Linear Programing Optimization

Problems
Problems where we have to minimize (or
maximize) a linear (objective) function subject
to certain conditions determined by a set of linear
inequalities (constraints).

1
 Two-variable LP
Model (Section 2.1)

2
Guidelines for formulating Linear
Programming model
• Identify and define the decision variable of the
problem
• Define the objective function
• State the constraints to which the objective
function should be optimized (i.e. Maximization
or Minimization)
• Add the non-negative constraints from the
consideration that the negative values of the
decision variables do not have any valid physical
interpretation
Reddy Mikks Problem
Reddy Mikks produces both interior and exterior paints from two
raw materials, M1 and M2. The following table provides the basic
data of the problem:

The daily demand for interior paint cannot exceed that for exterior
paint by more than 1 ton. Also, the maximum daily demand for
interior paint is 2 tons. Reddy Mikks wants to determine the
optimum (best) product mix of interior and exterior paints that
maximizes the total daily profit.
4
Mathematical Formulation
All OR models consist of three basic components:
1. Decision variables that we seek to determine.
2. Objective (goal) that we need to optimize
(maximize or minimize).
3. Constraints that the solution must satisfy.

5
Decision variables
Determine the optimum (best) product mix of
interior and exterior paints.
• x1: Daily production of exterior paint (tons)
• x2: Daily production of interior paint (tons)

6
Objective
• Maximize profit (z  objective function)
− Profit from exterior paint = 5x1 (thousand) dollars
− Profit from interior paint = 4x2 (thousand) dollars

𝒛 = 𝟓𝒙𝟏 + 𝟒𝒙𝟐 Objective

7
Constraints
• Availability of raw material M1
− 6 tons per ton of exterior paint and 4 tons per ton
of interior paint.
− Usage of raw material M1 = 6x1 + 4x2 tons/day

𝟔𝒙𝟏 + 𝟒𝒙𝟐 ≤ 𝟐𝟒 Material constraint

8
Constraints
• Availability of raw material M2
− 1 tons per ton of exterior paint and 2 tons per ton
of interior paint.
− Usage of raw material M2 = 1x1 + 2x2 tons/day

𝟏𝒙𝟏 + 𝟐𝒙𝟐 ≤ 𝟔 Material constraint

9
Constraints
• Demand limitation
− Daily production of interior paint cannot exceed
that of exterior paint by more than 1 ton.

𝒙𝟐 − 𝒙𝟏 ≤ 𝟏 Demand constraint

− Maximum daily demand of interior paint is 2 tons

𝒙𝟐 ≤ 𝟐 Demand constraint

10
Constraints

𝒙𝟏 ≥ 𝟎, 𝒙𝟐 ≥ 𝟎 Non-negativity constraints

11
Complete formulation
Maximize:
𝒛 = 𝟓𝒙𝟏 + 𝟒𝒙𝟐 Objective

subject to constraints…
𝟔𝒙𝟏 + 𝟒𝒙𝟐 ≤ 𝟐𝟒 Material constraint
𝟏𝒙𝟏 + 𝟐𝒙𝟐 ≤ 𝟔 Material constraint
𝒙𝟐 − 𝒙𝟏 ≤ 𝟏 Demand constraint
𝒙𝟐 ≤ 𝟐 Demand constraint
𝒙𝟏 ≥ 𝟎, 𝒙𝟐 ≥ 𝟎 Non-negativity constraints
12
Notes
• x1 and x2 that satisfy all five constraints constitute a
feasible solution.
• Otherwise, the solution is infeasible.
• Goal is to find the optimum (the best) feasible solution
that maximizes the total profit.
– Graphical method
– Algebraic generalization

13
Notes
• For the Reddy Mikks Problem
– Is x1 = 3 and x2 = 1 a feasible solution?
– Is x1 = 4 and x2 = 1 a feasible solution?

• How many feasible solutions are there?

14
Graphical LP Solution
Two steps:
1. Determination of the feasible solution space.
2. Determination of the optimum solution from among all the
points in the solution space.

15
Step 1: Feasible Solution Space Determination
𝒙𝟏 ≥ 𝟎, 𝒙𝟐 ≥ 𝟎 Non-negativity constraints
𝒙𝟐

𝒙𝟐 ≥ 𝟎

𝒙𝟏 ≥ 𝟎
𝒙𝟏
16
Step 1: Feasible Solution Space Determination
𝟔𝒙𝟏 + 𝟒𝒙𝟐 ≤ 𝟐𝟒 Material constraint
𝒙𝟐

𝒙𝟐 ≥ 𝟎

𝟔𝒙𝟏 + 𝟒𝒙𝟐 ≤ 𝟐𝟒

𝟔𝒙𝟏 + 𝟒𝒙𝟐 = 𝟐𝟒

𝒙𝟏 ≥ 𝟎
𝒙𝟏
17
Maximize 𝐳 = 𝟓𝐱 𝟏 + 𝟒𝐱 𝟐

18
Maximize 𝐳 = 𝟓𝐱 𝟏 + 𝟒𝐱 𝟐

19
Step 2: Optimum Solution Determination
• Infinite feasible solutions  procedure to determine
the optimum

20
21
Step 2: Optimum Solution Determination
Graphical procedure to determine the optimum.

Typical LPs may include hundreds of variables and

constraints. Then why study a two-variable LP?

22
Step 2: Optimum Solution Determination
• Optimum solution of an LP, when it exists, is
always associated with a corner point of the
solution space.

– True even if objective function (say Z = 6XI + 4X2) is

parallel to a constraint.

23
Z = 6XI + 4X2

24
Step 2: Optimum Solution Determination
• Optimum solution of an LP, when it exists, is
always associated with a corner point of the
solution space.

– True even if objective function is parallel to a constraint.

– Optimum occurs at B or corner point C.
– Any point on BC will be an alternative optimum.
– BC is totally defined by points B and C.

25
Step 2: Optimum Solution Determination
• Optimum solution of an LP, when it exists, is always
associated with a corner point of the solution space.
• Limits search for optimum from feasible points (infinite) to
corner points (finite).
Corner Point (x1, x2) z
A
B
C
D
E
F
26
Step 2: Optimum Solution Determination
• Optimum solution of an LP, when it exists, is always
associated with a corner point of the solution space.
• Limits search for optimum from feasible points (infinite) to
corner points (finite).
Corner Point (x1, x2) z
A (0, 0) 0
B (4, 0) 20
C (3, 1.5) 21 (MAXIMUM)
D (2, 2) 18
E (1, 2) 13
F (0, 1) 4
27
Step 2: Optimum Solution Determination
• Optimum solution of an LP, when it exists, is
always associated with a corner point of the
solution space.

• Limits search for optimum from feasible points

(infinite) to corner points (finite).
• Corner points increase, enumeration procedure
becomes computationally expensive.
• Basis for development of simplex method.

28
Step 2: Optimum Solution Determination
Another constraint is added to the Reddy Mikks problem:
x1 + x2 ≥ 1, and the problem is changed from
maximization to minimization. Find the optimal solution.

29
 x1 + x2 ≥ 1 (7)

Feasible region

30
 x1 + x2 ≥ 1 (7)

Z = 5XI + 4X2

31
Diet Problem (Example 2.2-2)
Ozark Farms uses at least 800 lb of special feed daily. The
special feed is a mixture of corn and soybean meal with
the following compositions:

The dietary requirements of the special feed are at least

30% protein and at most 5% fiber. The goal is to
determine the daily minimum-cost feed mix.
32
Diet Problem (Example 2.2-2)
Minimize
z = 0.3x1 + 0.9x2
Subject to constraints
x1 + x2 ≥ 800
0.09x1 + 0.60x2 ≥ 0.30(x1 + x2)
0.02x1 + 0.06x2 ≤ 0.05(x1 + x2)
x1 ≥ 0, x2 ≥ 0

33
Diet Problem (Example 2.2-2)
Minimize
z = 0.3x1 + 0.9x2
Subject to constraints
x1 + x2 ≥ 800
0.21x1 − 0.30x2 ≤ 0
0.03x1 − 0.01x2 ≥ 0
x1 ≥ 0, x2 ≥ 0

34
Minimize z = 0.3x1 + 0.9x2

Feasible region

35
Minimize z = 0.3x1 + 0.9x2

z = $ 1350

Feasible region

z = $ 900
x1 = 470.6 lb
x2 = 329.4 lb
z = $ 437.64

36
Maximize z = 5x1 + 7 x2
Subject to x2

2x1− x2 ≤ −1
−x1 + 2x2 ≤ −1
x1 , x2 ≥ 0

x1
No feasible solution
37
Maximize z = x1 + x2
Subject to x2

−x1 + 3x2 ≥ 30
−3 x1 + x2 ≤ 30
x1 , x 2 ≥ 0
z=70
z=50
z=30
z=20

x1
Unbounded solution
38
Summary
• Usually a LPP will have a unique optimal solution.
• The (unique) optimum solution occurs at one of the
“corners” points.

• Some LPPs may have:

– alternative optimum solutions
– no solution
– unbounded solutions

39
Outline
• Prove that optimum of a LPP lies at corner
points.
– Relationship between LPP and convex sets.
– Definition of a convex set.
– Definition of vertex/extreme/corner points.
– Definition of Convex Linear Combination (CLC).
– Proof: optimum of a LPP occurs at corner points.

40
Relationship between LPP and convex sets
The set of points satisfying all the constraints and non-
negativity criteria of a LPP, is a convex set.

Proof: self study

41
Convex set
• Convex set: if the line segment joining any two
distinct feasible points also falls in the set.

42
Convex set
• Which of the following are convex sets?

A B C

43
Convex set: line segment
• How do you write the equation of a line segment
between two points X (x1, x2) and Y (y1, y2) in two-
dimensions (R2)?

44
Convex set: line segment
• How do you write the equation of a line segment
between two points X (x1, x2) and Y (y1, y2) in two-
dimensions (R2)?

Line segment between x and y:

L = {(1 − α)X + αY: α ϵ [0,1]}

Can be extended to n-dimensions also (Rn)

45
Convex set: definition
A set (S ⊂ Rn) is convex:
if for all X and Y ϵ S, the points (or line segment
connecting X and Y):
L = {(1 − α)X + αY: α ϵ [0,1]} also belong to S

46
Example: convex set
Show that the set:
S = {(x, y)| x + y ≤ 1, x ≥ 0, y ≥ 0}
is convex.
Is the non-negativity condition essential for the proof ?

47
Example: convex set
Show that the set: S = {(x, y)| x + y ≤ 1} is convex.

(0, 1)

(1, 0) x
48
Example: convex set
Show that the set: S = {(x, y)| x + y ≤ 1} is convex.

Take 𝐚 = (ax , ay ) ϵ S & 𝐛 = (bx , by ) ϵ S

L = (lx , ly ) = (1 − α)a + αb, where α ϵ [0,1]

⇒ L = (1 − α)(ax , ay ) + α(bx , by )
⇒ L = (lx , ly )=((1 − α)ax +αbx , (1 − α)ay +αby )
lx + ly = (1 − α)(ax +ay ) + α(bx +by ) ≤ 1 − α+α
⇒ lx + ly ≤ 1 ⇒ L ϵ S
49
Example: convex set
Is the set:
S = {(x, y)| x ≥ 1 or y ≥ 2}
convex.

50
Example: convex set

y
S
(0, 2) (1, 2)

(1, 0) x
51
Example: convex set
In S = {(x, y)| x ≥ 1 or y ≥ 2}
Take a = (1, 0) ϵ S & b = (0, 2) ϵ S
L = (lx , ly ) = (1 − α)a + αb
where α ϵ [0,1]
⇒ L = (1 − α)(1,0) + α(0,2)
⇒ L= lx , ly = (1 − α, 2α)
⇒ lx ≤ 1 and ly ≤ 2
Since L∉S ⇒ S is not convex

52
Example: convex set
Is
X = {(x1, x2, x3)| 2x1 + 3x2 + x3 ≤ 5}
a convex set?

53
Example: convex set

54
Vertex/extreme/corner points
• A corner point of the convex set is a feasible point that
cannot lie on a line segment joining any two distinct
feasible points in the set.

55
Definitions
• A set is said to be open if it contains no
boundary point.
• A set is said to be closed if it contains all
boundary point
Example: corner points
Determine graphically the extreme points of the following
convex set:
𝐒 = {(𝐱𝟏 , 𝐱𝟐 )|𝐱𝟏 + 𝐱 𝟐 ≤ 𝟐, 𝐱 𝟏 ≥ 𝟎, 𝐱𝟐 ≥ 𝟎}

59
Example: corner points
Determine graphically the extreme points of the following
convex set:
S = {(x1, x2)| x1 + x2 ≤ 2, x1 ≥ 0, x2 ≥ 0}

60
Convex Linear Combination (CLC)
• A convex linear combination (CLC) of points X1, X2,
X3, … is defined as:

𝐗 = α1𝐗1 + α2𝐗2 + α3𝐗3 + …

α1 + α2 + α3 + … = 1
αi ≥ 0

61
Example: CLC
Determine graphically extreme points of the convex set:
S = {(x1, x2)| x1 + x2 ≤ 2, x1 ≥ 0, x2 ≥ 0}

Show that the entire feasible solution space can be

determined as a CLC (a.k.a. convex combination) of its
extreme points.
62
Example: CLC
Determine graphically extreme points of the convex set:
S = {(x1, x2)| x1 + x2 ≤ 2, x1 ≥ 0, x2 ≥ 0}
α3

α2

α1

Show that the entire feasible solution space can be

determined as a CLC of its extreme points.

63
Example: CLC
A CLC of corner points
X = α1𝐗1 + α2𝐗2 + α3𝐗3 = α1(0,0) + α2(2,0) + α3(0,2)
⇒ 𝐗 = (2α2,2α3)
where α2 ≥ 0, α3 ≥ 0 and α1 + α2 + α3 = 1
α3

α2

α1
64
Example: CLC
Let a1 = 2α2 and a2 = 2α3
⇒ 𝐗 = (a1, a2), where a1 ≥ 0, a2 ≥ 0 and

a1 a2
α1 + + = 1
2 2
⇒ a1 + a2 = 2(1 − α1)
⇒ a1 + a 2 ≤ 2

Thus, any convex (bounded) solution space is totally

defined once its corner points are known.
65
Example (CLC): 7.1a (4)
In the solution space in
Figure (drawn to scale),
express the interior point
(3, 1) as a convex
combination of the
extreme points A, B, C,
and D by determining
the weights associated
with the extreme points.

66
Example (CLC): 7.1a (4)

67
Outline
• Prove that optimum of a LPP lies at corner
points.
– Relationship between LPP and convex sets 
– Definition of a convex set 
– Definition of vertex/extreme/corner points 
– Definition of Convex Linear Combination (CLC) 
– Proof: optimum of a LPP occurs at corner points

68
Optimum occurs at corner points
The optimum of the objective function (z) of an
LPP occurs at a corner point of the feasible
solution space (SF), provided SF is non-empty and
bounded.

We have already agreed that SF is a convex

set.

70
Optimum occurs at corner points
Proof: Let the LPP be a maximization problem,
and the feasible region (SF) be a bounded convex
set.
Let the corner points of SF be X1, X2, X3, ….Xm.
Let z = f(X) assume its maximum value at point
X0 ϵ SF, that is:
f(X0) ≥ f(X) for all X ϵ SF

71
Optimum occurs at corner points
Since SF is non-empty, bounded convex set, any
point in SF is a CLC of its corner points; thus,
X0 = α1X1 + α2X2 + … + αmXm
α1 + α2 + ... + αm = 1
αi ≥0
Hence,
z = f(X0) = f(α1X1 + α2X2 + … + αmXm)
⇒ z = f(X0)= α1f(X1)+α2f(X2)+ … + αmf(Xm)
72
Optimum occurs at corner points
Let us choose that corner point which gives the
largest value of the objective function 
f(Xk) ≥ f(Xi) for i = 1,2,…m

Hence,
z = f(X0) ≤ α1f(Xk)+α2f(Xk)+ … + αmf(Xk)
⇒ z = f(X0) ≤ (α1 + α2 + ... + αm)f(Xk)
⇒ z = f(X0) ≤ f(Xk)
73
Optimum occurs at corner points
Thus,
1) z = f(X0) ≥ f(Xk)
2) z = f(X0) ≤ f(Xk)

⇒ f(X0)= f(Xk)

Maximum occurs at a corner point of SF.

74
 Maximize 𝐳 = 𝟓𝐱 𝟏 + 𝟒𝐱 𝟐

Show that the maximum lies on the

corner point C.

SF

75
Optimum occurs at corner points
The corner points of SF are A, B, C, D, E and F.

Corner Point (x1, x2) z

A (0, 0) 0
B (4, 0) 20
C (3, 1.5) 21
D (2, 2) 18
E (1, 2) 13
F (0, 1) 4

76
Optimum occurs at corner points
Let z = f(X) assume its maximum value at point X0 ϵ SF,
that is:
f(X0) ≥ f(X) for all X ϵ SF

 Any point in SF (X0) is a CLC of its corner points:

X0 = α1(0,0) + α2(4,0) + α3(3,1.5) + α4(2,2) + α5(1,2) + α6(0,1)
α1 + α2 + ... + α6 = 1
αi ≥0

77
Optimum occurs at corner points
Objective function: z = 5x01 + 4x02
X0 = α1(0,0) + α2(4,0) + α3(3,1.5) + α4(2,2) + α5(1,2) + α6(0,1)
x01 = 4α2 + 3α3 + 2α4 + α5
x02 = 1.5α3 + 2α4 + 2α5 + α6

⇒ z = f(X0 )
= 5(4α2 + 3α3 + 2α4 + α5) + 4(1.5α3 + 2α4 + 2α5 + α6)
⇒ z = f(X0 ) = α1 0 + α2 20 + α3 21 + α4 18 + α5 13 +
α6 4 ≤ 21 α1 + α2 + α3 + α4 + α5 + α6 = 21

78
Optimum occurs at corner points
Thus,
1) z = f(X0) ≥ f(C)
2) z = f(X0) ≤ f(C)

⇒ f(X0)= f(C)

Maximum occurs at the point C (3, 1.5).

79