ISOM 2700: Operations Management
Session 11: Midterm Exam Review
Dongwook Shin
Dept. ISOM, HKUST Business School
About Midterm Exam
• Time: Oct. 19 (Thuesday), 7:30pm to 9:30pm
• Please arrive at the venue at 7:15pm sharp
• Venue: LT-B
• Format: 50 multiple choice questions
• Coverage: Lecture notes from 1 to 10
• Remarks
• Closed-book, closed-note
• Cheat sheet is not allowed
• A formula sheet will be provided in the exam paper
• You can use a calculator
• Bring your student ID card
1
Process Analysis
• Utilization and bottleneck analysis
• Bottleneck is the resource with the highest utilization
• Ops in NYC Hospital: can you find the bottleneck?
2
Quality Management
• Control charts to check whether process is in-control
• X-bar and P charts
• Two types of errors
• Producer’s risk: type I error α = P(reject a good lot)
• Consumers’ risk: type II error β = P(accept a bad lot)
• Design Sampling Plan that achieves the desired α, β levels
• Parameters: α, β, AQL, LTPD
• Decisions: Sample size n and acceptance number c
• Six-sigma concept: process capability index
3
Parameters (AQL, 𝛼, LTPD, 𝛽)
• Acceptable quality level (AQL)
• Quality is considered high (acceptable) if the percentage of
defective is less than AQL
• Lot tolerance percent defective (LTPD)
• Quality is considered low (exceeding the tolerance) if the
percentage of defective is higher than LTPD
Should we reject the lot if the
The accept/reject decision is not based on
defective rate is between AQL and
AQL and LTPD, but on “n” and “c”
LTPD?
4
Concept Checker
A gourmet cookie maker wants to control the sugar in her
dough very tightly. She wants no more than 44% and no less
than 35% sugar in her dough that she bakes. She observes
that her process mean is at 40%
The maximum standard deviation her process can tolerate,
so that the process would be Six Sigma capable, would be
a. 1.2
b. 0.6666
c. 0.6
d. 0.5
e. 0.3333
5
Resource Management: Decision Tree
Method
• Tree is built forward
• Identify possible alternatives
• List possible states of nature for any chosen alternatives
• Tree is solved backward
• At each decision node, choose the alternative that is the best
according to the criteria
• Expected value of perfect information
• Willingness-to-pay to know the realized state of nature before
making decisions
6
Example: Expected Value of Perfect
Information
Payoff when Payoff
Alternatives locations for
New Bridge without
new warehouse
Built New Bridge
A 1 14
B 2 10
C 4 6
• Given that the probability of 0.6 for a new bridge and 0.4 for
no new bridge:
• If we want to maximize our payoff, what should be our decision?
• EV(A) = 0.6*1+0.4*14 = 6.2
• EV(B) = 0.6*2+0.4*10 = 5.2
• EV(C) = 0.6*4+0.4*6 = 4.8
• What should be the expected value of perfect information?
• (0.6*4+0.4*14)-6.2 = 1.8
7
Practice Set 3 Q6
The All-Colorful Company produces a variety of paint products. Demand is highly
seasonal, and the forecasted demand (in gallons) for the next 4 quarters is given as
follows:
Quarter 1 Quarter 2 Quarter 3 Quarter 4
Demand Forecast 300,000 850,000 1,500,000 350,000
The All-Colorful Company has a regular-time production capacity of 90,000 working
hours per quarter for Quarters 1, 2, and 4, and the regular-time production capacity of
Quarter 3 is 150,000 working hours. Besides regular-time production, the All-Colorful
Company can also use overtime production or subcontracting. The maximum overtime
working hours in any quarter is 20% of regular-time capacity. Each gallon requires 0.2
working hour of production. The subcontractor can supply a maximum of 200,000
gallons per quarter.
The All-Colorful Company has an initial inventory of 250,000 gallons at the beginning
of Quarter 1, and the ending inventory at the end of Quarter 4 should be 300,000
gallons. Regular-time cost is $10 per hour, overtime cost is $15 per hour,
subcontracting cost is $4 per gallon, and the inventory holding cost is $3 per gallon per
quarter. Assume no backorders or stockouts are permitted. Formulate a linear
program to determine the best production plan. 8
Practice Set 3 Q6
• The full solution is on canvas, but just a few comments
• In each quarter, supply must exceed demand
• (Quarter 1)
• Supply = 250,000 + 5(R1+O1) + C1 Leftover from Newly added
• Demand = 300,000 previous period products
• 250,000 + 5(R1+O1) + C1 >= D1
• (Quarter 2)
• Supply = (250,000 + 5(R1+O1) + C1 – D1) + 5(R2+O2) + C2
• Demand = 850,000
• (250,000 + 5(R1+O1) + C1 – D1) + 5(R2+O2) + C2 >= D2
• Can write the constraints for Quarters 3 and 4, but...
• It’d be convenient if we define It as the inventory at the end of
period t
• For each quarter, we may simply write the above as
It = It-1+5(Rt+Ot)+Ct-Dt 9
Shelby Shelving: LP Revisited
max 260 S + 245 LX (Revenue)
subject to:
(S assembly) S £ 1900
(LX assembly) LX £ 1400
(Stamping) 0.3 S + 0.3 LX £ 800
(Forming) 0.25 S + 0.5 LX £ 800
(Nonnegativity) S, LX ³0
Optimal solution: (S, LX) =(1900, 650), Revenue = $653,250
10
Questions for Students
• The Shelby Shelving company is considering an investment that will
increase the capacity of the model LX assembly line to 1500. The cost
of the investment is $85,000. Is it worthwhile?
• Yes
• No
• Cannot be determined from the available data
11
Questions for Students
• The Shelby Shelving is considering an investment that will increase
the the capacity of Forming to 900. The cost of the investment is
$30,000. Is it worthwhile?
• Yes
• No
• Cannot be determined from the available data
12
Questions for Students
• The Shelby Shelving is considering an investment that can reduce the
per-unit production cost of Model LX, which essentially increase the
profit margin for Model LX from 245 to 300. The cost of the
investment is $30,000. Is it worthwhile?
• Yes
• No
• Cannot be determined from the available data 13
Yes!
In general, the RHS value
does not necessarily
Can shadow price be represent the capacity of a
negative in maximization resource!
problems?
For instance, consider a
constraint like:
𝑆 − 2 𝐿𝑋 ≥ 0
14
Shadow Price and Slack
• Relaxing a non-binding constraint would not affect the
optimal objective value
• Slack > 0 (i.e., non-binding) => Shadow price = 0
• However, “Shadow price = 0” does not imply “Slack > 0”
• The optimal objective value is affected only if we relax a
binding constraint
• Shadow price > 0 => Slack = 0 (i.e., binding)
• However, “Slack = 0” does not imply “Shadow price > 0”
15
Shelby Shelving: Feasible Region
LX model
2500
A new constraint!
2000
What is the
shadow price for
1500 this constraint?
What is the slack?
1000
Feasible Region for
500 Shelby Shelving
0 500 1000 1500 2000 2500 3000 S model
16
Recap: M/M/1 Model Utilization
𝜌 = 𝜆/𝜇
Inter-arrival
Rate 𝜆 Single
Buffer
Server
Avg. Number Waiting Avg. Number in Service
𝐿! = 𝜌" /(1 − 𝜌) 𝐿# = 𝜌 = 𝜆/𝜇
Avg. Waiting time Avg. Service time
𝑊! = 𝐿! /𝜆 𝑊# = 1/𝜇
There are n
customers in
the system with
Avg. Number in the System
probability
𝐿 = 𝜌/(1 − 𝜌)
𝑃$ = 𝜌$ (1 − 𝜌)
Avg. Time in System
𝑊 = 𝐿/𝜆 17
Recap: M/M/s Model Utilization
𝜌 = 𝜆/𝑠𝜇
Server 1
Inter-arrival
Rate 𝜆 Server 2
Buffer
⋅⋅⋅
Server s
Avg. Number Waiting Avg. Number in Service
𝐿! (using table) 𝐿# = 𝑠𝜌 = 𝜆/𝜇
Avg. Waiting time Avg. Service time
𝑊! = 𝐿! /𝜆 𝑊# = 1/𝜇
Avg. Number in the System
𝐿 = 𝐿! + 𝐿# = 𝐿! + 𝜆/𝜇
Avg. Time in System
𝑊 = 𝐿/𝜆 18
Question from Students about M/M/1
• How can we deduce the probability that there are n
customers in the system
𝑃/ = 𝜌/ (1 − 𝜌)
• Balance equation (for stable queueing systems…):
(Rate from n to n+1) = (Rate from n+1 to n)
• To move from n to n+1, a customer must arrive at the system, which occurs at the rate
of 𝜆. Likewise, to move from n+1 to n, a customer must leave from the system, which
occurs at the rate of 𝜇. Hence the balance equation means 𝜆𝑃! = 𝜇𝑃!"# for all
n=0,1,2,3,…
• Equivalently, 𝑃!"# = 𝜌𝑃! for all n=0,1,2,3,…
• Since the sum of probabilities is one, we know that ∑𝑃! = 1
• This means 𝑃$ 1 + 𝜌 + 𝜌% + ⋯ = 1, from which you can get 𝑃$ = 1 − 𝜌
• Recursively, you can obtain that 𝑃! = 𝜌! 𝑃$ = 𝜌! 1 − 𝜌
19
Question from Students about M/M/1
• How can we deduce the average number of customers in
the system?
𝐿 = 𝜌/(1 − 𝜌)
• 𝐿 = ∑𝑛𝑃! = ∑𝑛𝜌! 1 − 𝜌 = 1 − 𝜌 ∑𝑛𝜌! = (1 − 𝜌)(𝜌 + 2𝜌% + 3𝜌& + ⋯ )
• 𝜌𝐿 = 1 − 𝜌 ∑𝑛𝜌!"# = (1 − 𝜌)(𝜌% + 2𝜌& + 3𝜌' + ⋯ )
• 𝐿 − 𝜌𝐿 = 1 − 𝜌 (𝜌 + 𝜌% + 𝜌& + ⋯ )
• 𝐿 = 𝜌 + 𝜌% + 𝜌& + ⋯ = 𝜌/(1 − 𝜌)
20
M/M/s vs. G/G/s
• If both the interarrival time and service time are
exponentially distributed, we can use M/M/s formula
• If either the interarrival time or service time is not
exponentially distributed, we cannot use M/M/s formula
• Whenever M/M/s model cannot be applied, use G/G/s as
an alternative
21
Practice Question
• The Judy Gray Income Tax Service is analyzing its customer service
operations during the month prior to the April filing deadline. On the
basis of past data it has been estimated that inter-arrival times of
customers is exponential with mean inter-arrival time of 12 minutes.
The time to complete a return for a customer is exponentially
distributed with mean of 10 minutes.
• If you went to Judy, how much time would you allow for getting your
return done? 1 hour
• On average, how much room should be allowed for the waiting area?
4.17
• If Judy stayed in the office 12 hours per day, how many hours on
average, per day, would she be busy? 10 hours
• What is the probability that the system is idle? 16.7%
• If the arrival rate remained unchanged but the average time in system
must be 45 minutes or less, what would need to be changed? At least
6.33 customers/hour
22
Practice Set 4 Q2
2. A cafeteria serving line has a coffee urn from which customers serve
themselves. Inter-arrival times at the urn follow an exponential
distribution at the rate of three per minute. In serving themselves,
customers take about 15 seconds, exponentially distributed.
(e) If the cafeteria installs an automatic vendor that dispenses a cup of
coffee at a constant time of 15 seconds, how does this change your
answers to a and b?
Check the solution to Practice Set 4
23
Reminders
• Use your time wisely
• Don’t get stuck with a question for too long. Make sure
you visit all the questions and get full credits on the easy
ones
24
