Professional Documents
Culture Documents
SUBWAY PROJECT REPORT-SılaFarukBegümMert
SUBWAY PROJECT REPORT-SılaFarukBegümMert
SYSTEM SIMULATION
REPORT
Subway Model
2
1) INTRODUCTION
Doctor's Associates Inc. Subway is an American franchise chain, founded in 1965 by Fred DeLuca
and Peter Buck, serving the fast food industry.
Subway's objective all over the world is to offer a wide range of tasty, quick and healthy meals. It's
really important that their menu offers a wide range of healthier food choices for a healthier lifestyle.
Food safety and food quality are of paramount importance to them.
We observed that the processes at Subway have queues and we decided that it is a appropriate
example for queueing models that it why we chose subway for our simulation. We aimed to reduce
waiting time in queue so we tried to find subway’s optimum employee and machine numbers for
maximum efficiency.
One of the problems we observed in the subway is the insufficient number of employees. We
observed that two employees could not keep up with more than one process. Therefore, we noticed
that queue can occur in every process. Accordingly, we observed that when customers saw the queue,
they may left the system.
The other problem is can be insufficient number of heater machines in the system. Because of that
there is one heater machine in the system, customers' waiting time increases and employee
productivity decreases. As a solution, we thought that increasing the number of employees will
increase customer satisfaction and system efficiency and we recommended increasing the number of
machines.
We had to create an Arena model to solve this problem and offer suggestions. For this, we have done
model conceptualization data collection, model translation, verification, validation, data analysis,
documentation and reporting.
First of all, we imagined the design of model of Subway. Then we collected data of customer arrival
times, waiting times, and service times for each process.
We observed that there are some differences between the imagined model design and the model
design in the actual system and we shaped our model according to actual system. Then for the
verification step, we analyzed the collected data on the input analyzer and validated the p values.
3
Then we created our Arena model. For the validation step, we compared the average waiting times in
the Report to the average waiting times we held and validated whether they were in the confidence
interval. We have done data analysis according to the report of Arena model. Finally, we have stated
all our steps in our report.
4) DATA COLLECTING
We collected data for 6 days and each days one hour. These data as below ;
Customer arrival time, order process time, heater process time, selection of material process time,
Arrival Time To Cashiers Queue, Payment Process Time
For instance, our data for the first day of the simulation;
4
5) INPUT ANALYSIS AND VERIFICATION
We observed whether the distributions were verified by fitting the process and interarrival times of the
data we collected in a program of the arena called input analyzer.
We opened the data that we saved to text files in the input analyzer. Thus, we observed distributions
for each process and examined p-values for chi square and kolmogorow tests for each process. So we
verified the data.
The distributions we have obtained in the input analyzer for each process are given below.
5
Graph 1.3 Selection of Ingredient Distribution On Input Analyzer
Firstly when we fit the data distribution is erlang distributiion came out when we first fit the payment
process. Thenwe changed from erlang distribution to weible distrubiton because p value is grather
then first distribution’s p value.
6) CONCEPTUAL MODEL
After entering the system, customers enter the order queue. Then the type of bread, its length and
which menu to choose is determined. Sandwiches are then prepared to enter the heater. Depending on
the menu and bread length, the sandwiches are processed in the heater for a constant time.
6
Simultaneous customers' sandwiches can be placed in the heater at the same time if the menu
selections are the same. Therefore, order process timings are long. After the sandwiches leave the
heater according to their arrival times, the customers start the material selection process. After the
material selection process, customers enter the cashier queue. Then the payment process starts and the
customer leaves the system.
7) ARENA MODEL
7
8) MODULES
We put the distribution parameters of interarrivals time of customers in the Create Module.
We put the distribution parameters of services time of order, payment and selection of ingredient in
the Process Module.
8
8.3 Decide Module:
The decide modüle is used by entities to make branching decisions, based on chance or the
truth/falsity of prescribed conditions.
On the other hand , we used decide module for heater process time. We used the Decide module to
indicate that the heater was operated at a fixed time for each menu and bread length. Hence, we create
9
a decison module we entered in this module by finding the percentage values for each fixed time.
When we calculate the percentages we follow these steps;
For instance, 15 cm meat menu stays in the heater for 35 sc. We calculated that these times
correspond to what percentages of the data we collacted, and we repeated these steps for each bread
length and type of food.
We need to enter 7 values for 8 processes because the last value must be connected to “Else”.
We filled the inside of each heater process with constant values as we did for meat15 above.
10
8.4 Dispose Module
The dispose module implements an entity that enter it are simply discarded.
9) RUN SET-UP
First of all, we took the number of replication as 3 randomly and we calculated replication length as;
We collected data for a total of 6 days, average one hour per day. Hence, we use time unit as second
in our simulation model, we took unit as second in runsetup part.
3600x6=21600 sc/total
Subway is open for 16 hours a day so we took hours per day in run setup part as 16 hours.
11
We can find the report results from here.
10)VALIDATION
Using the average and half with values in the report section of the arena, we found confidence
intervals for each process.
12
Confidence Intervals Processes Data Averages
28.6502<X<57.2902 Chicken15 57
23.4726<X<61.4326 Chicken30 112.917
41.7218<X<53.3018 Meat15 105.652
23.4322<X<59.3522 Meat30 109.434
30.6009<X<44.8209 Order 97.89326
1.3699<X<49.2699 Payment 47.904
31.049<X<55.989 Selection of Ingredients 46.561
38.7527<X<51.2927 Tuna15 58.687
-12.7001<X<68.8199 Tuna30 90.667
-69.772<X<179.708 Turkey 301.333
-26.4336<X<54.7364 Vegetable 124.5
We calculated the average of waiting in queue times for the given processes. We checked whether the
mean values in the data were within the confidence interval. We have observed that Chicken15,
payment and selection of ingredients processes are valid. We used third standard deviation(s)
calculations. We determined the process with the largest standard deviation and used that value in our
calculations. We took standard deviation for the selection process.
12.47= S/√
We took Ɛ=10.
R≥( ) =4.66
S/√
Replication
5 2.78( )=6.23
√
13
11) SUMMARY OF THE MODEL RESULT
11.1 ENTITY
14
As it is seen, the total time spent in the system is 311.09 sc. While the average number of customers
entering the system is 187, the number of customers leaving the system is 181. This may be due to
customers leaving the system outside the simulation period.
11.2 PROCESS
15
16
11.3 QUEUE
As shown above, average waitin time and half width values for each process are shown as arena
output. We calculated confidence interval using these average terms. So,we have done validation
using these calculations.
17
11.4 RESOURCE
Since there were two employees in total in Subway, whoever was available for each process was
working in that process.Therefore, we showed a single source in the model and set the capacity to
two. In the above resourse output, we can observe the utilization of this resource instantly.
To compare model one and two, we took the differences of the average waiting times of both
processes.
18
Model 2 Queue Results
Average waiting time for 2 model and their differences above. All results are positive and this is show
that our alternative system is succesful. When we increased the number of our employees to 3, our
utilization rate increased. Because the process times in the system decreased so that the system's idle
time increased.
For Model 1 ( 5 rep. With 2 employee) we calculated confidence interval to find and then we
repeated for Model2.( 5 rep. With 3 employee).
19
For Model 1 selection process ;
33.4607 <x< 51.3407 our calculating average waiting time for selection of ingredient is 46.561. So,
we used its halfwidth value(8.94) to calculate as 7.190.
6.1133 <x< 14. 5533 this interval’s halfwidth value is 4.22. Altough, data average is not on the this
interval and other process intervals we used selection process interval’s halfwidth value to calculate
as = 3.394.
( ) ( )
( )/ ( )+( ) = 5.69 → 6
20
42.4007-10.3333± (3.55)= 23.3699< ϴ1- ϴ2 <40.7649 for ϴ1- ϴ2 = 32.0674
And as it is seen above, in alternative system our number in and number out values more close to each
other than model 1.
When we add the system one more employee to improve the our model , the cost of system can
increase at least up to the minimum wage.
14) ASSUMPTIONS
Due to the lack of data and some time errors, we did not collect the order exit time,for example, the
distribution of the order was not suitable at first and we made assumptions to correct this problem.
Since many customers' orders were placed in the heater at the same time, it was difficult to observe
the actual order process time of each customer. Therefore, in cases where the queue is formed, we
have combined the time differences between the order start of the customers and used them. In this
way, the distribution was valid on the input analyzer. In Subway, heater times varied between fixed
values according to bread length and menu type. However, we have included in the heater constant
times the time it takes for the employee to remove orders from the heater. When we fit these data, we
see that the distributions are not suitable, so we decided to use the decide module. And we rounded
off all heater times to existing fixed times. We set these new heater times as constant values by
preparing separate processes for each menu.
21
15) CONCLUSION
Consequently, we observed long queue and wait times in order and other processes at the Subway in
our school, so we thought that the subway was appropriate for our project. In this report, we took
some steps to improve the queuing systems like Subway and we created a simulation in the arena.
First of all, we designed the system, then collected the data and created our arena model using this
data. We observed the distribution of the data we collected through the input analyzer and transferred
the results to our model. We have interpreted our model according to the outputs of the arena and
made inferences and recommendations for this.
16) RECOMMENDATIONS
Consequently, our first advice for the model to work better may be to keep the data smoother as we
mentioned in the assumptions, so that we get more realistic distributions and our simulation model
gives a more efficient result. Secondly, it was difficult for two employees on the subway to catch up
with each process. So there was a queue in every process. In order to reduce the length of these
queues and process teams, we have increased the number of alternastive employees to 3. In the
ordering and selection processes of the first two employees, the 3rd employee may also work at the
cash register. In this way, other employees prepare orders faster. As a third recommendation, we can
increase the number of heaters from one to two. Because the system has many queues, the prepared
sandwiches were waiting to enter the heater, so increasing the number of heaters increases the speed
of the system.
17) REFERENCES
https://www.subway.com/tr-TR
https://pdfs.semanticscholar.org/5331/fd12a37dbc8df828b32f69eb06b81bcb384b.pdf
22