23+ +linear+programming
TRANSCRIPT
-
8/3/2019 23+ +Linear+Programming
1/28
Quantitative Methods - 2010
Linear Programming (LP)
1QM 2010 - Transcend
George Dantzig
Father of Linear Programming
-
8/3/2019 23+ +Linear+Programming
2/28
Preliminary
Programming in the OR sense means to find
the values for several variables. (Not to beconfused with computer coding).
LP is part of a larger field called MathematicalProgramming (MP)
Objective of LP is to optimize (maximize orminimize) a function of several variables, subject
to constraints on these variables.
Huge variety of very large, practical problems canbe expressed and solved as LP problems
QM 2010 - Transcend 2
-
8/3/2019 23+ +Linear+Programming
3/28
Motivating Scenario
(Finance)
An NBFC is considering investing in shares of 2companies, A and B.
Current price per share: A: Rs.250, B: Rs.160
Expected annual returns per share are: A: Rs. 45
B: Rs. 37
NBFC internal limit for this industry: Rs. 10 L SEBI limits: A 3000 shares; B 4000 shares
Question: How many shares to buy in A and B?
QM 2010 - Transcend 3
-
8/3/2019 23+ +Linear+Programming
4/28
Motivating Scenario
(Advertizing) A marketing manager is considering 2 different
media, A and B, for promotion of a product.
Cost per insertion in each is known Medium A: Rs. 25,000; Medium B: Rs. 36,000
Expected no. of prospects per insertion Medium A: 7 L; Medium B: 9 L
Ad budget: Rs.350,000
Ad availability Medium A: 4 insertions; Medium B: 6 insertions
Question: How many insertions in A, B?
QM 2010 - Transcend 4
-
8/3/2019 23+ +Linear+Programming
5/28
Motivating Scenario
(Distribution)
Product availability (no. of cartons) at our
warehouses in
Sangli: 700; Lonavla: 450; Palghar: 800
Requirements at our customers in
Vapi: 630; Nasik: 750; Bombay: 570
Transport costs from each to each are known.
Question: How much to send from which
warehouse to which customer?
QM 2010 - Transcend 5
-
8/3/2019 23+ +Linear+Programming
6/28
General LP Problem
QM 2010 - Transcend 6
-
8/3/2019 23+ +Linear+Programming
7/28
General Structure
We have
n decision variables: x1 --- xn An objective function to be optimized involving the
decision variables: Z = c1x1 + --- +cnxn A set of constraints involving the decision variables:
a11x1 + a12x2 - - - +a1nxn b1 a21x1 + a22x2 - - - +a2nxn b2
- - - - - - - - - -
am1x1 + am2x2 - - - +amnxn bm
For solving, we have to add non-negativity constraints:
All x 0
QM 2010 - Transcend 7
-
8/3/2019 23+ +Linear+Programming
8/28
Structure for 2 Decision Variables
We have
2 decision variables: x1 and x2 An objective function to be optimized involving the 2
decision variables: Z = c1x1 + c2x2 A set of constraints involving the decision variables:
a11x1 + a12x2 b1 a21x1 + a22x2 b2
- - - - - - -
am1x1 + am2x2 bm
For solving, we have to add non-negativity constraints:
All x 0
QM 2010 - Transcend 8
-
8/3/2019 23+ +Linear+Programming
9/28
Graphical Representation(Maximizing Z - For 2 Decision Variables)
QM 2010 - Transcend 9
a11x1 + a12x2 b1
a21
x1
+ a22
x2 b
2
a31x1 + a32x2 b3
x1
x2
Z = c1x1 + c2x2
Feasible Region
Solution is x1 = x1*, and x2 = x2*
(x1*, x2*)
-
8/3/2019 23+ +Linear+Programming
10/28
Graphical Representation(Minimizing Z - For 2 Decision Variables)
QM 2010 - Transcend 10
x1
x2
(x1*, x2*)
Solution is x1 = x1*, and x2 = x2*
-
8/3/2019 23+ +Linear+Programming
11/28
Formulating A (2 Variable) LP Problem
Identify your decision variables
Let x1 = no of shares of A; x2 = no of shares of B
Write objective function
Maximize Z = 45x1 + 37x2 Write constraints
250x1 + 180x2 10 L ---- investment constraint
X1 + x2 4500 ----- SEBI constraint
Add non-negativity constraints
X1, x2 0
QM 2010 - Transcend 11
-
8/3/2019 23+ +Linear+Programming
12/28
Plotting The Constraints
Estimate range of variables and draw axes
X1: 0 to 4500, x2: 0 to 6000
Take any 2 arbitrary x1 in this range, write
constraints as equations, solve for x2 (1) 250(1000) + 160(x2) = 10 L; x2 = 4687
250(3000) + 160(x2) = 10 L; x2 = 1563
(2) x1 = 3000 (3) x2 = 4000
Plot the constraints lines
QM 2010 - Transcend 12
-
8/3/2019 23+ +Linear+Programming
13/28
Feasible Region For NBFC Problem
QM 2010 - Transcend 13
x1
x2
250(x1) + 160 (x2) = 10 L
X1 = 3000
X2 = 4000
0 1000 2000 3000 6000
0
10
00
2000
3000
6000
-
8/3/2019 23+ +Linear+Programming
14/28
-
8/3/2019 23+ +Linear+Programming
15/28
Objective Function
On Feasible region
QM 2010 - Transcend 15
x1
x2
250(x1) + 160 (x2) = 10 L
X1 = 3000
X2 = 4000
A
Z = 85,000
-
8/3/2019 23+ +Linear+Programming
16/28
Final Solution
Moving the Z line parallely up, we can see that the Zline thru point A will have a higher value than thru anyother point in the Feasible Region.
Calculate the coordinates of A:
Clearly, x2 of A = 4000 Put this value in the other line equation, to get x1 for A
250(x1) + 160(4000) = 10 L, so x1 = 3.60 L/250 = 1440
Find Z at A, by putting x1 = 1440, x2 = 4000 in Zequation Z = 45(1440) + 40(4000) = 224,800
Thus Max Z = 224,800, and solution is x1 = 1440, x2 =4000
QM 2010 - Transcend 16
-
8/3/2019 23+ +Linear+Programming
17/28
Illustrating Final Solution
QM 2010 - Transcend 17
x1
x2
250(x1) + 160 (x2) = 10 L
X1 = 3000
X2 = 4000
A
Z = 85,000
Z = 2,24,800
-
8/3/2019 23+ +Linear+Programming
18/28
Verification Of Final Solution
Point B has x1 = 3000 and x2 = 1563
Z at that point = 45(3000) + 40(1563) = 197,520
This less than the Z at A
All other points will have even less Z
So clearly Z at A is the maximum value, and
coordinates of A is the solution.
QM 2010 - Transcend 18
-
8/3/2019 23+ +Linear+Programming
19/28
Transportation Problem
(Special Case Of General LP Problem)
QM 2010 - Transcend 19
-
8/3/2019 23+ +Linear+Programming
20/28
General Structure
We have mxn table of m sources and n destinations. Table entries are transportation costs cij (per unit of some
item) from each destination to each source
Row totals Ai are availabilities at the sources
Column totals Rj are requirements at the destinations
We have mxn decision variables (how much to ship) : x11 --- xmn An objective function to be optimized involving the
decision variables: Z = c11x11 + --- +cmnxmn Constraints are shipments
Must not violate any availabilities
Must meet all requirements
QM 2010 - Transcend 20
-
8/3/2019 23+ +Linear+Programming
21/28
Transportation LP Table
QM 2010 - Transcend 21
Available
d1 d2 --- --- dn
s1 c11 c12 c1n A1s2 c21 A2
Sources --- ---
--- ---
sm cm1 cmn Am
Required R1 R2 --- --- Rm
Destinations
-
8/3/2019 23+ +Linear+Programming
22/28
Transportation LP Solution Table
QM 2010 - Transcend 22
Available
d1 d2 --- --- dn
s1 x11 x12 x1n A1
s2 x21 A2
Sources --- ---
--- ---
sm xm1 xmn Am
Required R1 R2 --- --- Rm
Destinations
X in rows must add up to Availability
X in columns must add up to Required
Some X can be zero.
-
8/3/2019 23+ +Linear+Programming
23/28
Try This Yourself
Cartons of milk powder are stored in 3 warehouses asfollows: Sangli: 700; Lonavla: 450; Palghar: 800
They are required at 3 customers as follows:
Vapi: 630; Nasik: 750; Bombay: 570 Transport costs per carton are as follows:
Suggest 3 distribution plans, and work out total cost foreach plan.
QM 2010 - Transcend 23
Vapi Nasik Bombay
Sangli 5 4 2
Lonavla 7 5 3
Palghar 2 8 6
-
8/3/2019 23+ +Linear+Programming
24/28
Assignment Problem
(Another Special Case Of General LP Problem)
QM 2010 - Transcend 24
-
8/3/2019 23+ +Linear+Programming
25/28
-
8/3/2019 23+ +Linear+Programming
26/28
Assignment Problem Table
QM 2010 - Transcend 26
m1 m2 --- --- mmj1 c11 c12 c1n
j2 c21
Jobs ---
---
jm cm1 cmn
Machines
-
8/3/2019 23+ +Linear+Programming
27/28
Assignment Problem Solution Table
QM 2010 - Transcend 27
m1 m2 --- --- mm
j1 x11 x12 x1n
j2 x21
Jobs ---
---
jm xm1 xmn
Machines
There can be only one 1 in each row
and in each column.
All other entries must be 0
-
8/3/2019 23+ +Linear+Programming
28/28
End Of
Linear Programming
QM 2010 - Transcend 28