Team Control Number

For office use only For office use only

T1 69377 F1
T2 F2
T3 Problem Chosen F3
T4 F4
D
2017
MCM/ICM
Summary Sheet

Fly Through Security Check

summary

We develop a computer simulation model to identify bottlenecks. We use arrival

times and average check time to compute wait time with loops. Results show wait time
continue increasing. Finally, we find Zone A and Zone B two bottlenecks. In addition,
we develop a queuing model to figure out measures to increase passenger throughput
and provide good travel experience.
We mainly propose two modifications to solve the congestion problem at OâĂŹ Hare.
One is to choose the model of one queue of multiple desks. The other is to change the
ratio between number of document checkpoints and security screening equipment.
We focus on validating the modifications above. For the first solution, one queue of
multiple checkpoints performs better. Average waiting time in Model (Three M/M/1/∞)
is 7.5(min). But average waiting time in Model(M/M/3/∞) is 1,89(min). For the sec-
ond solution, we determine the ratio between number of ID check counters and security
equipment according to number of ID checkpoints and security service time. For pre-
check passengers, the ratio should be 4: 5. For regular passengers, the ratio should be 4:
8. Finally, we consider the impact of cultural differences on passenger service time. We
use the given data as an intermediate value. And we simulate the data of another two
travel styles through increasing and decreasing the service speed by 50%, do a sensitivity
analysis and accommodate these differences.
Our model is applicable worldwide. In the future we can incorporate the constraints
of economic factors to further optimize our model.
Keywords: Queuing Model; Computer Simulation;
Team # 69377 Page 1 of 21

Contents

1 Introduction 2
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Our Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Assumptions 3
2.1 About the Given Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 About Our Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

3 Analysis of the Problem 3

3.1 Simplification of the Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Analysis of the Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.3 Solution Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

4 Establishment of the Model 5

4.1 M/M/s Waiting Queuing Model . . . . . . . . . . . . . . . . . . . . . . . . 5
4.1.1 Single Desk Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.1.2 Multi-service Desk Model . . . . . . . . . . . . . . . . . . . . . . . . 7
4.2 Exploration Model by Simulation . . . . . . . . . . . . . . . . . . . . . . . . 8
4.3 Multi-objective Optimization Model . . . . . . . . . . . . . . . . . . . . . . 8

5 Calculation and Analysis of the Model 9

5.1 Task A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.2 Task B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.3 Task C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

6 Strengths and Weaknesses 16

6.1 Strengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
6.2 Weaknesses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

7 Future Work 17

8 Policy and Procedural Recommendations 18

Appendices 19
Team # 69377 Page 2 of 21

1 Introduction

1.1 Background

The terrorists attacked the United States on September 11, 2001. Since then, the air-
port security has been significantly enhanced. Passengers need to remove shoes, belts,
jackets, metal objects, and electronic products. They put them in separate X-ray boxes
at the airport security checkpoint. The notebooks and some medical equipment also
need to be removed from their packs and be placed in separate containers. Because of
the lengthy and complex security checkpoints, the passengers wait for too long, or even
miss the flights. In order to change the situation that passengers wait too long, the TSA
has made some improvements, including some modifications to the checkpoint equip-
ment and the procedures, such as increasing staffs at the airport and introducing a new
measure. Approximately 45% of passengers enroll in Pre-Check. They pay $85 to get a
background check and enjoy a separate screening process. This change has improved
the airport’s long wait situation.

1.2 Our Works

• Evaluation model of waiting time

– In the process of the document check, we can use the waiting time of new
passengers to describe the flow of people checked by document (when the
waiting time for new arrivals increases, which means that the flow of people
going through the ID check is low).
– In the process of the security check, we can use the waiting time of the new
arriving passenger who is going to pass the baggage and body screening to
describe the flow of people (when the waiting time of additional passenger in-
creases, which means that the flow of people who going through the security
check is low)

• Queuing Theory
We study the quantity of channels, and different queuing ways to select the rea-
sonable the quantity of channel and the way of queue. Furthermore, we also need
to determine the average and variance of the waiting time of customers.

• Computer Simulation
Because the subject data is less, not suitable for statistical methods to find data
characteristics. Therefore, we establish a recursive algorithm model, make full use
of data analysis of passenger queuing time.

• Multi-objective Optimization Model

We try to reduce variance in wait time. For increasing passenger throughput, we
achieve it also by reducing passengers’ average wait time.
Team # 69377 Page 3 of 21

2 Assumptions

2.1 About the Given Data

• There is no time for placing items on the belt going into x-ray.

• There is no transition time between process stations.

2.2 About Our Model

• Passengers arrive at rate λ according to a Poisson process and move the process
from state i to i + 1.

• Service times are independent to each other and have an exponential distribution
with parameter µ in the M/M/s queue.

• There are s checkpoints (s is a constant), which serve from the front of the queue.
If there are less than s passengers, some of the checkpoints will be idle. If there are
more than s passengers, they will queue.

• There is no limit on the number of passengers wait in line.

• Security officers inspect passengers’ identification and boarding documents dur-

ing work time after passengers arrive at the checkpoint. (We will discuss other
situations in model modification)

• No congestion occur when passengers depart the checkpoint area.

• We use the average passenger wait time to estimate passenger throughput.

• We use the range of passenger wait time at certain probability to measure its vari-
ance.

3 Analysis of the Problem

The U.S. Transportation Security Agency (TSA) wants our team to solve the conges-
tion problem at airport. We should mainly consider two aspects:

• Passenger throughput.

• Passenger flying experience.

In order to expedite passenger throughput, we can shorten the time of security check
by improve the security process. In terms of passenger flying experience, we try to min-
imize both the average and variance of waiting time.
Team # 69377 Page 4 of 21

3.1 Simplification of the Problem

In the process, whether passengers proceed to the conveyor belt on the other side of
the X-ray scanner to collect their belongings or they go to Zone D to receive a pat-down
inspection because of failure of certain steps, they move forward freely. They do not
have to walk in a line. So we do not consider wait time in these two procedures.

3.2 Analysis of the Process

After our simplification of the problem, we divide the whole process into two parts.
In general, passengers will join in two queues and be checked twice.

• In the first section, passengers randomly arrive at the checkpoint and wait in a
queue until a security officer can inspect their identification and boarding docu-
ments.

• In the second section, the passengers then move to a subsequent queue for an open
screening line. Once the passengers reach the front of this queue, they prepare all of
their belongings for X-ray screening. Passengers must remove shoes, belts, jackets,
metal objects, electronics, and containers with liquids, placing them in a bin to be
X-rayed separately. Laptops and some medical equipment also need to be removed
from their bags and placed in a separate bin. Meanwhile the passengers process
through either a millimeter wave scanner or metal detector.

3.3 Solution Steps

We accomplish the task in several steps:

step 1: Build a model to review airport security checkpoints and staffing to identify po-
tential bottlenecks that disrupt passenger throughput.

step 2: Provide solutions that both increase checkpoint throughput and reduce variance
in wait time while maintaining the same standards of safety and security.

step 3: Validate our model by applying our modifications to the current process and
demonstrate the changes.

step 4: Make a sensitivity analysis on how cultural differences may impact the way in
which passengerâĂŹs process through checkpoints.

step 5: Propose policy and procedural recommendations for the security managers based
on our model.

step 6: Assess strengths and weaknesses.

step 7: Future prospects.

Team # 69377 Page 5 of 21

4 Establishment of the Model

4.1 M/M/s Waiting Queuing Model

Symbol Description

The common symbols in queuing theory are listed below in table 1.

Table 1: Symbol Description

Symbol Description Symbol Description

λ the average arrival rate L Average team length
µ the average service rate of a single service Lq Average queue length
P the probability of n people in system W Average length of stay
n the number of people in system Wq Average waiting time

Queuing system

There are two main queuing system in our model.

The queue security check ends and leaves

Passengers arrive
Checkpoint

The passenger being checked

Figure 1: Single-server Queuing System M/M/1

security check ends and leaves

Checkpoint
1
The queue security check ends and leaves
Passengers arrive Checkpoint
2
security check ends and leaves
Checkpoint
3

Figure 2: S Parallel Service Desk with One Queuing System M/M/s

Birth and Death process

A system of inter-arrival time and service time showed exponential distribution, we

[1]
denoted M .
Team # 69377 Page 6 of 21

Let E represent the number of times of entering state n, and L represent the number
[1]
of times of leaving state n. We have |E − L| ∈ {0, 1} . When the system arrives at steady
state, which means t, we have, therefore arrival rate = removed rate.

Balance Equation

Three conclusion equations:

λn−1 λn−2 · · · λ0
Cn = (n = 1, 2, · · · ) (1)
µn µn−1 · · · µ1
Pn = Cn P0 (n = 1, 2, · · · ) (2)
1 1
P0 = = (3)
∑ ∏
∞ n−1
λi
∑
∞
1+ 1+ Cn
µi+1
n=1 i=0 n=1

Mark Pn = P {N = n} (n = 0, 1, 2. · · · ) as the probability distribution of the queue

length N after the system reaches the steady state.

Assumprions

• The passenger arrives alone, and the time interval obeys the negative exponential
distribution with the parameter λ.

• Suppose there are s service stations in the system, and the service time of the ser-
vice desk is independent of each other and obeys the negative exponential distri-
bution with the parameter µ.

• Passengers take the initiative to select the free service desk. If there is no free ser-
vice desk, wait in queue.

4.1.1 Single Desk Model

The single-desk model is the simplest queuing system that we can use it to describe
the queuing status of a single security gate.
In this model, λn = λ, n = 1, 2, · · · and µn = µ, n = 1, 2, · · · mark ρ = λ
µ. And set
ρ < 1 so that ( )n
λ
Cn = , Pn = ρn Pn (n = 1, 2, · · · )
µ
Among them
( ∞
)−1 ( )−1
1 ∑ 1
P0 = = n
ρ = =1−ρ (4)
∑
∞
1−ρ
1+ ρn n=0
n=1
Team # 69377 Page 7 of 21

Therefore, the probability value of the number of passengers in the system under
equilibrium condition is

Pn = (1 − ρ)ρn n = 0, 1, 2, · · · (5)
λ
But this equation could obtain only in the condition that ρ = µ < 1.

Main Quantitative Indicators

Average team length L:

∞
∑ ∞
∑ ρ λ
L= nPn = n(1 − ρ)ρn = = (6)
1−ρ µ−λ
n=0 n=1

Average queue length Lq ;

∞
∑
Lq = (n − 1)Pn = L − (1 − P0 )
n=1 (7)
ρ2 λ2
=L−ρ= =
1−ρ µ(µ − λ)
Average length of stay W :
1
W = E(T ) = (8)
µ−λ
Average waiting time Wq :
1 λ
Wq = W − = (9)
µ µ(µ − λ)

4.1.2 Multi-service Desk Model

The stationary distribution of queuing systems is discussed below.

For multi-service desk system. There are two equations:λn = λ n = 0, 1, 2, · · · and
{
nµ n = 1, 2, · · · , s
µn =
sµ n = s, s + 1, · · ·
λ
Mark ρs = ρ/s = sµ ,if ρ < 1 , according to equation 1 and 2. We have:

 (λ/µ)
n
n! ( ) n = 1, 2, · · · , s
Cn = (λ/µ)s λ n−s (λ/µ)n (10)
 = s!sn−s n ⩾ s
s! sµ

So {
ρn
n! p0 n = 1, · · · , s
pn = ρn (11)
p
s!sn−s 0
n⩾s
Among them
[ s−1 ]−1
∑ ρn ρs
p0 = + (12)
n! s!(1 − ρs )
n=0
Team # 69377 Page 8 of 21

The probability of the number of passengers in the system under equilibrium is given
in equation 11 and 12. When n ⩾ s:
∞
∑ ρs
c(s, ρ) = pn = p0 (13)
n=s
s!(1 − ρs )

The equation 13 is called Erlang Waiting Formula. It gives the probability that a passen-
ger arrives at the system is waiting.

Main Quantitative Indicators

Average team length L and average queue length Lq :

p0 ρs ρs c(s, ρs )
L = Lq + ρ Lq = = (14)
s!(1 − ρs )2 1 − ρs

Average length of stay W and average waiting time Wq :

L Lq 1
W = Wq = =W − (15)
λ λ µ

4.2 Exploration Model by Simulation

Since the amount of data in the data table is too small, we use the simulation method
to calculate the sequence of each person’s waiting time.
We define the time for the it h passenger to arrive at the service desk as Ta (i) and the
average service time for the service desk to be Ts . So that we can get the waiting time
Tw (i) of the it h person.

Tw (i) = Ts + Tw (i − 1) + Ta (i − 1) − Ta (i) (16)

By analyzing the waiting time data and the arrival time data of each passenger, we
can get the distribution of personnel and the bottleneck position in the existing airport
security check.

4.3 Multi-objective Optimization Model

To solve the problem, we should consider both passenger flying experience and pas-
senger throughput. As for passenger flying experience, we aim at minimizing the pas-
sengers’ average waiting time. Meanwhile, we also try to reduce variance in wait time.
For increasing passenger throughput, we achieve it also by reducing passengers’ av-
erage wait time. Here we explain the principle. We use average arrival time interval
to calculate the speed of passengers’ arrival. Similarly, we utilize average security check
time to estimate the speed of security check. When average arrival time interval is equal
to average security check time, we say there is a balance between passengers’ arrival
and security check, so the average wait time remains the same. However, if passengers’
average wait time is shorter, in condition of the same speed of passengers’ arrival, the
Team # 69377 Page 9 of 21

speed of security check must be higher. In situation where passengers queue in line in
most of the time, passenger throughput heavily depends on the speed of security check
in a positive correlation. So if we take measures to successfully reduce passengers’ aver-
age wait time, we also improve passenger throughput. Meanwhile, we provide a better
passenger experience.
So after problem conversion, we are actually concerned with passengers’ average
wait time as well as variance in it. Our goal is to optimize the two targets simultaneously.

Model Application in Reality

In the situation, passengers arrive at airport when officers are at work. However, in
reality, there are already some people waiting in line when officials begin with security
check. So the actual average passenger wait time equals to the sum of the ideal average
passenger wait time and time of security check for passengers in front of them.

5 Calculation and Analysis of the Model

5.1 Task A

To deal with task a, we use our exploration model by computer simulation to explore
the flow of passengers through a security check point and identify bottlenecks. We use
the given data to complete the task. Although each column of data in the dataset is
independent of the others as the data were collected as several groups of people were
coming through the screening process, we can obtain some information such as average
time of document check process, millimeter wave scan and X-Ray scan. Furthermore,
we develop a simulation model to make full use of the data. We get data of wait time for
every passenger in Column A and B for further analysis. In particular, we interpret data
in these ways:

• We combine Column C and Column D to calculate the average time of Document

check.

• We divide data in Column H(time to get scanned property) into two groups in
ascending order of time. We consider the 45 percent data in the front come from
pre-check passengers because approximately 45% of passengers enroll in the pre-
check program.

We can see the passenger wait time in the figure3 and figure 4 is continuously in-
creasing. In other words, document checkpoints are congested because the arrival speed
is faster than the speed of document check. So we identify document checkpoint as a
bottleneck. We apply the same method to analyze baggage and body screening (showed
figure ??), and get the similar conclusion. However, the trend of rising up is more obvi-
ous compared to document check. Therefore, baggage and body screening checkpoint is
another bottleneck we identify.
Team # 69377 Page 10 of 21

X-axis values = arrival order X-axis values = arrival order

Y-axis values = time the people wait (s). Y-axis values = time the people wait (s).

Figure 3: TSA Pre-check Wait Time Figure 4: Regular Pax Wait Time

5.2 Task B

We develop five potential modifications to the current process to improve passenger

throughput and reduce variance in wait time. We model these changes to demonstrate
the impact of our modification. Each modification is followed by corresponding valida-
tion.

Modification 1

Method
We can increase the quantity of document Checkpoints to 8. document Check makes
no difference for pre-check passengers and regular passengers. Given the number of
pre-check passengers (45%) is approximately equal to the number of regular passengers
(55%), it is reasonable to set 4 as the quantity of document checkpoints for both pre-check
passengers and regular passengers.
Principle
With other conditions unchanged, when we increase the number of security check-
points, we can reduce the average passenger wait time. When we have only one check-
point, we cannot apply M/M/1 model because the speed of document check is slower
than the speed of passenger arrival. No balance will be achieved in this condition. We
cannot use the former formula in the queuing theory, either. So we use the data given in
the excel sheet to calculate the average waiting time of the customer by simulation, the
method we have explain before.
Verification
The first column in the chart is based on the data presented in the table and we
calculate the average passenger wait time. We find that the average passenger wait time
is long, meanwhile the passenger throughput is low.
Team # 69377 Page 11 of 21

X-axis values = arrival order

Y-axis values = time the people wait (s).

Figure 5: Baggage and Body Screening

The second column in the chart shows the result of our multi-service-based queuing
model. We select two document checkpoints, and the average passenger wait time (pre
check) is 7.5 seconds compared to the first one we calculated. The average wait time
in this model is significantly shorter. So we believe that the passenger throughput in-
creases. And the probability of the longest passenger wait time less than 33 seconds is
95%. So the passenger wait time range is relatively large.
The third column in the chart shows the result of our multi-service-based queuing
model. Now, we choose three document checkpoints. And the average passenger wait
time we calculate is 0.99 second. Compared with the previous two results we calculate,
the average waiting time has a noticeable shortening of the time. That is to say, the
passenger throughput increases. And the probability of the longest waiting time less
than 7 seconds is 95%. Comparing the 2 longest passenger wait time, we can see variance
of wait time reduce.
The fourth column in the chart shows the result of our multi-service-based queuing
model. At this time, we choose four document checkpoints. And the average passenger
wait time is 0.17second. Compared with the three results we calculated, the average
waiting time has a noticeable shortening of the time. In others words, we believe that
the passenger throughput increases. And the probability of the longest waiting time
less than 0.2 second is 95%. Comparing the 3 longest passenger wait time, we can see
variance of wait time.
When we continue to increase the number of document checkpointsïijŇthere is no
obvious decrease in average passenger wait time.
Team # 69377 Page 12 of 21

So when we increase the number of document checkpoints, we can increase the flow.
Comparing the results obtained by simulation with another three different model
in the queuing theory: M/M/2, M/M/3 and M/M/4(showed in table 2), we arrive at
a conclusion that when the number of security checkpoints is increased, the average
wait time of passengers will be greatly reduced, eventually, passenger throughput of the
airport will be increased.

Table 2: Comparing Data

Item s=1 s=2 s=3 s=4

Average team length L (people) - 2.04 1.36 1.27
Average queue length Lq (people) - 0.79 0.11 0.02
Average length of stay W (s) - 18.39 12.23 11.41
Average waiting time Wq (s) 51.048 7.15 0.99 0.17
Service Desk Usage ρ - 0.62 0.42 0.31
95% reliability maximum time (s) 106.9 33 7 0.2

Conclusion
At the same time, we can get the probability of wait time in a certain range. The
probability can measure variance in passenger wait time. In other words, we can obtain
the range of wait time on condition of a given probability. We choose the probability to
be 95% to calculate the range of passenger wait time for later comparison.

Modification 2

Method
One queue of multiple checkpoints performs better.
Principle
According to the "Queuing Theory", we have two methods of queuing. The one
is M/M/S model of queuing. In others words, passengers wait in line as soon as they
arrive, and they go to the free document checkpoint one by one. The other is s × M/M/1.
s × M/M/1 is the method of queuing that passengers can choose any one of document
checkpoint.
Verification
We use a simple example to compare the two models as follow.
The security checkpoint has three document checkpoints. The average passenger ar-
rival rate λ = 0.9person/min; Average service rate at the checkpoint µ = 0.4person/min.
The queuing problem is analyzed using the M/M/s/∞ and M/M/1/∞ models,
[2]
respectively, and get the results table .Through the following table 3, we understand If
the number of service desk and service rate unchanged, the single team queuing way
better than the multi-team queuing.
Conclusion
Team # 69377 Page 13 of 21

Table 3: The Result Table

Item M/M/3/∞ Three M/M/1/∞

Idle probability 0.0748 0.25 Each subsystem
The probability that the passenger must wait 0.57 0.75
Average team length 3.95 9 Whole system
Average queue length 1.70 2.25 Each subsystem
Average length of stay 4.39(min) 10(min)
Average waiting time 1,89(min) 7.5(min)

By comparing 2 methods of queuingïijŇWe know that one queue of multiple check-

points performs better.
So that we choose figure?? to be our queuing model.

Modification 3

Method
Optimize the ratio of the number between the security check equipment (including
the millimeter scan and X-ray scan) and document checkpoints to 5:4.
Principle
We apply the statistical analysis to the original process. When the number of the
document checkpoints is equal to that of security inspection equipment, the average
passenger wait time for document check is less than that for baggage and body security.
The average passenger wait time depends on the time the person pass the millimeter
wave body scanner as well as the time the item is scanned. We can shorten the aver-
age passenger wait time for baggage and body screening by increasing the number of
security check equipment (showed in table 4 and 5).
Verification
The figure shows the average passenger wait time for document check and baggage
and body screening. By comparing the average length of time, we select one with the
shortest average passenger wait time. In order to make the ratio of the average passen-
ger wait time close to 1, we determine the ratio of the number between baggage and
body screening and document check to be 5:4. We can also observe that when the ratio
between the two is 5:4, for example 4 security check ports and 5 sets of security equip-
ment, the variance range of average passenger wait time at the possibility of 95 percent
is very small.
Conclusion
To obtain the largest passenger throughput, we need to make the ratio of number be-
tween security check equipment and document checkpoints to be 5:4. In this condition,
we can also ensure the average passenger wait time remain a low value, which benefits
the passengers.
Team # 69377 Page 14 of 21

Table 4: Regular Pax Data Sheet

Regular Pax Document Check Baggage and Body Screening

Item s=4 s=4 s=5 s=6 s=7 s=8
Average team length L (people) 0.89 6 3.83 3.24 3.31 3.27
Average queue length Lq (people) 0 2.74 0.57 0.16 0.05 0.01
Average length of stay W (s) 11.29 68.2 43.53 38.88 37.58 37.2
Average waiting time Wq (s) 0.05 31.17 6.49 1.84 0.55 0.16
Service Desk Usage ρ 0.22 0.81 0.65 0.54 0.47 0.41
95% reliability maximum time (s) 0.1 123 40 13 1 0.2

Table 5: TSA Pre-Check Data Sheet

TSA Pre-Check Document Check Baggage and Body Screening

Item s=4 s=2 s=3 s=4 s=5
Average team length L (people) 1.27 3.36 1.72 1.54 1.5
Average queue length Lq (people) 0.02 1.87 0.23 0.04 0.01
Average length of stay W (s) 11.41 38.19 19.58 17.44 17.04
Average waiting time Wq (s) 0.17 21.24 2.63 0.5 0.1
Service Desk Usage ρ 0.31 0.75 0.5 0.37 0.3
95% reliability maximum time (s) 0.2 82 17 2 0.1

Modification 4

Method
Technology brings us new check methods, such as face recognition. Face recognition
greatly shortens the average service time in document checkpoints. So we can reduce
the average passenger waiting time in document checkpoints.
Principle
The average service time of face recognition is shorter than the average service time
of document checkpoints. So we can reduce the average passenger waiting time in doc-
ument checkpoints.
Verification
When we use face recognition, the average passenger wait time of document check
is 0.12 second, and the probability of the longest passenger wait time less than 1 second
is 95% (showed in table 6) while the average waiting time of document check without
facial recognition is 7.15s, and the probability of the longest passenger wait time less
than 33 second is 95% .It can be seen that the mean waiting time reduces. That is to say,
the passenger throughput increases, and the variation of the wait time reduces.
Conclusion
Replace the traditional inspection method with the new inspection equipment, such
as Face recognition.
Team # 69377 Page 15 of 21

Table 6: Document Check by Face Recognition

Item s=1
Average team length L (people) 0.12
Average queue length Lq (people) 0.01
Average length of stay W (s) 1.12
TSA Pre-Check
Average waiting time Wq (s) 0.12
Service Desk Usage ρ 0.11
95% reliability maximum time (s) 1
Average team length L (people) 0.09
Average queue length Lq (people) 0.01
Average length of stay W (s) 1.09
Regular Pax
Average waiting time Wq (s) 0.09
Service Desk Usage ρ 0.08
95% reliability maximum time (s) 0.4

Modification 5

Method
Encourage more people to participant in Pre-check.
Principle
Pre-check will not greatly shorten the average waiting time of passengers. But it
plays an important role in the baggage and body screening. Because the Pre-check are
not required to remove shoes, belts, or light jackets, and they also do not need to remove
their computers from their bags. So they save a lot of time. That is to say the waiting
time of queuing in the security check phase is shorter than the average waiting time of
the general queue, and the change of the waiting time is smaller.
Verification
Take the example of open queues with 4 security check. As for the Pre-check pas-
senger, the average passenger wait time is 0.5 second, and the probability of the longest
passenger wait time less than 2 seconds is 95%; While the average ordinary passenger
waiting time is 31.17 seconds, and the probability of the longest passenger wait time less
than 123 seconds is 95%. We can know that the average passenger wait time is reduced,
that is, the throughput increases, and the variation of the waiting time reduces.
Conclusion
More people being participant in Pre-check will reduce the average passenger wait
time and increase the passenger throughput.

5.3 Task C

Method
All over the world, different areas have different cultural norms that shape the local
rules of social interaction. In general, people tend to form certain characters because of
Team # 69377 Page 16 of 21

Table 7: Data Compare

Baggage and Body Screening

Item Regular Pax s = 4 TSA Pre-Check s = 4
Average team length L (people) 1.54 6
Average queue length Lq (people) 0.04 2.74
Average length of stay W (s) 17.44 68.2
Average waiting time Wq (s) 0.5 31.17
Service Desk Usage ρ 0.37 0.81
95% reliability maximum time (s) 2 123

their local cultural influence. Some people may pay more attention to efficiency, whereas
others do not attach much importance to it. In this specific example, some passengers ex-
pect to pass security check as quickly as possible. If they happen to choose regular check
points, then they will try to speed up during the process. They tend to do things such
as removing shoes, belts, jackets, metal objects, electronics, and containers with liquids
more quickly. However, othersâĂŹ action may be slower. Therefore, cultural differences
may impact the serve time in which passengerâĂŹs process through checkpoints. We
simulate three traveler styles, using the data searched from Beijing International Airport
as the middle one. And we generate data of another two traveler styles by increasing
and decreasing 50% the speed of the middle.
According the data we searched, the T3 terminal has 30 security checkpoints. It is
designed to send up to 3215 passengers per hour. And annual throughput is 17500000
every year.Meanwhile Beijing International Airport usually opening only 1/3 of the se-
[3]
curity gate .
So that we get some effective value for queue theory: λ = 0.0298 and µ = 0.0331.
Result
In order to accommodate these differences in a manner that expedites passenger
throughput and reduces variance the security system, we can change the number of
checkpoints based on task B. The sensitivity analysis result is showed in the following
figure. See the appendix for more details about the calculation results.

6 Strengths and Weaknesses

6.1 Strengths

• We make full use of the data and conditions given by the question and analyze
carefully.
Particularly, we develop an exploration model by computer simulation to calculate
the average passenger wait time. By detecting the trend of the average passenger
wait time, we can easily identify bottlenecks.

• Our model is strongly supported by queuing theory.

Team # 69377 Page 17 of 21

X-axis represents the average inspection time.

Y-axis represents the checkpoint number need to open

Figure 6: The Balance between Checkpoints Number and Average Check Time

As a traditional theory, queuing theory has a wide range of application, which has
been successfully validated.

• We simplified steps and models after rational analysis.

By proper problem conversion, we use the average passenger wait time to estimate
passenger throughput. And we use the range of passenger wait time at certain
probability to measure its variance.

6.2 Weaknesses

• We donâĂŹt take the economic factors into consideration and we ignore the air-
portâĂŹ profits due to lack of data.

• We assume passengers arrive at rate λ according to a Poisson process. However,

we ignore the impact of unexpected factors on queuing time at airport.

7 Future Work

A Multi-objective Programming Model Considering Economic and queuing

Factors

In general, increasing the quantity and quality of services can reduce customer wait-
ing costs. But it will also increase the cost of service agencies. Take the case of airports
Team # 69377 Page 18 of 21

for example, the cost of airports will be increased when the airport increases the opening
of the security checkpoints, and the cost of the airport will also increase when the ratio
of airport security equipment is increased. One of the objectives of the optimization is to
minimize the costs of both service agencies and passengers and determines the level of
service to achieve the optimal goal.
So the future research direction that we are going to study is that we need to take
economic factors into consideration. Then, we build the model to make the economic
factors, the airport throughput, the average passenger wait time to achieve optimal at
the same time.

Other Ways to Measure Variance in the Average Passenger Wait Time

Usually, it is not enough to use only one index to measure variance in the average
passenger wait time. Here we come up with two more ways for estimation.
The first one is to directly calculate the variance. If the value is large, then it means
the change is acute. However, this method will work only when you have considerable
amount of data. Otherwise, the results will not be reliable nor convincing.
[5]
The second method imitates the definition of travel time index - the ratio of travel
time in the peak period to the travel time at free-flow conditions. A value of 1.30 indi-
cates a 20-minute free-flow trip takes 26 minutes in the peak period. Similarly, we can
define an index C as the quotient of peak passenger wait time divided by average pas-
senger wait time. The value of C not only indicates the intensity of congestion, but also
suggests the ability of the security system to accommodate towards stability is poor.

8 Policy and Procedural Recommendations

After the analysis of airport congestion, we establish our simulation and queuing
models to solve the problem. Our policy and procedural recommendations cover the
sequence of airport security check and the number of security checkpoints. We propose
them based on the information we obtain such as the average check time of each kind
and average passenger wait time for different categories of security check.

Recommendation

1: Encourage more people to enroll in pre-check program.

2: Increase the number of document checkpoints at airport. (4 checkpoints for pre-check

passengers, 4 checkpoints for regular passengers)

3: change the queuing form to a queue of multi-service

4: Use the actual time of document check and baggage and body screening time to de-
termine the ratio of the number between document checkpoints and screening equip-
ment. (the ratio should be 4:5 for pre-check passengers while it should be 4:8 for
regular passengers)
Team # 69377 Page 19 of 21

5: Apply advanced technology. For instance, replace the traditional inspection method
(traditional document check) with the new inspection equipment, like face recogni-
tion.

In particular, we provide some proposals based on the data we get from the excel
dataset. In the situation, we assume the sum of the average security check in each period
to estimate the total average time of security check. we should encourage more people
to enroll in pre-check program, thus greatly shortening the time of security time. At the
same time, we should make the ratio of the number between people participate in the
pre-check program and passengers who go through regular process correspond with the
ratio of the number between pre-check security checkpoints and regular checkpoints that
are opened. We open the number of security mouth for a total of 8 (4 checkpoints for pre-
check passengers, 4 checkpoints for regular passengers), which can reduce the average
passenger wait time for the two types of passengers waiting to pass document check.
We use a queue of multi-service mode of service, which offer another way to decrease
the average passenger wait time. For pre-check passengers, we determine the ratio of
the number between document checkpoints and security check equipment ratio as 4:5.
So we open four document checkpoints for pre-check passengers. Correspondingly, we
open four document checkpoints for regular passengers, which means we should be
equipped with five sets of security equipment. Similarly, the opening of four regular
document checkpoints requires the optimization of 8 sets of security equipment. The
new ratio ensures that the average passenger wait time in the whole queuing phase is
relatively low and the variation of average passenger wait time is small.

References

[1] https://en.wikipedia.org/wiki/Queueing_theory#Example_of_M.2FM.2F1

[2] http://www.supositorio.com/rcalc/rcalclite.htm

[3] http://news.syd.com.cn/system/2013/08/13/010136939.shtml

[4] http://www.supositorio.com/rcalc/rcalclite.htm

[5] http://d2dtl5nnlpfr0r.cloudfront.net/tti.tamu.edu/documents/ums/congestion-

Appendices
Here are two tables.
Table 1 is a collection of data for Task C, which describes how the average waiting
time is stabilized by adjusting the number of security channels at different security times.
Table 2 is the raw data through computer simulation of the waiting time set.
Team # 69377 Page 20 of 21

Table 1
Team # 69377 Page 21 of 21

Table 2