
Dummy
1
B9
E 10
(9,16)
(0,9)
52
.
(16,26)
CPM: Widgetco Backward Pass (LS,LF)
(3,9)
(18,26)
C8
(26,38)
F 12
A6
5
D7
Dummy
1
B9
E 10
(9,16)
(0,9)
53
.
(16,26)
CPM: Widgetco Starts

Compute Slacks:
Activity
ES(i)
LS(i)
Activity A: 3
Activity D: 0
18
Activity C: 9
16
16
26
26
Activity B: 0
Activity E: 0
Activity F: 0
54
.
CPM: Widgetco Critical Path

An activity with a slack of zero is a critical activity
A path from node 1 to the finish node that consists
entirely of critical activities is called a critical path

Widgetco critical path: B-D-E-F, with a Makespan = 38
55
.
LP: Widgetco Critical Path

Decision variable:
Xj: the time that the event corresponding to node j occurs
Since our goal is to minimize the time required to
complete the project, we use an objective function of:

Z=XF-X1
Note that for each activity (i,j), i must occur before j occurs
& activity (i,j) must be completed.

56
.
LP: Widgetco Critical Path

Min Z =X6-X1
S.T. X3 X1+6
X2 X1+9
X5 X3+8
X4 X3+7
X5 X4+10
X6 X5+12
X3 X2
57
(Arc (1,3) constraint)

The optimal solution to this LP is Z=38, X1=0, X2=9, X3=9,

X4=16, X5=26, X6=38
.
Readings on Network Models

Chapter 8, Operations Research: Applications and
Algorithms (4th edition). W.L. Winston. Duxbury Press,

2003. ISBN 978-0534380588.
(optional) R. J. Vanderbei (2008) Linear Programming:
Foundations and Extensions, 3rd Ed., Kluwer. ISBN:
978-0-387-74387-5.
(optional) Bazaraa, Jarvis, and Sherali (1990) Linear
Programming and Network Flows, Wiley.
58
.

ALO

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

ALO

Загружено:

Авторское право:

Доступные форматы

INDE6372 Advanced Linear Optimization

INDE6372 Advanced Linear Optimization

basic feasible solution in which a basic variable = 0.

INDE6372 Advanced Linear Optimization

basic feasible solution in which a basic variable = 0.

cycling. This is where the simplex algorithm gets stuck in

INDE6372 Advanced Linear Optimization

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

Perform the pivot

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

xx1 xx2 xx3

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

Adapted from the textbook, Page 31.

10x1 57x2 9x3 24x4

10x1 57x2 9x3 24x4

x33 x44 x55

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

A Cycling Example: Continuation

Update the pivot

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

A Cycling Example: Continuation

Update the pivot

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

A Cycling Example: Continuation

One more iteration

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

A Cycling Example: Continuation

One more iteration

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

A Cycling Example: Continuation

One more iteration

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

Cycling Example: Continuation

Jiming Peng Department of Industrial Engineering University of Houston

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

Jiming Peng Department of Industrial Engineering University of Houston

INDE6372 Advanced Linear Optimization

Blands Rules (1977)

Entering variable selection: If more than one non-basic

INDE6372 Advanced Linear Optimization

Bland (1977) proposed the following two rules:

Leaving variable selection: If a tie occurs in M.R.T. test,

Advanced Linear Optimization

Blands rule applied

to the following tableau:

iming Peng Department of Industrial Engineering University of Houston

Cycling Example with Blands Rules

INDE6372 Advanced Linear Optimization

Blands Rules (1977)

INDE6372 Advanced Linear Optimization

A mathematical optimization problem has the following form

A solution x X is said to be optimal if

INDE6372 Advanced Linear Optimization

ng called a conic combination.