You are on page 1of 13

Assignment 1

Due: Beginning of class on 1/21/2010

Guidelines for Submission: Late HW will not be accepted. All the working has to be
shown on the homework. Turn in a hard-copy of your work at the beginning of class on
1/21/2010. Provide clear and complete definition of decision variables. Write down all
constraints and explain the constraints. Whenever you use Excel, turn in a hard-copy of
the Answer and Sensitivity reports.

PROBLEM 1:
Solve the advertising problem mentioned in Chapter 2, Problem 6 using LP. Use
Graphical Method to solve this problem.
Solution:
Notice how the problem states linearity assumption in number of ads
placed. This decision is with management team on how it interprets the
fractional solutions like 3.5 ads. But sometimes, as we will see in
next topic “Integer Programming”, this interpretation does not hold
ground.

Decision Variables
• X = Number of weekday ads placed by PMA in month of August
• Y = Number of Sunday ads placed by PMA in month of August
Objective Function
• Maximize Total Exposure
Constraints
• Total advertising Budget
• Minimum number of ads to be placed in weekdays
• Minimum number of ads to be placed on Sundays
• Maximum number of weekdays that exist in August
• Maximum number of Sundays that exist in August
• Non-negativity

Writing out everything in simple words like above, for yourself can
help you in doing the mathematical part of formulation easy.

Part (a)
LP Formulation
Maximize 30000X + 80000Y
Subject to
2000X + 8000Y <= 40,000 (Budget)
X >= 8 (Minimum of 8 weekdays)
Y >= 2 (Minimum of 2 Sundays)
X <= 26 (Number of weekdays in August)
Y <= 5 (Number of Sundays in August)
X >= 0
Y >= 0

Truly speaking, if same person sees the ad in weekday and that same
person also sees the ad on a Sunday, then total exposure should be
counted as only once unlike the double counting we do in this
formulation. This situation is quite possible in reality where in, a
person subscribes to the same newspaper for the entire week. However,
in this formulation we neglected that aspect and assumed that they are
independent. One possible justification for such modeling is that the
advertising agency has some method of determining that 30,000 weekdays
subscribers does not include the 80,000 Sunday subscribers and vice
versa.

The feasible region is given by yellow region in the following figure

Visually, move the Total Exposure line in the direction of increasing


exposure value as we want to maximize exposure. We can see that point A
is the optimal.

The point A lies at intersection of Y=2 and 2000X + 8000Y =40000.

Solving these two equations, we get, the point as (12, 2). At this
point, the optimal total exposure is 30,000*12 + 80,000*2 = 520,0000.

We can also verify or cross-check our visual observation. Since optimal


solution lies at extreme points, let us just check the extreme points
just to confirm our visual observation. At these three extreme points:
• (X, Y)=(8, 2): Exposure = = 30,000*8 + 80,000*2 = 400,0000
• (X, Y)=(8, 3): Exposure = 30,000*8 + 80,000*3 = 480,0000
• (X, Y)=(12, 2): Exposure = 30,000*12 + 80,000*2 = 520,0000

Therefore, (X*, Y*) = (12, 2) and Total exposure is 520,000


Part (b)
The two constraints are redundant because the budget constraint is
binding. You exhaust the advertising budget.

Part(c)
If the minimum restriction for Sunday ads is removed, then we can see
that point B becomes optimal.
This corresponds to new optimal solution: (20, 0).
At this point, the optimal total exposure is 30,000*20 + 80,000*0 =
600,0000.

Assume the reduced cost for Sunday ads is –c. Reduced cost means that
Sunday circulation has to increase by +c so that Y takes on positive
value. This means that Sunday circulation has to improve from 80,000 to
80,000+c. Another way to look at reduced cost is – “how much is the
change in the objective function for a unit increase in a variable with
0 value”. The way you do it is as follows. Let us assume Y takes a
value of 1.

Graphically, this means,


• Add a constraint a Y>=1
• Find the optimal objective value for this new feasible region

Add Y>=1, we can get a feasible region like this:

Moving around the objective line: 30,000X + 80,000Y, we see that


C(16,1) is the optimal point.
At this point, total exposure is: 30,000*16 + 80,000*1 = 560,000
Previous objective was 600,000.
Thus reduced cost is 560,000 – 600,000 = -40,000.

Again this means, if Sunday circulation increases to 80,000+40,000 =


120,000 then we will have positive values for Y.
You can feed this into Excel and check for yourself!!

PROBLEM 2:
Solve the two manufacturing problems pertaining to Golden Electronics in Chapter 2.
That is, Problems 2 and 3 on pages 101 and 102. You can use Excel or Graphical Method
to answer this question.
Solution:
Decision Variables
• X = Number of GE45 televisions manufactured
• Y = Number of GE60 televisions manufactured
Objective Function
• Maximize Total Expected Profit per Shift
Constraints
• 300 man-hours available in production area
o GE45 requires 2 man-hours in production
o GE60 requires 2 man-hours in production
• 240 man-hours available in assembly area
o GE45 requires 1 man-hour in assembly area
o GE60 requires 3 man-hours in assembly area
• Non-negativity

Problem 2, Part (a)


LP Formulation
Maximize 50X + 75Y
Subject to
2X + 2Y <= 300 (Production)
1X + 3Y <= 240 (Assembly)
X >= 0
Y >= 0

Using Excel Solver, we get:


(X*, Y*) = (105, 45)
Optimal Expected Profit per shift = $8,625

Part (b)

Shadow price for extra assembly hours is: $12.5


This means if GE can procure an extra one hour on assembly line, GE can
make profit of $12.5. (Make RHS of Assembly from 240 to 241 and check
if this is correct!!)

Part (c)

The allowable increase in objective coefficients for GE60 is +75. This


means per unit profit can go up to 75+75 = $150 and optimal solution
will not change. Increasing to $135 lies within this range, and thus
optimal solution will not change. If you increase per unit profit to
$300, then the optimal solution will change.
Problem 3, Part (a)
LP Formulation
Maximize 50X + 75Y
Subject to
2X + 2Y <= 300 (Production)
1X + 3Y <= 240 (Assembly)
30X + 45Y <= 80*60 (Quality Control in Minutes)
X >= 0
Y >= 0

Using Excel Solver, we get:


(X*, Y*) = (80, 53.33)
Optimal Expected Profit per shift = $8000

Part (b)

Here the objective is changed and it is not profit

LP Formulation
Maximize Y
Subject to
2X + 2Y <= 300 (Production)
1X + 3Y <= 240 (Assembly)
30X + 45Y <= 80*60 (Quality Control in Minutes)
X >= 0
Y >= 0

Using Excel Solver, we get:


(X*, Y*) = (0, 80)
The optimal number of GE60 sets is 80. As a self-exercise, try
interpreting the shadows prices and reduced cost in this model.

Part (c)

LP Formulation
Maximize 50X + 75Y
Subject to
2X + 2Y <= 300 (Production)
1X + 3Y <= 240 (Assembly)
30X + 45Y <= 80*60 (Quality Control in Minutes)
X - 3Y =0 (GE45 are three times as many as GE60)
X >= 0
Y >= 0

Using Excel Solver, we get:


(X*, Y*) = (106.67, 35.56)
Optimal Expected Profit per shift = $8000.
Notice that this profit is same as in Part 3(a) except that we are
looking for a specific mix of products.
PROBLEM 3:
Jason the jeweler makes diamond bracelets, necklaces, and earrings. He wants to work at
most 160 hours per month. He has 800 ounces of diamonds. The profit, labor time and
ounces of diamonds required to produce each product is given below. If demand for each
product is unlimited, how can Jason maximize his profit?
Product Unit Profit Labor hours Ounces of diamonds
per unit used per unit
Bracelet $300 0.35 1.2
Necklace $200 0.15 0.75
Earrings $100 0.05 0.5

Formulate this as LP and find and discuss the sensitivity of optimal solution using Excel
Solver. Remember to write down all the key components of LP clearly.

Solution:
Decision Variables
• B = Number of bracelets produced by Jason in a month
• N = Number of necklaces produced by Jason in a month
• E = Number of earrings produced by Jason in a month
Objective Function
• Maximize Total Profit in a month
Constraints
• 160 man-hours available in a month
• 800 ounces of diamonds available
• Non-negativity

LP Formulation
Maximize 300B + 200N + 100E
Subject to
0.35B + 0.15N + 0.05E <= 160 (Hours)
1.20B + 0.75N + 0.50E <= 800 (Diamonds)
B >= 0
N >= 0
E >= 0

Using Excel Solver, we get:


(B*, N*, E*) = (0, 1066.67, 0)
Notice that as we have not enforced integer restrictions on the
necklaces and we get fractional solutions as 1066.67 or 1066+2/3 =
3200/3.

Optimal Profit per month = $213,333.33

The way we can interpret this fractional solution is that in 1 month


Jason should make 3200/3 necklaces or in 3 months he should make 3200.

This would have been appropriate if the planning period can be extended
to 3 months. But sometimes we are restricted by time duration and such
interpretation may not be reasonable.
Sensitivity Analysis
Grabbing the values from Excel’s Answer and Sensitivity Reports:

Adjustable Cells
Objecti Allowabl Allowa
Final Reduced ve e ble
Coeffici Decrea
Cell Name Value Cost ent Increase se
$C$1 Number of
1 Bracelets 0 -20 300 20 1E+30
$D$ Number of 1066.666
11 Necklaces 667 0 200 1E+30 12.5
-
$E$1 Number of 33.33333 33.33333
1 Earrings 0 33 100 333 1E+30

Constraints
Constra Allowabl Allowa
Final Shadow int e ble
R.H. Decrea
Cell Name Value Price Side Increase se
$F$1
4 Hours 160 0 160 1E+30 0
$F$1 266.6666
5 Diamonds 800 667 800 0 800

1E+30 = 1 followed by 30 zeroes. It just means that it is a huge


number; we typically call it infinity for all practical purposes.

Changes in Objective Function Coefficients:


In sensitivity report, look under the “Adjustable Cells” section

Range of optimality for objective function coefficients:


• The profit per one bracelet can be increased at most by $20 and
decreased by infinite amount and the current optimal solution, that
is (B*, N*, E*) = (0, 1066.67, 0), will still be optimal. Of course
if you change the objective coefficients the optimal total profit
will change but not the optimal solution values, that is B*, N*, E*,
in this range.
So the range for profit coefficient for bracelet is (300-infinity,
300+20) under which the current solution is still optimal.

• The profit per one necklace can be increased by infinite amount but
can be decreased at most by $12.5. So the range for profit
coefficient for necklace is (200-12.5, 200+infinity) under which the
current optimal solution is still optimal.

• The profit per one earrings can be increased at most by $33.33 and
decreased by infinite amount and the current optimal solution, that
is (B*, N*, E*) = (0, 1066.67, 0), will still be optimal. So the
range for profit coefficient for earrings is (100-infinity,
100+33.33) under which the current solution is still optimal.

Reduced cost for variables that have zero value:


As we talked in the class, the reduced cost helps us identify the
change in objective function for forcing a particular variable with 0
value to take on at least value of 1.
• In this case, if you force the problem to sell 1 bracelet, your
profit is going to drop by $20.
• If you force the problem to sell at least 1 earring, the profit
is going to drop by $33.33

Changes in RHS of Constraints


In sensitivity report, look under the “constraints” section

Binding
• We see that both constraints “Hours” and “Diamonds” are binding and
hence there is no slack in these constraints. Final value refers to
the LHS of the constraints. Constraint R.H. Side refers to the RHS
of the constraint. If both of them are equal, the constraint becomes
binding.
Shadow Prices
• Shadow Price for “Hours” constraint is $0. What does this mean? This
means that procuring one additional hour will not increase profit by
anything. This is because, if you try with 161, then you will see
that the maximum number of hours needed is just 160. The additional
hour is not needed at all!
• Shadow Price for “Diamonds” constraint is $266.67. Notice that for
the RHS of this constraint, there is no allowable increase!! So what
shadow price mean here? It means that if you decrease one ounce of
diamond then the profit will decrease by $266.67. Try with 799 and
you will see the change in the profit drops by exactly $266.67.
PROBLEM 4:
The following is a special type of Linear Program, called Transportation Problem, similar
to what we discussed in the first class:
The distance between Boston, Chicago, Dallas, Los Angeles, and Miami, in miles, are
given in the following table.
Boston Chicago Dallas Los Angeles Miami
Chicago 983 0 1205 2112 1390
Dallas 1815 1205 0 801 1332
Miami 1539 1390 1332 2757 0

Each city needs 40,000 kilowatt hours (kwh) of power, and Chicago, Dallas and Miami
are capable of producing 70,000 kwh. Assume that shipping 1000 kwh over 100 miles
costs $4.00. From where should the power be sent to minimize the cost of meeting each
city’s demand? Formulate this as LP and find the optimal solution using Solver.

Solution:
This is very similar to what we have seen in the class for Standard
Transportation Problem except that here you have to determine the per
unit shipping cost.

We are given that to transport


1000 kwh – 100 miles - $4
So for 1000 kwh – 983 miles – how much does it cost?
We expect the cost to be higher so it should be $4 * 983/100.
This is to transport 1000kwh.
1000 kwh – $4 * 983/100
1 kwh – how much does it cost?
We expect the cost to be lower hence, we have $4 * 983/100 * 1/1000.

This is just the shipping cost of transporting 1kwh. Similarly, we can


calculate all the per unit shipping costs. We will use Excel to do
that. The formula you want to input is this:
Distance * 4 * 1/100* 1/1000

In this problem it is just I asked you to do one extra calculation of


coming up with shipping costs. Once you have the shipping costs, then,
this problem follows what we discussed in the class.

Decision Variables
• Xij = kwh shipped from Supply i to Destination j
• Where i = 1(Chicago), 2(Dallas), 3(Miami)
• And j = 1(Boston), 2(Chicago), 3(Dallas), 4(Los Angeles), 5(Miami)

Objective Function
• Minimize Total cost of power distribution
Constraints
• Total Supply available at each location
• Total demand needed at each of the destinations
• Non-negativity
LP Formulation
Maximize
0.00392 X11 + 0.0482 X13 + 0.08448 X14 + 0.0556 X15
+ 0.0726 X21 + 0.0482 X22 + 0.03204 X24 + 0.05328 X25
+ 0.06156 X31 + 0.0556 X32 + 0.05328 X34 + 0.11028 X35

Subject to

(Supply side constraints)


X11 + X12+ X13 + X14 + X15 <= 70,000 (Chicago power supply in kwh)
X21 + X22+ X23 + X24 + X25 <= 70,000 (Dallas power supply in kwh)
X31 + X32+ X33 + X34 + X35 <= 70,000 (Miami power supply in kwh)

(Demand side constraints)


X11 + X21+ X31 >= 40,000 (Boston power requirement in kwh)
X12 + X22+ X32 >= 40,000 (Chicago power requirement in kwh)
X13 + X23+ X33 >= 40,000 (Dallas power requirement in kwh)
X14 + X24+ X34 >= 40,000 (Los Angeles power requirement in kwh)
X15 + X25+ X35 >= 40,000 (Miami power requirement in kwh)

All variables are non-negative

Optimal solution and costs are given below. The optimal total shipping
cost in this case is: $3609.60

Boston Chicago Dallas Los Angeles Miami SHIPPED SUPPLY


Chicago 30000 40000 0 0 0 70000 <= 70,000
Dallas 0 0 30000 40000 0 70000 <= 70,000
Miami 10000 0 10000 0 40000 60000 <= 70,000
RECEIVED 40000 40000 40000 40000 40000
>= >= >= >= >=
DEMAND 40,000 40,000 40,000 40,000 40,000

INPUT Distance
Boston Chicago Dallas Los Angeles Miami
Chicago 983 0 1205 2112 1390
Dallas 1815 1205 0 801 1332
Miami 1539 1390 1332 2757 0

Calculatingthe cost per unit kwh (extrastep)


Cost ($/kwh)
Boston Chicago Dallas Los Angeles Miami
Chicago 0.03932 0 0.0482 0.08448 0.0556
Dallas 0.0726 0.0482 0 0.03204 0.05328
Miami 0.06156 0.0556 0.05328 0.11028 0
PROBLEM 5:
This problem is based on Graphical Method we discussed in the class.

The feasible region for a Linear Program (LP) is represented by the region ABC (shaded
in yellow color). Based on your knowledge of graphical method (without using the
Solver) answer the following questions (show all your calculations):

(a) What are the extreme points of this feasible region?


(b) Classify the following points: (2,1) and (3, 0.5)
(c) Line equations (3X-4Y=12, X+2Y=4, X=1) are provided in the legend of the
graph. Based on the feasible region ABC (shaded yellow color) write down all the
constraints for this LP.
(d) If my objective function is MAX 5X + 6Y, what is my optimal solution?
(e) If my objective function is MIN 5X + 6Y, what is my optimal solution?

You need to figure out the optimal solution without using the Solver.
HINT: Recall how we talked about how the optimal solution lies at the corner
points/boundary points of the feasible region.
Solution:
Part (a)
The extreme points are the corner points of the feasible region. They
are (1, 0), (1, 1.5) and (4, 0)

Part (b)
(2,1) is called a boundary point since it lies on the boundary line of
the feasible region. (3, 0.5) is also called a boundary point since it
lies on the boundary line of the feasible region.

Part (c)
The trivial constraints for the feasible region are:
X>0, Y>=0. (You do not really X>=0 as you will see next)

To figure out the sign, (that is >= or <=) for the line equations given
in the problem, do the following:
• Pick any point inside the feasible region (interior point)
• Plug in the x, y values and calculate LHS; compare LHS with RHS.
Depending on the relation between these two, figure out the sign

Let’s pick point (2, 0.5)


For line X=1:
LHS = X = 2, RHS = 1. LHS >= RHS. Hence, X >= 1

For line X+2Y=4:


LHS = X + 2Y = 2 + 2(0.5) = 3
RHS = 4
Here, LHS <= RHS. Hence constraint is X + 2Y <= 4

For line 3X-2Y=12:


LHS = 3X - 2Y = 3(2) - 2(0.5) = 5
RHS = 12
Here, LHS <= RHS. Hence constraint is 3X - 2Y <= 12

So if we put together all the constraints:


X >= 1
X+2Y <= 4
3X-2Y <= 12
X >= 0 (You can see that this is not really required)
Y >= 0

Part (d)
From Graphical Method, we know that solution has to be at one of corner
points or boundary points. Let us check all the corner points first:
• (X, Y) = (1, 0) implies profit is 5(1) + 6(0) = 5
• (X, Y) = (1, 1.5) implies profit is 5(1) + 6(1.5) = 14
• (X, Y) = (4, 0) implies profit is 5(4) + 6(0) = 20

For MAX 5X + 6Y, the optimal solution is (4,0) and corresponding


optimal value is 20.

Part (e)
We need to check the corner points again here but we already did it in
Part (d). This time we pick the Minimum.
For MIN 5X + 6Y, the optimal solution is (1,0) and corresponding
optimal value is 5.

You might also like