You are on page 1of 18

SMU Classification: Restricted

Week 4 – Optimization class exercises

Linear programming

Practice 1

Your company sells 2 models of Jacuzzi tubs: A1 and B2.

In the manufacturing process, you buy pre-fabricated tubs, and add a pump and tubing, to create
the final product. From a manufacturing perspective, the main difference between A1 and B2 is the
amount of tubing and man hours required (both models only require 1 pump)

In the next production run, you have: 200 pumps, 2,880m of tubing, 1,566 man hours of labour

A1 requires: 12m of tubing and 9hrs of labour per unit

B2 requires: 16m of tubing and 6hrs of labour per unit

Expected profit per tub is $350 for A1 and $300 for B2. You are confident that you will be able to sell
all tubs produced

REQUIRED

You need to decide how many of each to produce to maximize profits during the next production
cycle.

File: Week 4 - Class exercise.xlsx, Linear Programming tab

Step-by-step solution

The summarized LP model is as follows:

MAX: 350X1 + 300X2

Subject to: X1 + X2 ≤ 200

9X1 + 6X2 ≤ 1566

12X1 + 16X2 ≤ 2880

X1 ≥0

X2 ≥0
X1 and X2 are integers

Where X1 = no of A1 to produce and X2 = no of B2 to produce

Based on the LP model, input the following values into the yellow cells in the spreadsheet model:

1
SMU Classification: Restricted

Insert the following formulas:

E4: =(C4*C3)+(D4*D3)

E7: =(C7*C3)+(D7*D3)
Copy and paste to E8:E9

Next, go to Data  Analyze  Solver. Input the values as shown:

The solver should return the following output

2
SMU Classification: Restricted

You should manufacture 122 units of A1 and 78 units of B2. This translates to a total profit of
$66,100.

3
SMU Classification: Restricted

Goal programming

Practice 2

Your company wants to build a new convention center in a vacant building that it owns. Based on
market research carried out by a consultant, the convention center should ideally have 5 small
conference rooms (each 400 sq ft), 10 medium rooms (750 sq ft), and 15 large rooms (1,050 sq ft).

Also, you learn that if the new convention center had 25,000 sq ft of space, it would be among the
largest in the region - this would be a key advertising point for the convention center.

Your developers tell you that it will cost $18,000 to build a small room, $33,000 to build a medium
room, and $45,150 to build a large room.

The company has allocated an initial budget of $1,000,000 to developing this new convention center

REQUIRED

As the CFO of the company, how many large, medium, and small rooms should you recommend be
built?

File: Week 4 - Class exercise.xlsx, Goal Programming tab

Step-by-step solution

The summarized GP model is as follows:

MIN:
1 1 1 1 1
¿+ ¿+ ¿+ ¿+ ¿
5 10 15 25000 1000000

Subject to:
−¿ ¿ +¿¿
X1 +d 1 - d 1 = 5
−¿ ¿ +¿¿
X2 + d 2 - d 2 = 10
−¿ ¿ +¿¿
X3 + d 3 - d 3 = 15
−¿ ¿ +¿¿
400X1 + 750X2 + 1050X3 + d 4 - d 4 = 25000
−¿ ¿ +¿¿
18000X1 + 33000X2 + 45150X3 + d 5 - d 5 = 1000000

−¿ ¿ +¿¿
Where d i , di ≥ 0 and Xi ≥ 0 and integers for all i

Next, input the following weights in cells B17:F18

4
SMU Classification: Restricted

In the rest of the model, fill up the yellow cells as follows:

Insert the following formulas:

E6: =SUMPRODUCT(B6:D6,B2:D2)
F6: =SUMPRODUCT(B6:D6,B3:D3)
B9: =B6+B7-B8; copy and paste to C9:CF9
B13: =B7/B10, copy and paste to C13:F13
B14: =B8/B10, copy and paste to C14:F14
B20: =SUMPRODUCT(B13:F14,B17:F18)

Go to Data  Analyze  Solver. Input the following:

5
SMU Classification: Restricted

The following solution should be generated

6
SMU Classification: Restricted

FURTHER ANALYSIS

Change the weights to the following:

Repeat the GP analysis using solver. The following solution should be generated:

7
SMU Classification: Restricted

Increasing the magnitude of the weight assigned to the overachievement of the cost goal (relative to
the other goals) indicates that exceeding the budget goal is viewed as being ten times more
undesirable that the over/underachievement of the various other goals for which positive weights
are assigned. Accordingly, we see that this solution provides for a significant reduction in the cost
(reduces from $1,097,250 to $1,006,950). At the same time, we also note that this solution involves
tradeoffs in the achievement of other goals (particularly in the number of large rooms and the
overall size of the convention center).

In a GP analysis, we will continue to adjust the weights assigned until we arrive at a solution that is
acceptable to the decision maker (this ‘acceptable’ solution is not absolute and may differ depending
on unique circumstantial factors).

8
SMU Classification: Restricted

MOLP

Practice 3

You are the CFO of TOBO Mines which operates two coal mines in Western Australia – Mine A and
mine B. In the next year, you have forecasted an increase in demand for coal:

Type of Coal Incremental Yearly


Demand (Tons)

High grade 48

Medium grade 28

Low grade 100

To meet growing demand, you are planning to schedule extra shifts of workers at the mines. Due to
manpower constraints, only 1 additional shift of workers can be scheduled each month at each
mine.

Production per shift (per month) at each mine is as follows:

Other details associated with adding a shift:

Mine A Mine B
Cost (per month) $40,000 $32,000
Gallons of toxic water 800 1,250
produced (per month)
Life threatening accidents (per 0.20 0.45
month)

REQUIRED

While you want to increase production to meet the increased demand for coal, you also want to
ensure that you don’t create more pollution that is necessary and also keep the number of life-
threatening incidents to a minimum.

For how many months should you schedule extra shifts in mines A and B in the next year?

9
SMU Classification: Restricted

File: Week 4 - Class exercise.xlsx, MOLP tab

Step-by-step solution

The intermediate MOLP model is as such (refer to seminar notes for more details):

The spreadsheet model represents a straight forward implementation of the model (refer to
comments embedded in specific cells for further details).

As we have learnt in practice 1, a LP model can only have one objective function. How do we then
implement our 3 objective functions?

If these objectives had target values, we could then treat them the same way we did goals in goal
programming models. Unfortunately, the problem does not specifically provide us with goals.
However, if we solved our model to find the solution that minimizes an individual objective (say
production costs), then that solution would be a reasonable value to use as a goal for the production
cost objective. Solving for the other 2 objectives individually would similarly give us reasonable
values to use as targets for the toxic water and accidents objectives. With this, we could then view
our MOLP problem in the format of a GP model, with three goals.

10
SMU Classification: Restricted

Let’s start with the Production cost objective. To obtain the minimum production cost that can be
realized in this problem, input the following Solver parameters:

The solution obtained (minimum production cost) is $244,000. The solution can be obtained by
running shifts for 2.5 months at mine A and for 4.5 months at mine B.

Next, obtain solutions for the other 2 objective functions in the same way. You should obtain
solutions of 6,950 gallons for the toxic water objective and 2.00 for the accidents objective.

Note that these 3 solutions are in conflict with one another (e.g. S1 has the lowest cost but highest
accident rate while S3 has the highest cost and lowest accident rate). This is unsurprising given that
MOLP problems often involve trade-offs  to improve the value of one objective, we have to
sacrifice the values of others.

Given the above, we can now redefine our goals as:

 Goal 1: The total cost of production should be approximately $244.

11
SMU Classification: Restricted

 Goal 2: The gallons of toxic water produced should be approximately 6,950.

 Goal 3: The number of life-threatening accidents should be approximately 2.0.

Our updated objective function, where we seek to minimize the sum of weighted percentage
deviations, is:

We know that the actual production cost


cannot be smaller than $244,000 (we
had earlier determined that $244k is the
minimum production cost), so the %
deviation is computed as shown.

Constraints are:

Recall from our lecture notes that the optimal solution to an LP problem always occurs on a
corner point of the feasible region. Hence, if we use the above (linear) objective function
(together with the other stated constraints which are also linear in nature), we will always obtain
a solution on a corner point regardless of the weights that we assign. Specifically, we will confine
ourselves to the following three possible (corner point) solutions:

12
SMU Classification: Restricted

To enable ourselves to explore points other than corner points along the curve (because we
want to explore the trade-offs among the constraints), we will need to introduce a minimax
objective (Q).1

A minimax objective seeks to minimize the maximum weighted percentage deviation from the
target values for given goals. In this problem, our minimax objective is given by:

Correspondingly, our revised (final) MOLP model is:

1
Further reading: https://personalpages.manchester.ac.uk/staff/jian-bo.yang/JB%20Yang%20Book_Chapters/
YangYang_MSM_Minigoal.pdf

13
SMU Classification: Restricted

With this, we can modify our spreadsheet to implement the revised model.

Input the appropriate target values in cells E8:E10

Insert the following formula in cell F8 and copy to cell F10: =(D8-E8)/E8
Assign appropriate weights (start with: +1 for all) to cells G8:G10
Insert the following formula in cell H8 and copy to cell H10: =F8*G8

Designate cell B18 as the minimax variable.

Input the following Solver parameters:

14
SMU Classification: Restricted

15
SMU Classification: Restricted

Adjust the weights to explore various other solutions.

Note:

- The solution obtained above (X1 = 4.23, X2 = 2.88) does not occur at a ‘corner point’. The
introduction of Q allows us to explore non-corner point solutions. The decision maker might find
this solution more appealing than other solutions which occur at corner points.
- The solution also does not occur inside the feasible region. An MOLP model allows a decision
maker to explore non-dominated solutions (which occur along the boundary line of the feasible
region). Dominated solutions (which occur inside the feasible region) will not occur in an MOLP.

16
SMU Classification: Restricted

NLP

Practice 4

You are the CFO of TLC Corp, and are trying to determine how to allocate your company’s R&D
budget for the next year. Six projects are under consideration. You believe that the success of each
project depends, to a large extent, on the number of engineers assigned to each project.

Each project proposal includes an estimate of the probability of success as a function of the number
of engineers assigned to the project. Each probability function is in the form:

Pi = probability of project i if assigned Xi engineers = Xi/(Xi + ei), where ei is a positive constant for
project i that determines the shape of the probability function.

Details of the project are as follows:

REQUIRED

You have agreed to hire up to 25 engineers to assign to the projects and are willing to allocate up to
$1,700 of the R&D budget to cover start-up costs of the selected projects.

You now need to use NLP to determine the project selection and resource allocation strategy that
will maximize the expected NPV.

File: Week 4 - Class exercise.xlsx, NLP tab

Step-by-step solution

Insert the following formula in cell D7 and copy to cell D12: =C7-$C$14*B7

Insert the following formula in cell C13: =SUM(C7:C12)

Insert the following formula in cell I13: =SUMPRODUCT(B7:B12,I7:I12)

Insert the following formula in cell F7, and copy to F12: =C7/(C7+E7)

Insert the following formula in cell H7, and copy to H12: =F7*G7

Insert the following formula in cell H13: =SUM(H7:H12)

Input the following Solver parameters:

17
SMU Classification: Restricted

18

You might also like