Prof. M.

Hamdy El wany

1
st
Year Product i on Engi neer i ng
Depar t ment
2011

Operations Research
1
Lecture 5
Integer Programming
2
Introduction
An integer programming problem (IP) is an
LP in which some or all of its variables are
required to be non-negative integers.

Many real-life situations may be
formulated as IPs.

Mainly 3 classes depending on variable
type:

3
Pure Integer Programming
An IP in which all variables are required to
be integers is called a pure integer
programming problem. For example,
Max z = 3x
1
+ 2x
2

s.t x
1
+ x
2
6
x
1
, x
2
0,
x
1
, x
2
are integers


4
Mixed Integer Programming
An IP in which only some of the variables
are required to be integers is called a
mixed integer programming problem. For
example:
Max z = 3x
1
+ 2x
2

s.t x
1
+ x
2
6
x
1
, x
2
0,
x
1
is an integer


5
01 IP or Binary IP
An IP problem in which all the variables must
equal 0 or 1 is called a 01 IP or Binary IP.
The following is an example of a 01 IP:
Max z = x
1
- x
2
s.t. x
1
+ 2x
2
2
2x
1
- x
2
1
x
1
, x
2
= 0 or 1
01 IPs occur in surprisingly many situations.


6
Example 5.1
7
Stockco is considering four investments.
Stockco has only $14,000 for investment.
Stockco wants to maximize Net Income (NI)
obtained.
Investment 1 2 3 4
Initial Cost 5,000 7,000 4,000 3,000
NI 16,000 22,000 12,000 8,000
Example 5.1 (Contd)
8
Decision Variables:
for i = 1, 2, ., 4
Objective Function:
Maximize Z = 16X
1
+ 22X
2
+ 12X
3
+ 8X
4

Subject to:
5X
1
+ 7X
2
+ 4X
3
+ 3X
4
< 14
X
i
= 0 or 1 (For i=1, 2, 3, 4)
All numbers are in terms of thousands of dollars.

How Does it Differ?
9
max z = 21x
1
+ 11x
2
s.t. 7x
1
+ 4x
2
13
x
1
, x
2
0
x
1
, x
2
integer

How Does it Differ? (Contd)
Feasible is not a convex set.
Feasible region is the set of points:
S = {(0, 0), (0, 1), (0, 2), (0, 3), (1, 0), (1, 1)}.
If the feasible region for a pure IPs LP relaxation
is bounded, then the feasible region for the IP
will consist of a finite number of points.
Could be solved by enumeration: The optimal
solution here is x
1
= 0, x
2
= 3 with a value of z =
33.




10
Solving this Simple Example
11
Unfortunately, IPs are usually much harder
to solve than LPs.
Could be solved by:
Enumeration of all possible solutions.
Graphical Solution.
With the help of LP Relaxation.
The LP obtained by omitting all integer or 01
constraints on variables.

Solving this Simple Example (Contd)
Suppose that a naive analyst suggests the
following approach for solving an IP:
First solve the LP relaxation;
Then round off (to the nearest integer) each variable
that is required to be an integer and that assumes a
fractional value in the optimal solution to the LP
relaxation.
Applying this approach here, we first find the optimal
solution to the LP relaxation: x
1
= 13/7, x
2
= 0.

12
Solving this Simple Example (Contd)
Rounding this solution yields the solution x
1
= 2,
x
2
= 0 as a possible optimal solution. But it is
however infeasible.
Even if we round x
1
downward (yielding the
candidate solution x
1
= 1, x
2
= 0),

For some IPs, it can even turn out that every
roundoff of the optimal solution to the LP
relaxation is infeasible.
13
LP Relaxation
The concept of LP Relaxation of an integer
programming problem plays a key role in the
solution of IPs.
DEFINITION: The LP obtained by omitting all
integer or 01 constraints on variables is called
the LP relaxation of the IP.
This means that the feasible region for any IP
must be contained in the feasible region for the
corresponding LP relaxation.


14
LP Relaxation (Contd)
For any IP that is a max problem, this implies
that;
Optimal z-value for LP relaxation optimal z-
value for IP

If you solve the LP relaxation of a pure IP and
obtain a solution in which all variables are
integers, then the optimal solution to the LP
relaxation is also the optimal solution to the IP.
15
The Branch-and-Bound Method
In practice, most IPs are solved by
using the technique of branch-and-
bound.
Branch-and bound methods find the
optimal solution to an IP by efficiently
enumerating the points in a sub-
problems feasible region.

16
The Branch-and-Bound Method
Example 5.4:
Max z = 8x
1
+ 5x
2
s.t. x
1
+ x
2
6
9x
1
+ 5x
2
45
x
1
, x
2
0
x
1
, x
2
integer
The branch-and-bound method begins by
solving the LP relaxation of the IP.
We call the LP relaxation sub-problem 1.

17
The Branch-and-Bound Method
18
The Branch-and-Bound Method
Unfortunately, the optimal solution to the LP
relaxation is z = 165/4, x
1
= 15/4, x
2
= 9/4 (see
Figure).
We know that (optimal z-value for IP) (optimal
z-value for LP relaxation). This implies that the
optimal z-value for the IP cannot exceed 165/4.
Thus, the optimal z-value for the LP relaxation is
an upper bound for Telfas profit.

19
The Branch-and-Bound Method
Our next step is to partition the feasible region
for the LP relaxation in an attempt to find out
more about the location of the IPs optimal
solution.
We arbitrarily choose a variable that is fractional
in the optimal solution to the LP relaxationsay,
x
1
.

20
The Branch-and-Bound Method
Now observe that every point in the feasible
region for the IP must have either x
1
3 or x
1

4. (Why cant a feasible solution to the IP have 3
< x
1
< 4?)
With this in mind, we branch on the variable
x
1
and create the following two additional
subproblems:
Subproblem 2: Subproblem 1 + Constraint x
1
4.
Subproblem 3: Subproblem 1 + Constraint x
1
3.

21
The Branch-and-Bound Method
Observe that neither subproblem 2 nor
subproblem 3 includes any points with x
1
= 15/4.
This means that the optimal solution to the LP
relaxation cannot recur when we solve
subproblem 2 or subproblem 3.

22
The Branch-and-Bound Method
23
The Branch-and-Bound Method
From Figure, we see that every point in the feasible
region for the Telfa IP is included in the feasible
region for subproblem 2 or subproblem 3.
Also, the feasible regions for subproblems 2 and 3
have no points in common.
Because subproblems 2 and 3 were created by
adding constraints involving x
1
, we say that
subproblems 2 and 3 were created by branching on
x
1
.

24
The Branch-and-Bound Method
We now choose any subproblem that has not
yet been solved as an LP.
We arbitrarily choose to solve subproblem 2.
From Figure above, we see that the optimal
solution to subproblem 2 is z = 41, x
1
= 4, x
2
=
9/5 (point C).
Our accomplishments to date are summarized in
the Figure below.

25
Telfa Subproblems 1 and 2 Solved
26
The Branch-and-Bound Method
A display of all subproblems that have been created
is called a tree.
Each subproblem is referred to as a node of the
tree, and each line connecting two nodes of the
tree is called an arc.
The constraints associated with any node of the
tree are the constraints for the LP relaxation plus
the constraints associated with the arcs leading
from subproblem 1 to the node.
The label t indicates the chronological order in
which the subproblems are solved.

27
The Branch-and-Bound Method
The optimal solution to subproblem 2 did not
yield an all-integer solution, so we choose to use
subproblem 2 to create two new subproblems.
We choose a fractional valued variable in the
optimal solution to subproblem 2 and then
branch on that variable.

28
The Branch-and-Bound Method
Because x
2
is the only fractional variable in the
optimal solution to subproblem 2, we branch on
x
2
. We partition the feasible region for
subproblem 2 into those points having x
2
2
and x
2
1.
This creates the following two subproblems:
Subproblem 4 = Subproblem 1 + x
1
4 and x
2
2
= Subproblem 2 + x
2
2.
Subproblem 5 = Subproblem 1 + x
1
4 and x
2
1
= Subproblem 2 + x
2
1.

29
Feasible Regions Subproblems 4, 5
30
Telfa Subproblems 1, 2, and 4 Solved
31
Feasible Regions Subproblems 4, 5
32
Subproblems 1, 2, 4, and 5 Solved
33
Feasible Regions Subproblems 6,7
34
After Six Subproblems
35
Solve Sub-problem 3
36
Final Branch-and-Bound Tree
37
Fathoming
If it is unnecessary to branch on a subproblem,
then it is fathomed. The following three
situations result in a subproblem being
fathomed:
The subproblem is infeasible;
The subproblem yields an optimal solution in which
all variables have integer values; and
The optimal z-value for the subproblem does not
exceed (in a max problem) the current LB.

38
Another Example
39
Solving MIP Problems
40
Modify the method described by branching only
on variables that are required to be integers.
Optimum solution needs integer values for only
integer variables.
For Example
Max z = 2X
1
+ X
2

s.t. 5X
1
+ 2X
2
8
X
1
+ X
2
3
X
1
, X
2
0, X
1
integer


Branch-and-Bound tree for Mixed IP
41
Solving BIP Problems
42
Lecture 6
Multi-Objective Linear
Programming
43
Introduction
Consider the impact of multiple, and conflicting
objectives
Mathematical representation for a multi-
objective programming problem:
Minimize (or maximize) set of objectives:

Subject to set of constraints
=
1
,
2

() = 0 () 0
44
Introduction
Multiple-objective techniques :
Weighting or utility methods.
Ranking or prioritizing methods.
Efficient solution (or generating) methods.

Goal programming approach (a ranking or
prioritizing method) is the technique under
study.
45
Introduction
Reasons for goal programming:
The model development is relatively simple and
straightforward.
Minor modifications may be employed so as to
encompass the alternative approaches.
The method of solution is quite simple and is, in fact, just
a refinement to the two phase Simplex method.
The goal programming models, and variations, have
already found extensive implementation in actual
problems since the early 1950s.
The model and its assumptions seem consistent with
typical real-world problems.

46
Terminology
Efficient solution
Results form the model and it is called with that name
because it may not be optimum with respect to all the
conflicting objectives of the problem.
Objective
An objective is a relatively general statement that reflects
the desires of the decision maker.
Aspiration level
Is a specific value associated with a desired or acceptable
level of achievement of an objective.
Used to measure the achievement of an objective and
generally to anchor the objective to reality.

47
Terminology Cont.
Goal
An objective in conjunction with an aspiration
level is termed a goal.

Goal deviation
The difference between what we accomplish
and what we aspire to is the deviation from
our goal.

48
Goal formulation





Goal Type Goal Programming Form Deviation Variables to Be Minimized

f
i
(x)= mathematical representation of objective i as a function of
the decision variables .

b
i

= value of the aspiration level associated with objective i.

i
=positive deviation.

i
=negative deviation
49
Goal formulation (Contd)
Example
Let us assume that the profit of a firm may be
expressed as a linear function of two variables:

If our aspiration level is to obtain at least $1,000 of
profit per time period, the goal maybe written as:

Adding negative and positive deviation variables

1
should be minimized.



5
1
+ 7
2
=
5
1
+ 7
2
1000
5
1
+ 7
2
+

= 1000
50
The Achievement Function
The achievement function answers the question of:
How good the solution is?
Some of the measures used to evaluate the goodness of a solution
include:
How well does it minimize the sum of weighted goal deviations?
How well does it minimize some polynomial (or other nonlinear)
form of the goal deviations?
How well does it minimize the maximum (worst) goal deviation?
How well does it lexicographically minimize an ordered (i.e.
ranked or prioritized) set of goal deviations?
Various combinations of the above.
51
The Achievement Function (Contd)
We shall measure the achievement in terms of the
lexicographic minimization of an ordered set of goal
deviations, wherein within each set of goals at a
particular rank, weights may be used.
Lexicographical minimum
Given an ordered array a of nonnegative elements a
k
s,
the solution given by a
(1)
is preferred to a
(2)
if a
k
(1)
< a
k
(2)

and all higher-order elements (i.e. a
1
, a
2
, a
k-1
) are
equal. If no other solution is preferred to a, then a is the
lexicographic minimum.
52
Model Construction
The steps in the construction of the model are
summarized then as:
Step 1: Formulate the baseline model.
Step 2: Specify aspiration levels for each and every
objective.
Step 3: Include negative and positive deviation
variables for each and every goal and constraint.
Step 4: Rank the goals in terms of importance.
Priority 1 is always reserved for the rigid constraints.
Step 5: Establish the achievement function.
53
Model Construction (Contd)
Example
A small cosmetics firm until recently made only nail
polish remover. However, when an employee
accidentally dropped a jar of peanut butter into the
remover, it was found that the resultant mixture
could temporarily remove facial wrinkles.
Thus, the firm now produces two product: Wipe out
(the nail polish remover), and Ageless (The wrinkle
remover). The new, improved formulae for each
product require differing amounts of two base
chemicals (whose names are a company secret) as
shown in the coming.
54
Model Construction (Contd)
Product
Profit per
Gallon
Pounds of
Chemical A
Required per
Gallon
Pounds of
Chemical B
Required per
Gallon
Pounds of
Peanut Better
per Gallon
Ageless
wrinkle
remover
80 4 4 1
Wipe out nail
polish
remover
100 5 2 0
Amounts
available per
day (pounds)
- 80 48 6
55
Model Construction (Contd)
The daily availability of these two chemicals
cannot be exceeded, as there is only one
supplier, who is producing at maximal capacity.
Note that, although the daily supply of peanut
butter is unlimited, the owner does not want to
purchase more than 6 pounds per day, so as to
keep secret the fact that this ingredient is going
into production.

56
Model Construction (Contd)
When the company president was questioned, the
following facts were determined:
The daily availability of chemicals A and B cannot, in any
way, be exceeded. That is, these are the rigid constraints.
The firm would like to maintain daily profits at a level
above $800.
The daily amount of peanut butter ordered should be
kept under 6 pounds (that way, it may be disguised as
shipments to the company cafeteria).
The total number of gallons produced per day of both
products should be minimized so as to simplify shipping
and handling.
First, we start by constructing the baseline model:
57
Model Construction (Contd)
Decision variables:
X
1
= gallons per day of Ageless produced
X
2
= gallons per day of Wipe out produced
Goals and constraints:
4X
1
+ 5X
2
< 80 (1)
4X
1
+ 2X
2
< 48 (2)
80X
1
+ 100X
2
> 800 (3)
X
1


< 6 (4)
Minimize X
1
+ X
2
(5)
X
1
, X
2
> 0


58
Model Construction (Contd)
The only relation for which we need an aspiration level is
(5). We shall assume that the aspired level is 7 gallons per
day.
Now we can form our linear goal programming model as:
Find X
1
and X
2
so as to lexicographically minimize
a = {(
1
+
2
), (
3
), (
4
), (
5
)}
Subject to
4X
1
+ 5X
2
+
1
-
1
= 80
4X
1
+ 2X
2
+
2
-
2
= 48
80X
1
+ 100X
2
+
3

3
= 800
X
1
+
4
-
4
= 6
X
1
+ X
2
+
5
-
5
= 7
X , , > 0

59
Methods of solution
There are 2 basic approaches to the
solution of the linear goal programming
model with an achievement function that
is to be lexicographically minimized.
Graphical method
Sequential Linear Goal Programming (SLGP)
60
Graphical Method
Find X
1
and X
2
so as to lexicographically
minimize
a = {(
1
+
2
), (
3
), (
4
), (
5
)}
Subject to,
G1: 4X
1
+ 5X
2
+
1
-
1
= 80
G2: 4X
1
+ 2X
2
+
2
-
2
= 48
G3: 80X
1
+ 100X
2
+
3

3
= 800
G4: X
1
+
4
-
4
= 6
G5: X
1
+ X
2
+
5
-
5
= 7
X, , > 0
61
Graphical Method (Contd)
G1: 4X
1
+ 5X
2
+
1
-
1
= 80
G2: 4X
1
+ 2X
2
+
2
-
2
= 48
G3: 80X
1
+ 100X
2
+
3

3
= 800
G4: X
1
+
4
-
4
= 6
G5: X
1
+ X
2
+
5
-
5
= 7
62
Graphical Method (Contd)
G1: 4X
1
+ 5X
2
+
1
-
1
= 80
G2: 4X
1
+ 2X
2
+
2
-
2
= 48
G3: 80X
1
+ 100X
2
+
3

3
= 800
G4: X
1
+
4
-
4
= 6
G5: X
1
+ X
2
+
5
-
5
= 7
63
Graphical Method (Contd)
G1: 4X
1
+ 5X
2
+
1
-
1
= 80
G2: 4X
1
+ 2X
2
+
2
-
2
= 48
G3: 80X
1
+ 100X
2
+
3

3
= 800
G4: X
1
+
4
-
4
= 6
G5: X
1
+ X
2
+
5
-
5
= 7
64
Graphical Method (Contd)
G1: 4X
1
+ 5X
2
+
1
-
1
= 80
G2: 4X
1
+ 2X
2
+
2
-
2
= 48
G3: 80X
1
+ 100X
2
+
3

3
= 800
G4: X
1
+
4
-
4
= 6
G5: X
1
+ X
2
+
5
-
5
= 7
65
Graphical Method (Contd)
G1: 4X
1
+ 5X
2
+
1
-
1
= 80
G2: 4X
1
+ 2X
2
+
2
-
2
= 48
G3: 80X
1
+ 100X
2
+
3

3
= 800
G4: X
1
+
4
-
4
= 6
G5: X
1
+ X
2
+
5
-
5
= 7
66
Sequential Linear Goal Programming (SLGP)
Lexicographically minimize
a = {(
1
+
2
), (
3
), (
4
), (
1
+ 1.5
2
)}
Subject to,
X
1
+
1
-
1
= 30
X
2
+
2
-
2
= 15
8X
1
+ 12X
2
+
3

3
= 1000
X
1
+ 2X
2
+
4

4
= 40
X, , > 0.

67
SLGP Example Level 1
We begin by formulating the model
associated with priority level 1 only.
Minimize a
1
= {(
1
+
2
)}
Subject to,
X
1
+
1
-
1
= 30
X
2
+
2
-
2
= 15

68
SLGP Example Level 1 (Contd)
C
N
0 0 1 1
C
B
V X
1
X
2

1

2
X
B

0
1
1 0 -1 0 30
0
2
0 1 0 -1 15
0 0 -1 -1 0
Initial tableau for priority level 1.
X
1
= 0
1
= 30
2
= 0
X
2
= 0
2
=15
1
= 0
a
1
*=
1
+
2
= 0
This initial tableau is
also optimal, and thus
the solution to the first
level is:
69
SLGP Example Level 2
We now move to the formulation of priority
level 2, which is given as,
Minimize a
2
=
3

Subject to,
X
1
+
1
-
1
= 30
X
2
+
2
-
2
= 15
8X
1
+ 12X
2
+
3

3
= 1000

1
+
2
= a*
1
= 0
X, , > 0

70
SLGP Example Level 2 (Contd)
C
N
0 0 0
C
B
V X
1
X
2

3
X
B

0
1
1 0 0 30
0
2
0 1 -1 15
1
3
8 12 -1 1000
8 12 -1 1000
Initial tableau for priority level 2
71
SLGP Example Level 2 (Contd)
V X
1

2

3
X
B

1
1 0 0 30
X
2
0 1 0 15

3
8 -12 -1 820
8 -12 -1 820
Second tableau for priority level 2
72
SLGP Example Level 2 (Contd)
V X
1

2

3
X
B

X
1
1 0 0 30
X
2
0 1 0 15

3
- 8 -12 -1 580
- 8 -12 -1 580
Third and optimum tableau for priority level 2.
X
1
= 30
1
= 0
3
= 0
X
2
= 15
2
=0
3
= 580
a
2
*=
3
= 580
The solution is, thus, for
level 2:
73
SLGP Example Level 3
Note that the indicator row elements under the
non-basic variables
1
,
2
, and
3
are all negative in
this optimal tableau.
They should degrade the previous solution for level
2.
Column dropping rule: Any non-basic variable that
has a negative indicator row value in the optimal
tableau may be dropped (and its corresponding
column in the tableau dropped) from the problem,
as the introduction of such a variable would
degrade the solution.
74
SLGP Example Level 3 (Contd)
Thus,
1
,
2
, and
3
may all be dropped, giving
rise to the following formulation for level 3:
Minimize a =
4

Subject to,
X
1
= 30
X
2
= 15
8X
1
+ 12X
2
+
3
= 1000
X
1
+ 2X
2
+
4

4
= 40

3
= 580
75
SLGP Example Level 3 (Contd)
The first four constraints are those associated
with levels 1, 2, and 3, while the final constraint
is added to assure that any solution will not
degrade the result obtained for priority 2.
The foregoing problem is trivial, as the values of
X
1
and X
2
are fixed by the first two constraints.
That is, the solution is:
X
1
= 30
3
= 580
4
= 20
X
2
= 15
4
=0
a
3
* =
4
= 20
76
SLGP Example Level 4
Moving to priority level 4, note that,
a
4
=
1
+ 1.5
2

But
1
and
2
have already been dropped (i.e.
set to zero) in the problem. Thus, it is
unnecessary to formulate or solve the model for
level 4 as the solution is fixed:

X
1
*= 30
X
2
* = 15
a
4
* = (0, 580, 20, 0)
77
Lecture 7
Inventory Models
78
Introduction
Inventory is one of the most expensive assets of
many companies, representing as much as 40%
of total invested capital.
A firm can reduce costs by reducing on-hand
inventory levels.
Production may stop and customers become
dissatisfied when an item is out of stock
79
Introduction Cont.
All organizations have some type of inventory
planning and control system
Bank
Hospital
Government agencies
Schools
In inventory management
Decide whether to produce goods or to purchase them.
Forecast demand.
Determine the inventory necessary to service that
demand
80
Functions of Inventory
To provide a stock of goods to meet anticipated
customer demand and provide a selection of
goods.
To decouple suppliers from production and
production from distribution.
To take advantage of quantity discounts,
because purchases in larger quantities may
reduce the cost of goods or delivery.
81
Functions of Inventory Cont.
To hedge against inflation and upward price
changes.
To protect against delivery variation due to
weather, supplier shortages, quality problems,
or improper deliveries.
To permit operations to continue smooth/v with
the use of work-in-process inventory
82
Types of Inventory
To accommodate the functions of inventory,
firms maintain four types of inventories:
Raw material inventory
Materials that purchased but not processed
Work-in-process inventory
Components or raw material that has undergone some
change but is not completed
83
Types of Inventory Cont.
Maintenance/repair/operating supply (MRO)
inventory
Inventories devoted to maintenance/repair/operating
supplies necessary to keep machinery and processes
productive
Finished goods inventory.
Finished goods inventory is completed product awaiting
shipment.
May be inventoried because future customer demands are
unknown.

84
ABC Analysis
Also called Pareto analysis
It discuss how inventory items can be classified
The Pareto principle states that there are a
critical few and trivial many.
The idea is to establish inventory policies that
focus resources on the few critical inventory not
the many trivial ones.
85
ABC Analysis Cont.
Class A items are those on which the annual dollar
volume is high.
Represent about 15% of the inventory.
Represent 70% to 80% of the total dollar usage.
Class B items inventory items of medium annual
dollar volume.
Represent about 30% of inventory.
Represent 15% to 25% of the total value
Those with low dollar volume are Class C
Represent only 5% of the annual dollar
Represent about 55% of the total inventory items.
86
ABC Analysis Cont.
87
Inventory Models
Inventory cost elements
Holding Costs
Are the costs associated with holding or carrying inventory
over time. (Space, lighting, ..)
Ordering/Setup Costs
Include costs of supplies, forms, order processing, clerical
support and so forth. (Phone call, transportation cost,.)
Is a part of what is called setup costs.
Setup cost is the cost to prepare a machine or pr for
manufacturing an order.
88
Inventory Models Cont.
Two main types of inventory models:
Deterministic Inventory Models
Basic economic order quantity (EOQ) model.
Production order quantity model.
Quantity discount model.
Probabilistic Models with Constant Lead Time

89
EOQ Model
Basic economic order quantity (EOQ) model
Is one of the oldest and most commonly known
inventory-control techniques.
Model assumptions
Demand is known, constant, and independent.
Lead timethat is, the time between placement and receipt of
the orderis known and constant.
Receipt of inventory is instantaneous and complete.
Quantity discounts are not possible.
The only variable costs are the cost of setting up or placing an
order and the cost of holding or storing inventory over time .
Stock-outs (shortages) can be completely avoided if orders are
placed at the right time.
90
EOQ Model Cont.
91
EOQ Model Cont.
The objective of most inventory models is to
minimize total costs
92
EOQ Model Cont.
Q = Number of pieces per order
Q* = Optimum number of pieces per order (EOQ)
D = Annual demand in units for the inventory item
S = Setup or ordering cost for each order
H = Holding or carrying cost per unit per year
L = Lead time
D = Slope (units/day)
93
EOQ Model Cont.
Annual setup cost

Annual holding cost

Annual Purchase cost

The Total Annual cost

EOQ


=

2

=
=

2
+

94
EOQ Model Cont.
Reorder Points (ROP)
It answer when to order question
=
=

95
EOQ Model Cont.
Example
Company markets painless hypodermic needles
to hospitals would like to reduce its inventory
cost by determining the optimal number of
hypodermic needles to obtain per order. The
annual demand is 1000 units; the setup or
ordering cost is $10 per order: and the holding
cost per unit per year is $.50.
96
EOQ Model Cont.
r units/orde 200
0.5
) 2(1000)(10
H
2DS
Q* = = =
= =

= =

= 1000/2200 = 5
= 250/5 = 50
= (1000/200) (5) + (200/2) (0.50) = 50 + 50 = $ 100
97
EOQ Model Cont.
Notice that we ignored the purchase price here in
calculating the total annual costs as they are
constant for any ordering policy, except for the case
of quantity discount models which we will discuss
later on
If we assume that the annual demand and the price
per hypodermic needle are known values (for
example, 1,000 hypodermics per year at P = $10)
and annual cost should include purchase cost,

= 50 + 50 + (1000 10) = $ 10,100
98
POQ Model
Production Order Quantity (POQ) Model
Consider that the firm may receive its inventory over
a period of time
This model is applicable under two situations:
When inventory continuously flows or builds up over a
period of time after an order has been
When units are produced and sold simultaneously.
99
POQ Model Cont.

100
POQ Model Cont.
Q = Number of pieces per order (run size)
H = Holding cost per unit per year
p = Daily production rate
d = Daily demand rate, or usage rate
t = Length of the production run in days

101
POQ Model Cont.
Annual inventory holding cost

Average Inventory Level

Maximum inventory level


=
= /2
= 1

102
POQ Model Cont.
The Total Annual cost


Optimum run size

= PD )
p
d
H(1
2
Q
S
Q
D
+ +

2
1

103
POQ Model Cont.
Example
Nathan Manufacturing Inc. makes and sells
specialty hubcaps for the retail automobile
aftermarket. Nathans forecast for its wire wheel
hubcap is 1000 units next year, with an average
daily demand of 4 units. However the production
process is most efficient at 8 units per day. So the
company produces 8 per day but uses only 4 per
day. Given the following values, solve for the
optimum number of units per order (Note This
plant schedules production of this hubcap only as
needed about 250 days per year)
104
POQ Model Cont.
Annual Demand = D = 1000 units per year
Setup Cost = $10/setup
Holding Cost = H = $0.50 per unit per year
Daily production rate = p = 8 units per day
Daily Demand rate = 4 units per day

2
1

1000

(10)
0.50 1
4
8

105
Quantity Discount Model
Quantity Discount Models
Increase sales by offering quantity discounts
106
Quantity Discount Model Cont.
The equation for total annual inventory cost:
Total cost = Setup cost + Holding cost + Product cost


Where:
Q = Quantity ordered
D = Annual demand in units
S = Ordering or setup cost per order or per setup
P = Price per unit
H = Holding cost per unit per year

107
Quantity Discount Model Cont.
Quantity Discount Models
Steps
Step 1: For each discount, calculate a value for optimal order
size Q*.
Step 2: For any discount, if the order quantity is too low to
qualify for the discount, adjust the order quantity upward to
the lowest quantity that will qualify the discount.
Step 3: Using the total cost equation, compute a total cost
for every Q* determined in steps 1 and 2.
Step 4: Select the Q* that has the lowest total cost, as
computed in step 3.
108
Quantity Discount Model Cont.
Example
A Discount Store stocks toy race cars. Recently, the store
has been given a quantity discount schedule for these
cars (The one shown in the previous table). The normal
cost for the toy race cars is $5.00. For orders between
1,000 and 1,999 units, cost drops to $4.80; for orders of
2,000 or more units, the unit cost is only $4.75.
Furthermore, ordering cost is $49.00 per order. Annual
demand is 5,000 race cars, and inventory carrying charge,
as a percentage of cost, labeled I, is 20% or 0.2. What
order quantity will minimize the total inventory cost?
109
Quantity Discount Model Cont.
Step 1

1
=

5000

49

0.2

5.00

= 700 /

2
=

5000

49

0.2

4.80

= 714 /

3
=

5000

49

0.2

4.75

= 718 /
110
Quantity Discount Model Cont.
Step 2


Step 3


Step 4
Q
1
* = 700
Q
2
* = 1000 adjusted
Q
3
* = 2000 adjusted
TC
1
= $ 25,700
TC
2
= $ 24,725
TC
3
= $ 24,822.50
Q
2
* = 1000 units.
111
Probabilistic model
Probabilistic models with constant lead time
Apply when product
demand is not known
but can be specified by
means of a probability
distribution
= +
112
Probabilistic model Cont.
Example
Rivera Optical has determined that its reorder
point for eyeglass frames is 50 (d X L) units. Its
carrying cost per frame per year is $5, and stock-
out (or lost sale) cost is $ 40 per frame. The
store has experienced the following probability
distribution for inventory demand during the
reorder period. The optimum number of orders
per year is six.
113
Probabilistic model Cont.
Number of Units Probability
30 0.2
40 0.2
50 0.3
60 0.2
70 0.1
1.0
114
Probabilistic model Cont.






Safety
Stock
Additional
Holding Cost
Stock-out Cost
Total
Cost
20 (20)*($5) = $100 $0 $ 100
10 (10)*($5) = $ 50 (10)*(0.1)*($40)*(6) = $240 $ 290
0 $ 0
(10)*(0.2)*($40)*(6) +(20)*(0.1)*($ 40)*(6) =
$ 960
$ 960
The safety stock with the lowest total cost is 20
frames. Therefore, the new reorder point = 50 + 20
= 70 frames.
115
Probabilistic model Cont.
Example
A regional Hospital stocks a code blue
resuscitation kit that has a normally distributed
demand during the reorder period. The mean
(average) demand during the reorder period is 350
kits, and the standard deviation is 10 kits. The
hospital administrator wants to follow a policy that
results in stock-outs occurring only 5% of the time.
What is the appropriate value of Z?
How much safety stock should the hospital maintain?
What reorder point should be used?
116
Probabilistic model Cont.
The following figure may help you visualize the
example:
117
Probabilistic model Cont.
(a) We use the properties of a standard normal
curve to get a Z value for an area under the
curve of 0.95 (or 1 - 0.05).Using a normal table,
we find a Z value of 1.65 standard deviations
from the mean.
(b) Safety Stock = x - because Z = (X - )/o
Then safety stock = Zo = 1.65 * 10 = 16.5 kits, which
is the situation for figure 6.

118
Probabilistic model Cont.
(c) The reorder point =
Expected demand during lead time + SS = 350 + 16.5
= 366.5 pr 367 kits.
Now, assumes that both an estimate of expected demand
during lead times and its standard deviation are available.
When data on lead time demand is not at hand, these
formulas cannot be applied and we need to determine if:
(a) demand is variable and lead time is constant; or
(b) only lead time is variable; or
(c) both demand and lead time are variable. For each of these
situations, a different formula as follows:

119
Probabilistic model Cont.
(a) If only the demand d is variable, then


(b) If only lead time is variable, then

(c) If both are variable, then


= +

= =

= ( ) +( )
=

+
2

2

120
Lecture 8
Queuing Theory
121
Introduction
Waiting in queues -waiting lines- is one of the most
common occurrences in everyone's life, companies
are able to reduce waiting time and provide faster
service by increasing their service capacity, which
usually means adding more servers.
However, increasing service capacity in this manner
has a monetary cost, and therein lies the basis of
waiting line analysis, the trade-off between the cost
of improved service and the cost of making
customers wait.

122
Introduction
Queuing analysis is a probabilistic form of analysis,
not a deterministic technique. Thus, the results of
queuing analysis, referred to as operating
characteristics, are probabilistic.
These operating statistics (such as the average time
a person must wait in line to be served) are used by
the manager of the operation containing the queue
to make decisions.

123
Waiting Line Analysis
Operating characteristics
average values for characteristics that describe the
performance of a waiting line system
Queue
A single waiting line
Waiting line system consists of
Arrivals
Servers
Waiting line structures

124
Elements of a Waiting Line
Calling population
Source of customers
Infinite - large enough that one more customer can always
arrive to be served
Finite - countable number of potential customers
Arrival rate ()
Frequency of customer arrivals at waiting line system
Typically follows Poisson distribution
125
Elements of a Waiting Line (cont.)
Service time
Often follows negative exponential distribution
Average service rate =
Arrival rate () must be less than service rate ()
or system never clears out
126
17-127
Elements of a Waiting Line (cont.)
Queue discipline
Order in which customers are served
First come, first served is most common
Length can be infinite or finite
Infinite is most common
Finite is limited by some physical
Basic Waiting Line Structures
Channels are the number of parallel servers
Single channel
Multiple channels
Phases denote number of sequential servers the customer must
go through
Single phase
Multiple phases
Steady state
A constant, average value for performance characteristics that system will
reach after a long time
128
Essential Features of Queuing Systems
Departure
Queue
discipline
Arrival
process
Queue
configuration
Service
process
Renege

Balk
Calling
population
No future
need for
service
129
Arrival Process
Static Dynamic
Appointments Price
Accept/Reject
Balking Reneging
Random
arrivals with
constant rate
Random arrival
rate varying
with time
Facility-
controlled
Customer-
exercised
control
Arrival
process
130
Queue Configuration

Multiple Queue Single queue
Take a Number
Enter
3 4
8
2
6 10
12
11
5
7
9
131
Queue Discipline

Queue
discipline
Static
(FCFS rule)
Dynamic
selection
based on status
of queue
Selection based
on individual
customer
attributes
Number of
customers
waiting
Round robin Priority Preemptive
Processing time
of customers
(SPT rule)
132
Service Facility Arrangements
Service facility Server arrangement
Parking lot Self-serve
Cafeteria Servers in series
Toll booths Servers in parallel
Supermarket Self-serve, first stage; parallel servers, second
stage
Hospital Many service centers in parallel and series, not all
used by each patient
133
Kendall Notation
A/S/NS/B/K/SD
A,S=Interarrival time, service time distribution
M = Exponential
C = Constant
E
k
= Erlang
H
k
= Hyperexponential
D = Deterministic
NS=Number of servers
B=Number of Buffers
K=Population size

134
Kendall Notation
SD=Service Discipline
FCFS,FCLS
Defaults B= , K= SD=FCFS
M/M/1 = M/M/1/ / /FCFS

135
Multi-server queue
136
Multiple Single-Server Queues
137
Operating Characteristics
NOTATION OPERATING
CHARACTERISTIC
L Average number of customers in the system
(waiting and being served)
L
q
Average number of customers in the waiting line
W Average time a customer spends in the system
(waiting and being served)
W
q
Average time a customer spends waiting in line

P
0
Probability of no (zero) customers in the system
P
n
Probability of n customers in the system
Utilization rate; the proportion of time the system is in use
138
Queuing System Design Objective
Cost Relationship in Waiting Line Analysis
Service capacity
Optimum
Cost of service
capacity
customer
waiting Cost
Total cost
C
o
s
t

Total
cost
Customer
waiting cost
Capacity
cost
= +
139
Waiting Time vs. Service System Utilization
System Utilization
A
v
e
r
a
g
e

n
u
m
b
e
r

o
r

t
i
m
e

w
a
i
t
i
n
g

i
n

l
i
n
e

0
100%
140
Single-Server Models
All assume Poisson arrival rate
Variations
Exponential service times
General (or unknown) distribution of service times
Constant service times
Exponential service times with finite queue length
Exponential service times with finite calling
population

141
Single-Server Model Assumptions
Poisson arrival rate
Exponential service times
First-come, first-served queue discipline
Infinite queue length
Infinite calling population
= mean arrival rate
= mean service rate

142
Waiting Line Performance Measures for M/M/1
Average queue time, W
q
= - W = / ( - )
Average time in system, W = 1 / -
Average queue length, L
q
= - L =
2
/ ( - )
Average number of customers in system, L = / -
Average system utilization rate, = /
Probability of idle service facility, P
0
= 1 - = 1 /
Probability of exactly n customers in system, P
n
= (1 - )
n

=(1 / ) ( / )
n

143
M/M/1 Example
Drive-up window at a fast food restaurant.
Customers arrive at the rate of 25 per hour. The
employee can serve one customer every two
minutes. Assume Poisson arrival and exponential
service rates.
A) What is the average utilization of the employee?
B) What is the average number of customers in line?
C) What is the average number of customers in the system?
D) What is the average waiting time in line?
E) What is the average waiting time in the system?
F) What is the probability that exactly two cars are in the
system?

144
M/M/1 Example
A) What is the average utilization of the employee?

= 25 cust / hr
=
1 customer
2 mins (1hr / 60 mins)
= 30 cust / hr
= =
25 cust / hr
30 cust / hr
= .8333
145
M/M/1 Example
B) What is the average number of customers in line?



C) What is the average number of customers in the system?






4.167 =
25) - 30(30
(25)
=
) - (
=
2 2

q L
5 =
25) - (30
25
=
-
=

L
Note: L= L
q
+ = 4.167 + 0.833
146
M/M/1 Example
D) What is the average waiting time in line?

mins 10 = hrs .1667 =
25) - 0 3 0( 3
25
=
) - (
=

q W
E) What is the average waiting time in the system?
mins 12 = hrs .2 =
25 - 30
1
=
-
1
=

W
147
M/M/1 Example
F) What is the probability that exactly two cars are in
the system?
n
) )( - (1 = P
n

.1157 = )
30
25
)(
30
25
- (1 = P
2
2
148
Waiting Lines - Summary
Queuing System Design Rules
Tradeoff between capacity cost and cost of waiting
Increasing average utilization (=/) may cause
very long waiting times
Model Characteristics
Arrival process distribution
Service process distribution and discipline
Number of servers
Queue capacity/length
Model Presented
M/M/1

149
Constant Service Times
Constant service times
occur with machinery
and automated
equipment
Constant service times
are a special case of the
single-server model with
undefined service times

150
Constant Service Times


P
0
= 1 -
Probability that no customers
are in system
Average number of
customers in system
L

= L
q
+


Average number of
customers in queue
L
q
=

2

2( - )
151
Constant Service Times (Contd)


=
Probability that the
server is busy
Average time customer
spends in the system
W = W
q
+
1

Average time customer
spends in queue
W
q
=

L
q


152
M/C/1 Example
Assume a drive-up window at a fast food
restaurant. Customers arrive at the rate of 25 per
hour. The employee can serve one customer every
two minutes. Assume variable arrivals and a
constant service rate.
Determine:
A) What is the average utilization of the employee?
B) What is the average number of customers in line?
C) What is the average number of customers in the
system?
D) What is the average waiting time in line?

153
M/C/1 Example
A) What is the average utilization of the employee?

.8333 =
cust/hr 30
cust/hr 25
= =
cust/hr 30 =
mins) (1hr/60 mins 2
customer 1
=
cust/hr 25 =

154
M/C/1 Example
B) What is the average number of customers in line?


C) What is the average number of customers in the system?


D) What is the average waiting time in line?

2.0833 =
25) - (2)(30)(30
(25)
=
) - ( 2
=
2 2

Lq
mins 5 = hrs .08333 =
25
08333 . 2
=

=

Lq
Wq
mins 7 = hrs .116667 =
30/hr
1
+ hrs .08333 =

1
+ =

Wq Ws
155
Finite Queue Length
A physical limit exists on length of queue
M = maximum number in queue
Service rate does not have to exceed arrival rate
( > ) to obtain steady-state conditions

156
Finite Queue Length

P
0
=
Probability that no
customers are in system
1 - /
1 - (/)
M + 1

Probability of exactly n
customers in system


P
n
= (P
0
) for n M
n
L = -
Average number of
customers in system
/
1 - /
(M + 1)(/)
M + 1

1 - (/)
M + 1

157
Finite Queue Length

Let P
M
= probability a customer will not join system
Average time customer
spends in system
W

=
L
(1 - P
M
)
L
q
= L -
(1- P
M
)

Average number of
customers in queue
Average time customer
spends in queue
W
q
= W -
1

158
Finite Queue Length Example
First National Bank has waiting space for only 3
drive in window cars. = 20, = 30, M = 4 cars
(1 in service + 3 waiting)


159
Finite Queue Length Solution



Probability that no
cars are in the
system
P
0
= = = 0.38
1 - 20/30
1 - (20/30)
5

1 - /
1 - (/)
M + 1

P
n
= (P
0
) = (0.38) = 0.076
Probability of
exactly 4 cars in
the system
20
30
4


n=M
L = - = 1.24
Average number
of cars in the
system
/
1 - /
(M + 1)(/)
M + 1

1 - (/)
M + 1

160
Finite Queue Length Solution
Average time a car
spends in the system
W

= = 0.067 hr
L
(1 - P
M
)
L
q
= L - = 0.62
(1- P
M
)

Average number of
cars in the queue
Average time a car
spends in the queue
W
q
= W - = 0.033 hr
1

161
Finite Calling Population
Arrivals originate from a finite (countable)
population, N = population size

Probability of exactly n
customers in system
P
n
= P
0
where n = 1, 2, .., N
n


N!
(N - n)!
Average number of
customers in queue
L
q
= N - (1- P
0
)
+

Probability that no
customers are in system
P
0
=
n = 0
N!
(N - n)!
E
N
n


1
162
Finite Calling Population
W
q
=
L
q

(N - L)
Average time customer
spends in queue
L

= L
q
+ (1 - P
0
)
Average number of
customers in system
W

= W
q
+
Average time customer
spends in system
1

163
Finite Calling Population Example
20 trucks which operate an average of 200 days
before breaking down
( = 1/200 day = 0.005/day)
Mean repair time = 3.6 days
( = 1/3.6 day = 0.2778/day)

164
Finite Calling Population Solution
Probability that no
trucks are in the system
P
0
= 0.652
Average number of
trucks in the queue
L
q
= 0.169
Average number of
trucks in system
L = 0.169 + (1 - 0.652) = .520
Average time truck
spends in queue
W
q
= 1.74 days
Average time truck
spends in system
W = 5.33 days
165
Multiple-Server Model
Two or more independent servers serve a single waiting
line. Poisson arrivals, exponential service, infinite calling
population
s >




Computing P
0
can be time-consuming.
Tables can used to find P
0
for selected values of and s.


P
0
=
1
1
s!


s
s
s -
E
n=s-1
n=
0
1
n!


n
+
166
Multiple-Server Model
Probability of exactly
n customers in the
system
P
n
=
P
0
, for n >
s
1
s! s
n-s


n
P
0
, for n > s
1
n!



n
Probability an arriving
customer must wait
P
w
= P
0
1
s!
s
s -
s


Average number of
customers in system
L

= P
0
+
(/)
s

(s - 1)!(s - )
2



167
Multiple-Server Model
W

=
L

Average time customer
spends in system

= /s
Utilization factor
Average time customer
spends in queue
W
q
= W - =
1

L
q


L
q
= L
-


Average number of
customers in queue
168
Multichannel (M/M/S) Model Characteristics
Type: Multichannel system
Input source: Infinite; no balks, no reneging
Arrival distribution: Poisson
Queue: Unlimited; multiple servers
Queue discipline: FIFO (FCFS)
Service distribution: Negative exponential
Relationship: Independent service & arrival
E Service rates > arrival rate
169
Multiple-Service Model Example
Student Health Service Waiting Room
= 10 students per hour
= 4 students per hour per service representative
s = 3 representatives
s = (3)(4) = 12

170
Multiple-Service Model Solution
P
0
= 0.045 Probability no students are in the system
Number of students in the service area
L = 6
L
q
= L - / = 3.5
Number of students waiting to
be served
Average time students will wait in
line
W
q
= L
q
/ = 0.35 hrs
Probability that a student must wait
P
w
= 0.703
Waiting time in the service area
W = L / = 0.60
171
Multiple-Service Model Solution
Add a 4th server to improve service
Recompute operating characteristics
P
0
= 0.073 prob of no students
L = 3.0 students
W = 0.30 hour, 18 min in service
L
q
= 0.5 students waiting
W
q
= 0.05 hours, 3 min waiting, versus 21 earlier
P
w
= 0.31 prob that a student must wait

172
l ecture 9
Basic Simulation Modeling
173
What is Simulation?
It refers to a broad collection of methods and
applications to mimic the behavior of real
systems.
174
Examples of Modeled Systems
A manufacturing plant with machines, people.
Transport devices, conveyor belts and storage
systems.
A bank with different kinds of customers, servers,
and facilities like teller windows, automated teller
machines (ATMs), and safety deposit boxes.
A distribution network of plants, warehouses, and
transportation links.
A supermarket with inventory control, checkout,
and customer service.

175
Ways of Studying a System
System
Experiment with the
Actual System
Experiment with a
Model of the System
Physical
Model
Mathematical
Model
Simulation
Analytical
Solution
176
When to use Simulation?
Studying complex systems and experimenting
them.
Valuable insight is needed by changing
simulation inputs and observing the resulting
outputs.
To experiment with new designs and policies
prior to implementation.
Verifying analytical solutions is needed.
Job learning without cost.
Planning visualizations.
177
When NOT to use Simulation?
If the problem can be solved analytically.
If the problem is easier to be solved by
direct experiments.
If simulation costs more that it saves.
If resources or time is not available.
If systems behavior is too complex or cant
be defined.
178
Simulation Advantages
Experimenting with the real system is too expensive or
sometimes impossible.
Experiments are made without disrupting the real world
system.
Studying a certain parameter over a long period of time.
Better understanding of systems.
Easy manipulation of models.
Bottle-neck analysis.
179
Simulation Disadvantages
Model building requires special
training.
Simulation results may be difficult to
interpret.
Simulation modeling and analysis can
be time consuming and expensive.
180
Types of Systems
Discrete-system state variable.
181
Types of Systems (Contd)
Continuous-system state variable.
182
Systems Components
Entity: object of interest in the system.
Attribute: property of an entity.
Activity: represents a time period of specified
length.
The state of a system: collection of variables
necessary to describe the system at any time,
relative to the objectives of the study.
An event: instantaneous occurrence that may
change the state of the system.

183
Examples of System s and Components
System Entities Attributes Activities Events State Variables
Production Machines
Speed;
capacity;
breakdown
rate
Welding;
stamping
Breakdown
Status of machines
(busy, idle or down)
Inventory Warehouse Capacity Withdrawing Demand
Levels of inventory;
backlogged
demands
Banking Customers
Checking
account
balance
Making
deposits
Arrival;
departure
Number of busy
tellers; queue length
184
Simulation Models
There are a lot of ways to classify
simulation models, but one useful
way is along these three
dimensions:
Static vs. Dynamic
Deterministic vs. Stochastic
Continuous vs. Discrete

185
Simulation Methods
First simulation used to be performed by
hand, then evolved to various types of
computer simulation packages, as follows:
1. By Hand
2. Programming in General-Purpose Languages
3. Simulation Languages
4. High-Level Simulators

186
Example of Simulation by Hand
Customer arrival and service times in a bank (I/P):

Customer Number Arrival Time (Minutes) Service Time (Minutes)
1 3.2 3.8
2 10.9 3.5
3 13.2 4.2
4 14.8 3.1
5 17.7 2.4
6 19.8 4.3
7 21.5 2.7
8 26.3 2.1
9 32.1 2.5
10 36.6 3.4
187
Example of Simulation by Hand (Contd)
Manual Simulation Outputs:
Customer
Number (n)
Arrival
Time (2)
Service
Time (3)
Departure
Time (4)
Time in Queue
= 4
n-1
- 2
n

Time in Bank
= 4
n
-2
n

1 3.2 3.8 7.0 0.0 3.8
2 10.9 3.5 14.4 0.0 3.5
3 13.2 4.2 18.6 1.2 5.4
4 14.8 3.1 21.7 3.8 6.9
5 17.7 2.4 24.1 4.0 6.4
6 19.8 4.3 28.4 4.3 8.6
7 21.5 2.7 31.1 6.9 9.6
8 26.3 2.1 33.2 4.8 6.9
9 32.1 2.5 35.7 1.1 3.6
10 36.6 3.4 40.0 0.0 3.4
188
Example of Simulation by Hand (Contd)
Event-oriented description of bank teller simulation.

189
Example of Simulation by Hand (Contd)
Graphic portrayal of bank teller simulation.

190
Steps in a Computer Simulation Study
1. Problem Formulation
2. Model Building
3. Model Translation
4. Verification
5. Validation
6. Experimentation
7. Analysis of Results
8. Implementation and Documentation

191