Вы находитесь на странице: 1из 8

Session ETD 441 WAITING LINES: SYSTEM SIMULATION FOSTERS COMPANY PERFORMANCE

Hamid Khan1 and Musatya Bere 2 Abstract


At the request of a business a graduate project was undertaken to solve productivity problem. Several production techniques were investigated to apply, including system simulation, which could be used to effectively analyze business productivity problems. In this case System Simulation was used for the queuing situation for customers arriving and waiting in the drivethrough of a fast food franchise. This business project focuses on analyzing the time spent by customers in a busy drive through queue of a fast food outlet adjacent to an urban university. The research involved first of collecting data, which included customer inter-arrival times, customer ordering time, and service times. This data was then used to build a model of the system that would behave in a steady state situation at peak time to benefit management decisions. The GPSS/H simulation software package was used to build the model as well as analyze the data. Index Terms Modeling, System Simulation, Queuing Theory, Service Optimization.

INTRODUCTION
This application project report uses all methodologies suggested by Agnus [1]. According to Schriber [2] customers waiting for a service are common in banks, hospitals and food outlets which are targets of ineffective systems that can be improved by simulation. Whenever customers arrive to the bank faster than they could be served, some will have to wait in line before being served. Waiting for a service is undesirable. As such management is faced with the problem of reducing waiting time. The project is aimed at using some of the techniques that can be used to study the waiting line problem.

PROBLEM S TATEMENT FOR PROJECT FORMULATION


Long waiting lines of cars with customers wanting to buy food through the Drive Through windows at McDonalds (Montgomery) food outlet near an urban university are a common phenomenon during lunchtime. Long waiting lines are not favorable because customers may turn away (balk away to competitors), and this means lost business. The application of simulation for solving waiting lines problems will produce some important information that will help management in decision making so as to reduce customer waiting times, hence protecting balking of customers to competitors.

PROJECT CONCEPTION PHASE


1. Project purpose: The purpose of this project is to determine ways in which McDonalds (Montgomery) food outlet could improve its customer service by reducing customer-waiting time. 2. Tasks involved: 1. 2. 3. 4. 5. 6.

Interviewing the managers Interviewing the customers Determining customer arrival patterns Determining service times Determining the number of identical servers Determining the way customers are served from the queue

22

Associate Professor, Northern Kentucky University, Highland Heights, KY 41099 Graduate Student, Master of Technology, Northern Kentucky University, Highland Heights, KY 41099 January 28 thru 31, 2003, Tucson, Arizona ETD441-15

2003 CIEC Conference

Session ETD 441


7. 8. 9. 10. 3. Determining the maximum number of customer allowed in the queue Developing queuing model using GPSS simulation package. Evaluating the results Validating the method for generalization.

Brainstorming the method Basically the aim is to find the average time a car spends waiting in a line when it enters the system. Three distinct methods can be used: e.g. Actual Experimentation, Analysis and Simulation. Actual Experimentation: the stopwatch is used to determine the cars inter-arrival time, time it takes to place an order, and the time it takes to wait for the food to be served before departing with the food. Method is too expensive and time consuming. Mathematical Analysis: Using mathematical models to compute average time spent in the system. This method relies on too many assumptions, hence could be inaccurate. Simulation: Using a model that generates random arrival of cars and the duration of service. Models are simple to use and many software packages are available. After the brainstorming session the method chosen was simulation due to the availability of modern software packages like GPSS/H and its capability to answer many What if business questions by simulated responses. What if questions are the hall marks of creativity suggested by Csikszentmihalyi [3].

4.

Risks involved in simulated solutions: 1. 2. The results may be close approximation of the real situation hence care must be taken to use the results as a basis for making informed decisions. More time must be spent in gathering actual data sample for averages.

Alternative solutions: 1. The data could be gathered over a longer period of time. 2. Applying a very stable safety factor to the results is a must. 5. Tasks schedules and budget constraints: The project stipulates that all the tasks should be completed in sixty days. The tasks will be carried out during daytime. There are no real budget constraints in this project as the scope of the project is investigative application of simulation. 6. Documents to be prepared during the Conception Phase: 1. 2. 3. Project purpose to get management support List of tasks to get approvals Schedule of tasks to manage time.

PROJECT S TUDY PHASE


Purpose and objective of the study phase was to develop ways to study the project selected during conception phase that effectively directs the final design of a chosen project. Study Phase activities planned-- Tasks accomplished: Collecting customer arrival times data Collecting ordering times data Collecting service times data 2003 CIEC Conference ETD441-16 January 28 thru 31, 2003, Tucson, Arizona

Session ETD 441


Project Coordination and Control Project is controlled through the use of a schedule which was approved by management. System Optimization

Series Servers Customer Inter-Arrivals W waiting line place order food-prep customer leaves system

Figure 1 Single-Channel waiting line at McDonalds Figure 1 above illustrates the single-channel waiting line at McDonalds restaurant. The customer arrives at the order placement point, places the order and then move to the food service point were the food is served. The customer leaves the system after getting the food and simulation clock is decremented by one. A customer who arrives when there is someone on the service point has to wait. When more customers arrive than the servers could handle, customers wait in the line too. This is an undesirable phenomenon in fast food business in which McDonalds is operating. The following tables show the record of time that was taken at various points of the system. These points are (1) order placement point, (2) food service point and (3) total time a customer spends in the system.

Study of Order Placement Time Pattern (Production Study) for 50 cars

Ordering Point Data


Customer Number 1 2 49 50 Time Execution Begins 10:55:37 10:56:03 12:11:26 12:14:58 Time Execution Completes 10:55:55 10:57:06 12:11:50 12:15:10 Total Time Total Time 18 59 24 12 1457

Hence from above, average time to place an order = 1457/50 = about 30 seconds Since Inter-Arrival time is a poisson distribution, for simulation calculations the system advance time in ordering may be 60 10 seconds.

2003 CIEC Conference ETD441-17

January 28 thru 31, 2003, Tucson, Arizona

Session ETD 441


Study of Customer Service Time (Production Study)

Customer Service point data Customer Time since last arrival (Sec) 26 267 212 Time service begins 10:55:37 10:56:03 12:11:26 12:14:58 Time service ends 10:57:07 10:57:24 12:15:10 12:17:50 Customer Service Time (Sec) 90 81 104 112 3830

1 2 49 50 Total

Average time to pick up an order = Average service time = 3830/50= 76.6 seconds Since service time is a poisson distribution, for simulation calculations the system advance time in ordering is 80+25 seconds. 1. Rules for the Study and Design Phases The study is based on 50 customers in the system. Customers are served on a first come first served order. The customer arrival rate is determined at the point the customers arrive at the ordering point. There is only a single channel-waiting line. The customers arrivals occur randomly and independently of other arrivals. Potential Solutions to envision

2.

Add more servers at more cost Add another order taking point Have automatic ordering language assemblers 4. Gathering and Evaluation of Information Interviewed managers about their proposals Interviewed the customers about how they feel about the waiting times (whether short or long)

5. Selection of the Best Solution Based on costs involved, there is a trade-off in the additional cost of a rapid service against waiting cost. The rapid service cost may include adding another server into the system. On the other hand cost of waiting means lost business as potential customers leave without being served. Management has a role to determine this optimum cost. Depends on whether the customers feel that they are really taking time to be served. Availability of resources.

6. Documentation Provided Project description Design Specification Design description with sketches

2003 CIEC Conference ETD441-18

January 28 thru 31, 2003, Tucson, Arizona

Session ETD 441


DESIGN PHASE
Purpose: to develop the details for the working plans for proposed project solution. Goal: to prepare a complete set of working plans for the selected design. Simulation Model of a McDonalds waiting line The customer arrivals and departures are as represented on Figure 1. As depicted before, a customer that arrives to a server when it is busy have to wait until it is free to accept new customer. A resulting waiting line is a queue. In a simulation model it is required to have a time-oriented view of the whole queuing system (Customers waiting to be served and the customer being served). Figure 3 explains all this explicitly.

EXPLANATION OF THE BLOCK EXECUTIONS GENERATE Block This block creates transactions (customers) and introduces them into the model. Its operands 60,10 represent timing in which customers arrive to the system. This means that customers arrive at a rate of 60 10 seconds. SNACKTIME QUEUE This queue is the overall queue of the system. It is represented by the number of cars waiting to place and order up to the last customer just before leaving the system with the order. It is formed when the car arrives at the INTERCOM and finds it busy. SEIZE Block This block is associated with the facility. It main purpose is to request control of the facility and initiate the waiting process. When transactions arrive, they immediately try to capture the facility, when it is busy then it waits. This is all done at the SEIZE block. ADVANCE Block This is the service point. In the model there are two advance blocks represented by INTERCOM (order taking point) and the WINDOW (order pickup point). The operands in the blocks represent the time it takes to perform a service. RELEASE Block This is associated with the facility too. It ensures that transactions give up control of the facility once they have been served. TERMINATE block It is used to destroy Transactions (remove them from the model). This block represents a point were the customer has been served with the food and is now removed from the system. The operand here is 1. This means that customers leave the system one by one after being served.

IMPLEMENTATION / R EPORT PHASE (S IMULATION R ESULTS )


GPSS/H RELEASE 2.0 (AY130) 25 Apr 2002 19:15:32 FILE: SIMULATE MCDONALD.gps

ENTITY DICTIONARY (IN ASCENDING ORDER BY ENTITY NUMBER; "*" => VALUE CONFLICT.) Facilities: 1=INTERCOM Queues: 1=SNAKTIME 2=WINDOW 2=WINDOW

*** WARNING: THE FOLLOWING ENTITIES HAVE BEEN DEFINED BUT NOT EXPLICITLY REFERENCED: Blocks: NEWBLOK1 NEWBLOK2

2003 CIEC Conference ETD441-19

January 28 thru 31, 2003, Tucson, Arizona

Session ETD 441


SYMBOL VALUE EQU DEFNS CONTEXT NEWBLOK1 NEWBLOK2 7 9 9 Block 11 Block Facility Facility Queue Queue 5 7 10 13 4 14 9 11 REFERENCES BY STATEMENT NUMBER

INTERCOM 1 WINDOW 2 SNAKTIME 1 WINDOW 2

STORAGE REQUIREMENTS (BYTES) COMPILED CODE: 348 COMPILED DATA: 40 MISCELLANEOUS: 0 ENTITIES: 428 COMMON: 10000 ----------------------TOTAL: 10816 Simulation begins. RELATIVE CLOCK: 4072.7948 ABSOLUTE CLOCK: 4072.7948 BLOCK CURRENT TOTAL 1 69 11 2 69 12 3 69 13 4 1 69 5 68 6 68 NEWBLOK1 18 68 8 50 NEWBLOK2 50 10 50 BLOCK CURRENT 50 50 50 TOTAL

Studying and Understanding the Simulation Results --AVG-UTIL-DURING-FACILITY TOTAL AVAIL UNAVL TIME TIME TIME INTERCOM 0.862 WINDOW 0.970 QUEUE ENTRIES AVERAGE CURRENT PERCENT SEIZING PREEMPTING TIME/ XACT STATUS AVAIL XACT XACT 69 50.872 AVAIL 69 50 78.987 AVAIL AVERAGE TIME/UNIT 642.492 526.107 $AVERAGE 642.492 533.959 QTABLE CURRENT NUMBER CONTENTS 19 18

MAXIMUM AVERAGE TOTAL ZERO PERCENT CONTENTS ENTRIES ENTRIES ZEROS TIME/UNIT SNAKTIME 20 10.885 69 0 WINDOW 18 8.784 68 1 1.5 RANDOM ANTITHETIC INITIAL CURRENT STREAM VARIATES POSITION POSITION 1 OFF 100000 100257

SAMPLE CHI-SQUARE COUNT UNIFORMITY 257 0.91 2976 IN USE 3104 USED (MAX)

STATUS OF COMMON STORAGE 7024 BYTES AVAILABLE

Simulation terminated. Absolute Clock: 4072.7948 Total Block Executions: 780 Blocks / second: 780000 Microseconds / Block: 1.28

Definitions/Explanation of results and Presentation to the Franchise Management


FACILITY REPORT The facilities were the INTERCOM representing the order taking point and the WINDOW representing the order pickup point.

--AGV-UTIL-DURING--TOTAL TIME 2003 CIEC Conference ETD441-20 January 28 thru 31, 2003, Tucson, Arizona

Session ETD 441


Shows the fraction of time in which the facility was doing work. This represents the utilization. For the INTERCOM it was 86.2% and for the WINDOW it was 97%. --AGV-UTIL-DURING--AVAIL TIME Shows the fraction of available simulated time the facility was in a state of capture. The facility was always available and the available time is equal to simulated time. ENTRIES Indicates the number of times the facility was captured during simulation. The INTERCOM was captured 69 times while the WINDOW was captured 50 times. AVERAGE TIME/XACT Shows the average holding time per capture of the facility. For the INTERCOM it took about 50.8seconds to place an order and it took 78.98seconds to pick the order. CURRENT STATUS Indicate the facilitys working status. Both Facilities were available during the simulation time. PERCENT AVAIL Shows the fraction of total time the facility is working. The INTERCOM was 69% working. SEIZING XACT Shows the id number of Transaction (if any) that has the Facility captured at the time report was produced. No transaction was captured. PREEMPTING XACT Shows the id number of the Transaction (if any) holding the Facility in a state of preemption at the time the report was produced. No transaction held the facility. MAXIMUM CONTENTS This shows the maximum size of queue membership. The maximum number of cars that were at McDonalds outlet on the time of study was 20 cars. This is represented by the SNACKTIME maximum contents. The queue membership at the window was 18 cars. AVERAGE CONTENTS This shows the number of transactions (cars) that were members of the queue on average. The average number of cars was 10 for the SNAKETIME queue and 8 for cars on the Window queue. TOTAL ENTRIES This represents the total number of Transactions that became queue members. Although the study was carried on 50 cars, the simulation program took in more cars. The simulation program recorded 69 cars in the SNAKETIME queue and 68 cars in the WINDOW Queue. ZERO ENTRIES This is the number of cars that had zero time in the queue. Only one car passed the window queue without waiting. This must have been the first car to be served.

AVERAGE TIME/UNIT This number represents the time taken by a car in the queue. For the SNAKETIME queue it is 10 minutes and for the WINDOW Queue it was 8 minutes $AVERAGE TIME/UNIT It represents the average duration of Queue membership, excluding cases in which cars passed through the queue in zero simulated time. For the WINDOW queue it was 9 minutes.

CONCLUSION
2003 CIEC Conference ETD441-21 January 28 thru 31, 2003, Tucson, Arizona

Session ETD 441

Simulation has presented the necessary information for management to act on. The results showed an average of 10 cars waiting to be served at any given time during lunchtime. The results also showed that each car had to wait for an average of 10 minutes. This is an undesirable feature for a fast food restaurant. An opportunity to solve the waiting time could be derived through using the facilities to full capacity. The percentage utilizations of the facilities showed that they were not utilized to full capacity. The WINDOW facility was utilized 97%. This value is unjustifiable given that there were an average of 10 cars waiting to be served. This meant 13% of the time the facility was idle while at the same time it had customers wanting to be served. The explanation to this could be that the food was not prepared when the customer arrived at the window, hence the need to wait. As such management need to make decisions to reduce waiting time based on these simulation results. Such decisions would involve increasing the number of servers. Again this has a cost element and management need to determine the optimum cost that should be incurred to deal with the waiting line problem. The other option would be to make waiting bearable. This could involve notifying the customers how much they should expect to wait in the queue. The other option also could be to train all the employees to work fast or to always give an impression that they are working fast. Simulation has managed to present the information to the management. Discussion with the management showed clear concern and they promised to act to improve the situation. Since the franchise serves the university clientele in the rush hour, the situation was taken seriously and it was promised that they will act within their resources to reduce waiting time and queue length.

ACKNOWLEDGMENT
The author is obliged to John Wiley and sons who provided him with a copy of an Introduction to Simulation by Thomas Schriber. The book contained a copy of GPSS/H rel.2.software by Wolverine Software corporation, which was used for this simulation..

REFERENCE
1. Angus Robert B. (2000), Norman A. Gundersen, Thomas P. Cullinane. Second Edition, Planning Performing, and Controlling Projects . Principles and Applications. Prentice-Hall, Inc. New Jersey. 2. Schriber, Thomas J.. (1991). An Introduction to Simulation Using GPSS/H. john Wiley, New York. 3. Csikszentmihalyi Mihaly, (1997). Creativity. Flow and the Psychology of Discovery and Invention. Harper Perennial. New York.

2003 CIEC Conference ETD441-22

January 28 thru 31, 2003, Tucson, Arizona

Вам также может понравиться