SimQuick Process Simulation with Excel

Second Edition

by David Hartvigsen

Mendoza College of Business Administration University of Notre Dame

Prentice Hall Upper Saddle River, NJ 07458

Table of Contents Preface Motivation………………………………………………………………………………………………….x

How to use the booklet ………………………………………………………………………………..x Chapter 1: Introduction …………………………………………………………………………………1 Sec. 1: What is process simulation?……………………………………………………………….2 Sec. 2: What does SimQuick do?…………………………………………………………………..3 Sec. 3: How does SimQuick incorporate uncertainty? ……………………………………..4 Sec. 4: System requirements and “installation”……………………………………………….6 Chapter 2: Waiting Lines……………………………………………………………………………….8 Sec. 1: Solving a problem with SimQuick………………………………………………………9 Example 1: A Bank……………………………………………………………………………………..9

Modeling the process (with a process flow map) ………………………………..10 Entering the model into SimQuick ……………………………………………………10 Interpreting SimQuick results …………………………………………………………..15 Improving the process, Variation 1……………………………………………………19 Improving the process, Variation 2……………………………………………………22

Sec. 2: Additional waiting line processes ……………………………………………………..23

Example 2: A grocery store checkout………………………………………………..23 Example 3: A call center………………………………………………………………….24 Example 4: A fast-food restaurant drive-thru ……………………………………..25 Example 5: A bank with a discrete distribution…………………………………..26 Example 6: A bank with changing demand ………………………………………..27

Sec. 3: Waiting lines with Decision Points……………………………………………………29

Example 7: An airport security system………………………………………………30 Example 8: A department of motor vehicles ………………………………………33 Example 9: A hospital emergency room…………………………………………….34

Sec. 4: Waiting lines with priorities and resources…………………………………………36

Example 10: Cross-trained workers at a fast-food restaurant………………..36 Example 11: Value meals at a fast-food restaurant ……………………………..40

Chapter 3: Inventory in Supply Chains……………………………………………………..44 Sec. 1: A periodic review inventory policy …………………………………………………..46

Example 12: Grocery store inventory………………………………………………..46 Sec. 2: Reorder point inventory policies……………………………………………………….49

Example 13: An electronics superstore ……………………………………………..49 Example 14: A warehouse ……………………………………………………………….53 Example 15: Two stores and a warehouse………………………………………….54

Sec. 3: More complex inventory policies ……………………………………………………..56

Example 16: An appliance store ……………………………………………………….57 Example 17: A department store……………………………………………………….62

Chapter 4: Manufacturing …………………………………………………………………………..69 Sec. 1: Linear flow processes ……………………………………………………………………..70

Example 18: A generic linear flow process ………………………………………..71 Example 19: A manufacturing cell ……………………………………………………73

Sec. 2: Assembly/disassembly processes………………………………………………………75

Example 20: Box manufacturing ………………………………………………………75 Sec. 3: Batch and job shop processes …………………………………………………………..79

Example 21: Pharmaceutical manufacturing ………………………………………79 Example 22: A single machine job shop…………………………………………….83

Sec. 4: Quality and reliability in processes……………………………………………………86

Example 23: A quality control station ……………………………………………….86 Example 24: A machine with breakdowns …………………………………………89

vii

Chapter 5: Project Management …………………………………………………………………91 Example 25: A software development project………………………………………….92 Example 26: A house addition project…………………………………………………….96

Appendix 1: The Steps in a Simulation Project ………………………………………..98 Appendix 2: Enhancing SimQuick with Excel Features………………………….99 Appendix 3: Using Custom Schedules ……………………………………………………………….. 105 Appendix 4: SimQuick Reference Manual ………………………………………………108

Preface Motivation Process simulation is one of the conceptually simplest and most often applied techniques in Operations Management and Management Science, yet it has not been widely taught to business students. A key reason for this is that performing process simulation requires the use of software, and the software that is available tends to be complex and expensive. Even the more graphics-based packages, although often beautifully designed, frequently have an enormous number of features that place an unnecessary burden on students (and instructors) in classes that are not devoted to simulation. SimQuick is a computer package for process simulation that is easy to learn (most of its features can be learned in under an hour of class time or independent reading) and inexpensive. It is aimed primarily at business students and managers who want to understand process simulation and be able to quickly analyze and improve real-world processes. SimQuick is flexible in its modeling capability; that is, it is not a “hardwired” set of examples; it requires true modeling. In addition, SimQuick runs in the widely-known Microsoft Excel spreadsheet environment (it is an ordinary Excel 2000 file with some hidden macros). Hence, users of Excel will already be familiar with much of the interface, and the results are already in the spreadsheet, ready for analysis. This booklet accompanies SimQuick. It presents the basics of process simulation by having the reader construct, run, and analyze simulations of realistic processes using SimQuick. Chapter 1 contains a brief introduction to process simulation and the concepts underlying SimQuick. The next four chapters contain a variety of examples of process simulation. These examples are organized as follows: waiting lines (Chapter 2), inventory in supply chains (Chapter 3), manufacturing (Chapter 4), and project management (Chapter 5). Each example is followed by an exercise. All of the examples and exercises have been designed with business students and managers in mind. In addition to presenting the basics of process simulation, this booklet introduces a number of key concepts from the analysis of processes: service level, cycle (or waiting) time, throughput, bottleneck, batch size, setup, priority rule, and so on. The booklet also introduces some key trade-offs from the analysis of processes: number of servers vs. service level, inventory level vs. service level, working time variability vs. throughput, batch size vs. service level, and so on. These notions are presented through computer models that the reader constructs and experiments with using SimQuick. How to use the booklet The booklet is self-contained; that is, all technical terms involving processes or operations are defined. (The reader is assumed to have a rudimentary understanding of how to use Excel on the level of knowing how to save files and how to enter information into cells.) The chapters are

organized around typical topics in Operations Management and Management Science courses so that this booklet can easily be used in these types of courses. The reader should first read Chapter 1 (which contains a conceptual explanation of process simulation and SimQuick) and Section 1 of Chapter 2 (which contains a step-by-step explanation of how to use SimQuick by completely working through a simple example). After this the reader has a lot of freedom: The remaining sections in Chapters 2, 4, and 5 can be read in any order (except Example 7 should be read before Example 11). However, the sections in Chapter 3 build on one another and should be read in sequence. The bulk of Chapters 2 through 5 consists of examples of processes that can be modeled using SimQuick. When needed, an example discusses how to build the SimQuick model. Each example is followed by an exercise. A very quick treatment of process simulation could consist of working through Example/Exercise 1, followed by Examples/Exercises 2-4 for waiting lines and Example/Exercise 18 for manufacturing. With just this material, many real-world processes can be easily modeled and studied. Adding Example/Exercise 7 with Decision Points would allow the modeling of many more types of processes. Adding Examples/Exercises 12 and 13 would provide a quick introduction to the modeling of inventory in supply chains. The booklet contains four appendices. Appendix 1 contains a list of the basic steps in conducting a simulation project. Appendix 2 contains tips on how to enhance SimQuick by using some of the features built into Excel. These tips are tied to examples in the booklet. Appendix 3 describes how to use a feature of SimQuick called Custom Schedules. Appendix 4 contains a succinct description of all the features of SimQuick and can be used for reference. Hence, the features of SimQuick are presented in two ways: through examples and in a reference manual. Solutions to exercises: Instructors are provided with complete solutions (in Excel) to every exercise. These solutions may be distributed to the students at the instructor’s discretion. Web site: Refer to www.prenhall.com/hartvigsen for additional information on SimQuick, this booklet, technical support, and process simulation in general. Over the past four years, I have used SimQuick in the classroom with executive MBAs, full-time MBAs, and undergraduate business students. After a one-hour introduction in class (basically, covering Section 1 of Chapter 2), the students successfully solve a variety of modeling problems with little help. This introduction also serves as a launching pad for term projects, whereby students identify and analyze real-world processes of their choice.

New to the 2nd edition: The SimQuick software in the second edition has been changed in a number of ways from its first incarnation. These changes include the following (the manual in Appendix 4 contains all the features of SimQuick):

• SimQuick runs considerably more quickly. This allows more simulations to be performed in a reasonable amount of time, which leads to more accurate results. As a result, more simulations are allowed.

• SimQuick accepts larger models: up to 250 elements (new tables can be added by pasting copies of the 20 tables given for each element).

• Essentially arbitrary statistical distributions can be constructed using discrete distributions (introduced in Example 5).

• A “View Model” button has been added to the Control Panel. Clicking this button produces a copy of the tables used in the model in a format convenient for printing and logic checking.

• The initial number of objects at Buffers can be generated randomly. This allows models to capture demand patterns that change over time (see Example 6).

• Decision Points can have up to ten outputs. • The simulation details appearing on the Results worksheet can be suppressed, which

allows SimQuick to run faster, especially when many simulations are performed. • The method for using Custom Schedules for modeling arrivals and departures has been

simplified (see Appendix 3). The text has also been updated; these changes include the following:

• Four completely new models have been added. Two of these models deal with services: airport security and a hospital emergency room (Examples 7 and 9); and two deal with inventory in supply chains: a base stock policy used in an appliance store (Example 16) and a periodic review policy used in a department store (Example 17).

• The key introductory section (Section 1 of Chapter 2) on how to use SimQuick has been streamlined (by removing the introduction of Decision Points). As a result, learning the basics of SimQuick is even easier than in the first edition. This makes possible a shorter treatment of process simulation, if the instructor wishes.

• Decision Points are now introduced in a separate section (Section 3 of Chapter 1) via an airport security example. This example is more timely than the ATM bank example from the first edition, which has been removed. Thus the topic of Decision Points may be skipped altogether.

• The issue of how to model observed data with a statistical distribution, particularly when it does not follow a built-in SimQuick distribution, is an important topic and is addressed as a new variation on the introductory bank example (see Example 5).

• The initial conditions of a simulation (particularly how much inventory is initially in the buffers) are of critical importance. This topic is now explicitly dealt with as a new variation on the bank example (see Example 6).

• An important issue in manufacturing environments is that, over the long run, inventory will pile up to fill the available space. This issue is now addressed in the exercise on linear flow processes (see Exercise 18).

• The general model for project management has been simplified. • An appendix has been added that lists the basic steps in a simulation project.

To Nancy

Acknowledgments The design of SimQuick was inspired by the breakthrough simulation product XCELL, so I want to begin by acknowledging its authors: Richard Conway, William L. Maxwell, John O. McClain, and Steven L. Worona. Next, I want to thank my editor Tom Tucker at Prentice Hall. He has been indispensable in helping me to define this project and bring it to fruition. I want to thank the following reviewers for their careful reading and excellent suggestions on a draft of the first edition: Sue Abdinnour-Helm, Arundhati Kumar, Larry Meile, Kelly B. Nichols, Jeffrey L. Rummel, and Billy M. Thornton. I’d also like to thank the reviewers of the second edition for their thoughtful suggestions based on their classroom use of SimQuick: C.H. Aikens, Stephen N. Chapman, Christos Koulamas, Michael Schwartz, and Billy M. Thornton. I still haven’t been able to implement all the great ideas I’ve received; I wish there was more time! I want to thank my colleagues Lee Krajewski and Hojung Shin for a number of helpful discussions of this project. I want to thank Kristin Ann Steffeck for her copy editing on the first edition. Finally, I want to thank my many students (executive MBAs, full-time MBAs, and undergraduate business students) of the past four years who have made many helpful suggestions during the development of this software and booklet. In particular, I’d like to single out Douglas Wait, Ben Gaw, Patrick Dahman, Xuejun Zhang, and Katy Delany for their insights.

Chapter 1: Introduction Learning objectives: • To understand the idea of process simulation • To understand the general structure of SimQuick models • To understand the role of uncertainty in process simulation • To get SimQuick running on your computer

Overview This chapter contains a brief definition of process simulation, an overview of how SimQuick works, and instructions for how to run SimQuick on a computer. Most of the details of how to use SimQuick are covered in Chapter 2, Section 1. Section 1: What is process simulation? Process simulation is a widely used technique for improving the efficiency of processes. Following are some examples of processes and some related efficiency problems that can be addressed with process simulation (and SimQuick): Examples of processes: • People moving through a bank or post office • Telephone calls moving through a call center • Parts moving along an assembly line or through a batch process or a job shop • Inventory moving through a retail store or warehouse • Products moving by trucks, trains, or ships through a supply chain • A software development project Examples of efficiency problems: • How many tellers are needed to keep waiting times at a bank reasonably short? • What effect will a new answering system have on how long customers wait at a call center? • What effect will a new just-in-time (JIT) inventory system have on the number of units

produced per day on an assembly line? • What is the best batch size to use in a factory? • What is the best delivery policy for goods at a warehouse? • How much inventory should be kept on the shelves in a grocery store? • How many machines should each worker operate in a manufacturing cell? • How should inventory be distributed along a supply chain? • What is the expected duration of a software development project? With process simulation, you begin by building a computer model of a real-world process. Your initial goal is to have the computer model behave in a way similar to the real process, except much more quickly. You then try out various ideas for efficiency improvements on the computer model and use the best ideas on the real process. Thus, a lot of time and money can be saved. Simulation is particularly useful when there is uncertainty in a process (e.g., the arrival times of customers, the demand for a product, the supply of parts, the time it takes to perform the work, the quality of the work). With uncertainty, it is often difficult to predict the effects of making changes to a process, especially if there are two or more sources of uncertainty that interact.

Section 2: What does SimQuick do? SimQuick allows you to perform process simulation within the Excel spreadsheet environment. There are three basic steps involved in using SimQuick (for a more detailed list of steps, see Appendix 1): 1. Conceptually build a model of the process using the building blocks of SimQuick

(introduced below). 2. Enter this conceptual model into SimQuick (this is done by filling in tables in a special

Excel spreadsheet). 3. Test process improvement ideas on this computer model. The building blocks in SimQuick are objects, elements, and statistical distributions. Objects typically represent things that move in a process: people, parts in a factory, paperwork, phone calls, e-mail messages, and so on. Elements typically represent things that are stationary in a process. There are five types of elements: Entrances: This is where objects enter a process. Entrances can represent a loading dock at a warehouse, a door at a store, and so on. You must specify when objects arrive at an Entrance and in what numbers (using a statistical distribution or an explicit “custom” schedule). Buffers: This is where objects can be stored. Buffers can represent a location in a warehouse or factory where inventory can be stored, a place where people can stand in line at a post office, a memory location in a computer for e-mail messages, and so on. You must specify how many objects a Buffer can hold. Work Stations: This is where work is performed on objects. Work Stations can represent machines in a factory, cashiers in a store, operators at a call center, computers in a network, and so on. You must specify how long a Work Station works on an object (using a statistical distribution). Decision Points: This is where an object goes in one of two or more (up to 10) directions. Decision Points can represent the outcome of a quality control station, different routings in the processing of a mortgage application, and so on. You must specify a rule for deciding in which direction an object goes (using a statistical distribution). Exits: This is where objects leave a process. Exits can represent a loading dock at a warehouse, a customer buying a product at a store, and so on. You must specify when objects depart from an Exit and in what numbers (using a statistical distribution or an explicit “custom” schedule). The third building block, statistical distributions, is discussed in the next section. A SimQuick model describes how the objects move between the elements. You have a great deal of freedom in constructing models using the building blocks of SimQuick; hence, you can

model a variety of real processes. However, because the number of building blocks is small and they are simple, some complex real-world processes will be difficult to model in SimQuick. For such situations, a more powerful simulation package may be needed. When a SimQuick simulation begins, a “simulation clock” starts in the computer and runs for the designated duration of the simulation. While this clock is running, a series of events sequentially takes place. There are three types of events in SimQuick: the arrival of objects at an Entrance, the departure of objects from an Exit, and the finish of work on an object at a Work Station. Whenever an event occurs, SimQuick moves objects from element to element as much as possible. SimQuick keeps track of various statistics during the simulation (e.g., the mean inventory at each Buffer) so you can analyze what happened when the simulation is over. Section 3: How does SimQuick incorporate uncertainty? As in a real process, the timing of events in SimQuick can be uncertain or random. Here is an example: Suppose you are entering a model into SimQuick that contains a Work Station. You must specify how much time this Work Station works on an object. What do you do if this time varies in a random fashion at the real work station? A typical approach is to observe the real work station and record a list of real working times. Following are four common possibilities and the ways in which SimQuick models them. Case 1: The list of real working times has a “bell-shaped” histogram:

Then, a list of numbers taken randomly from a normal distribution, with the same mean and standard deviation as your list, is likely to have a similar-looking histogram. So, for example, if your list of numbers has a mean of 3 minutes and a standard deviation of 1 minute, then you would enter Nor(3,1) into SimQuick. Thus, you are instructing SimQuick to randomly pick each working time for this Work Station from the normal distribution with mean of 3 and standard deviation of 1.

Histogram

0 5

10 15 20 25 30 35 40

0.7 9

1.1 7

1.5 6

1.9 4

2.3 2

2.7 0

3.0 9

3.4 7

3.8 5

4.2 3

4.6 2

5.0 0

5.3 8

5.7 6

Real working times

Fr eq

ue nc

Case 2: The list of real working times has a histogram that is “skewed to the right” as follows:

Then, a list of numbers taken randomly from an exponential distribution, with the same mean as your list, is likely to have a similar-looking histogram. So, for example, if your list of numbers has a mean of 3 minutes, then you would enter Exp(3) into SimQuick. Thus, you are instructing SimQuick to randomly pick each working time for this Work Station from the exponential distribution with mean of 3. Case 3: The list of real working times has a histogram whose bar heights are all roughly the same:

Then, a list of numbers taken randomly from a uniform distribution, with the same minimum and maximum values as your list, is likely to have a similar-looking histogram. So, for example, if your list of numbers has minimum and maximum values of 1 and 6, then you would enter Uni(1,6) into SimQuick. Thus, you are instructing SimQuick to randomly pick each working time for this Work Station from the uniform distribution with minimum and maximum values of 1 and 6.

Histogram

0 20 40 60 80

100

1.3 2

2.5 4

3.7 6

4.9 8

6.2 0

7.4 2

8.6 4

9.8 6

11 .08

12 .30

13 .52

14 .74

15 .96

17 .18

Real working times

Fr eq

ue nc

Histogram

0 20 40

1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 Mo re

Real working times

Fr eq

ue nc

Case 4: The list of real working times is described by a histogram with at most ten bars:

Histogram

0 10 20 30

0 1 2 3 4 5 6 7 8

Real working times

Fr eq

ue nc

A discrete distribution in SimQuick has up to ten output numbers, each chosen with a specified probability. To model a histogram as above, we choose one number from each interval as an output number and we set its probability to be proportional to the height of its bar. Then, a list of numbers taken randomly from this output list, according to the associated probabilities, is likely to have a similar-looking histogram. In SimQuick this is modeled with the Dis function. The details of how to use the “Nor,” “Exp,” “Uni,” and “Dis” functions are provided in Chapters 2 through 5 and Appendix 4. To input fixed schedules, see Appendix 3. Section 4: System requirements and “installation” System requirements: To run SimQuick, you must have a PC that is running Microsoft Excel, version 2000 or later. Excel can be run from your hard drive or from a network. “Installing” and running SimQuick: The disk accompanying this booklet contains a file called “SimQuick-v2.xls.” It is a standard Excel spreadsheet file (with some special worksheets and hidden macros). To use SimQuick, simply launch the application Excel and, within Excel, open your copy of SimQuick-v2.xls (see note below). You are now ready to go! It is probably most convenient to put a copy of SimQuick-v2.xls on your computer or network space and open this copy when you want to use SimQuick. Note: When opening SimQuick-v2.xls, you may see a window with the message: “The workbook you are opening contains macros…” If this occurs, just click the “Enable macros” button. If Excel does not allow you to enable the SimQuick macros, do the following: Look under Tools / Macro / Security / Security Level. If the level is set at high, change it to medium. Saving a SimQuick model: After opening SimQuick-v2.xls, you may save your work at any time just as you do with any Excel spreadsheet: Simply click on “Save As” under the “File” menu, rename SimQuick-v2.xls., and specify a location on your computer or network space.

If a problem arises with your copy of SimQuick-v2.xls (e.g., the formatting gets changed or a worksheet gets deleted), just replace it with a fresh copy from the disk that accompanies this booklet.

Chapter 2: Waiting Lines Learning objectives: • To understand the basics of using SimQuick • To model, simulate, and analyze a variety of waiting line processes • To understand the following performance measures: service level, mean cycle (or waiting)

time, and mean inventory (or number of customers in line) • To analyze the trade-off between number of servers and service level • To understand the SimQuick elements: Entrances, Work Stations, Buffers, and Decision

Points • To understand the SimQuick statistical distributions: Nor, Exp, and Dis • To understand the advanced SimQuick features: resources and priorities

Overview In this chapter, we consider waiting line processes (also called queueing systems). In a typical process of this type, customers arrive at a service in a random fashion. They may be arriving at a bank (as in our first example), a fast-food restaurant, a car wash, or even via the phone at a 1-800 customer support center. After arriving, the customers typically get in a line, wait awhile, and then are served in some way. They may then leave the process or get in another line to be served again later. Management is typically interested in determining the right number of servers or adjusting the time to serve customers so they don’t have to wait too long and so the fraction of customers able to enter the process is sufficiently high. Hence, the key performance measures of mean cycle (or waiting) time and service level are introduced in this chapter. The first section of this chapter discusses a simple waiting line process: a bank. This section is a “must read” because it contains a thorough description of how to use SimQuick and many of its features. In particular, this section shows how to model a process with SimQuick elements, how to enter a model into SimQuick, how to run a number of simulations, and how to analyze the results. The key SimQuick elements called Entrances, Buffers, and Work Stations are introduced, as well as the statistical distributions Nor and Exp. The second section contains examples that illustrate the variety of waiting line processes and associated issues that can be modeled with SimQuick. Examples 2, 3, and 4 are straightforward variations of the basic bank model. They involve a grocery store, a call center, and a fast-food restaurant. Examples 5 and 6 extend the basic bank model in two ways. Example 5 introduces the notion of discrete distributions and how they can be used in SimQuick. With these distributions we can model processes where SimQuick’s built-in distributions (e.g., the normal or exponential distributions) are not appropriate. Example 6 addresses the issue of how to model a waiting line process where the demand pattern changes over time. The third section introduces a new SimQuick element called a Decision Point. With this element we route objects in models in several different directions. The basics are presented in Example 7, an airport security system. Two more examples follow: a department of motor vehicles and a hospital emergency room. The last section introduces two additional features of SimQuick: resources and priorities. With these features, more complex waiting line processes can be modeled. This section contains two examples of fast-food restaurant processes. Section 1: Solving a problem with SimQuick Example 1: A bank Consider the following process within a small bank: customers enter the bank, get into a single line, are served by a teller, and finally leave the bank. Currently, this bank has one teller working from 9 A.M. to 11 A.M. Management is concerned that the wait in line seems to be too long. Therefore, they are considering two process improvement ideas: adding an additional teller

during these hours or installing a new automated check-reading machine that can help the single teller serve customers more quickly. Modeling the process (with a process flow map) Our first step in modeling this process is to construct a process flow map of the bank using the elements of SimQuick (see below). This is a conceptual step, so it can be done anywhere you prefer (e.g., on a piece of paper or on a computer). (The map below was made using some simple drawing tools within Excel; see Appendix 2 for details.) The flow map for the bank contains four elements, which are represented by boxes. For each element, the top line indicates the element type and the bottom line is a name (we follow this convention throughout the booklet). In this case, objects represent people and the arrows on the map indicate how the objects move between the elements. Note that the final element is a Buffer called Served Customers. An object entering this Buffer corresponds to a customer leaving the bank. The Buffer gives us an easy way to count our simulated served customers. You might have expected to see an Exit here, but with an Exit, objects leave the model according to a specified schedule instead of when they are ready to leave. (For example, an Exit can be used to model finished goods leaving a factory on trucks that depart periodically according to a schedule; Exits are introduced in Chapter 3.)

Process Flow Map for Bank Entrance Buffer Work Station Buffer Door Line Teller Served Customers

Entering the model into SimQuick We’re now ready to start entering our model into SimQuick. Launch the application Excel on your computer and, within Excel, open your copy of the SimQuick-v2.xls file. This file is on the disk accompanying this booklet; you can copy it or save it to your computer or network space (see Chapter 1, Section 4, for details). You should see the following screen, which is called the Control Panel:

You can save your work at any time. To do so, click on “File” in the menu and then “Save As.” Enter a new name (usually something to remind you of the process you are modeling) and designate a location on your computer or network space. Observe that there are a number of buttons on the Control Panel. In particular, there is one button for each type of element. We will be clicking on these buttons to enter information for each element in our model. This can be done in any order, but let’s do it in the same order in which the objects move. So click on the “Entrances” button. You should see the following screen:

For each Entrance in your model, you must fill in one table (working from left to right). So let’s fill in table #1. To begin, type Door into the “Name” cell.

In the next cell down, we specify when objects arrive at this Entrance. Traditionally, this is done by specifying the amount of time between arrivals (the so-called interarrival time). To do this, suppose we have spent some time observing the door of our bank and have compiled a list of actual times between customer arrivals. We discover that this list of numbers has a mean of 2 minutes and a histogram with the same shape as an exponential distribution (see Chapter 1, Section 3). (Thus, customers tend to arrive at the bank every 2 minutes.) The interarrival times of customers can then be approximated by numbers, generated randomly, from an exponential distribution with a mean of 2. Thus, we enter Exp(2) in this cell. General principle: Interarrival times of people at services can often be closely approximated by the exponential distribution. In the next cell down, we specify how many objects arrive at a time. In this case, let’s assume people usually arrive at the bank one at a time; thus, we enter a 1 here. (If there was uncertainty about the number of objects entering, we could enter one of our four distributions; in this case, SimQuick would round to give integers.) Next, we specify where objects go after this Entrance. From the process flow map, we see that objects go next to the Buffer whose name is Line, so under “Output destination(s)” enter Line (these rows should be filled in from the top down). The table should now look as follows:

Name → Door Time between arrivals → Exp(2) Num. objects per arrival → 1 Output destination(s) ↓

Line

Now click on “Return to Control Panel,” followed by “Buffers.” You should see the following screen:

In table #1, enter the name Line. In the next cell down, we must specify the maximum number of objects that can fit into this Buffer at one time. This is a small bank, so let’s say we can estimate this size as 8. So enter 8. The next cell down asks for the initial number of objects in this Buffer at the beginning of each simulation. Since the bank opens at 9 A.M., we enter 0 here. Now we have to specify where objects go next, the “Output destination(s)”; so enter Teller here. You are also asked to specify the “Output group size.” Because people leave the line one at a time, enter a 1 here. If people were leaving the line two at a time, then you would enter a 2 here (this feature is useful when, for example, you are putting objects into batches in a factory). The table should look as follows:

Name → Line Capacity → 8 Initial # objects → 0 Output Output destination(s) ↓ group size ↓

Teller 1

Note: If, at some time during the simulation, an object arrives at the Entrance and the Buffer is full (i.e., it contains 8 objects), then the object does not enter the process. Furthermore, it does not enter the process later in the simulation; hence, it effectively goes away. For our bank example, this represents a customer who arrives at the bank but immediately leaves because the line is too long (this is sometimes referred to as balking). Now, click on the “Return to Control Panel” button followed by the “Work Stations” button. You should see the following screen:

In the first table, enter the name Teller. Next, we describe the “Working time” of this teller. Let’s assume we have observed tellers in action for several days and discovered that their service time per customer can be approximated by numbers, generated randomly, from a normal distribution with a mean of 2.4 minutes and a standard deviation of .5 minutes. So enter Nor(2.4,.5) for the “Working time.”

For “Output destination(s),” enter Served Customers. For “# of output objects,” enter 1 because people leave the teller one at a time. (If, for example, this Work Station represented a machine that split objects into identical pieces, then we would enter a bigger integer.) The final two columns are not relevant for this model, so you can leave them blank. The table should look as follows:

Name → Teller Working time → Nor(2.4,.5)

Output # of output Resource Resource destination(s) ↓ objects ↓ name(s) ↓ # units needed ↓ Served Customers 1

Now click on the “Return to Control Panel” button followed by the “Buffers” button. In table #2, enter the name Served Customers. For the “Capacity,” enter the word Unlimited, which sets the capacity to a very large number that is sure to exceed the number of customers served from 9 A.M. to 11 A.M. Enter 0 as the “Initial # objects.” Objects in this Buffer have no output destination, so the Buffer tables should look as follows:

1 2

Name → Line Name → Served Customers Capacity → 8 Capacity → Unlimited Initial # objects → 0 Initial # objects → 0 Output Output Output Output destination(s) ↓ group size ↓ destination(s) ↓ group size ↓

Teller 1

Click on “Return to Control Panel.” The Control Panel contains the last two cells to be filled in. “Time units per simulation” asks us for the duration of each simulation. Because each simulation represents two hours and the time units we have been using for the Entrance and Work Station already refer to minutes, enter 120 here. General principle: A time unit in SimQuick can represent any real time interval: 1 second, 3.5 seconds, 1 minute, 1 hour, 2.5 days, and so on. However, time units must be used consistently throughout a SimQuick model (in all statistical distributions and in “Time units per simulation”). “Number of simulations” asks us for the number of times we want to simulate the 2-hour period. Because each simulation is based on randomly generated numbers, each simulation can yield different results. Hence, you typically want to do more than one simulation and to analyze the results by using means (and possibly some other statistics). Let’s do 30 simulations. This part of the Control Panel should look as follows:

General principle: As the amount of uncertainty in a model increases (i.e., the number of statistical distributions used and the amount of their variability), the number of simulations should increase. We are now ready to go, so click on the “Run Simulation(s)” button. It will probably take a few seconds (this depends on the speed of your computer). It will take longer as you increase the number of elements, the time units per simulation, and the number of simulations (see Appendix 4 for limits). Some messages will appear (in the lower right portion of the Control Panel), telling you what SimQuick is doing. If SimQuick seems to be taking too long, you may hit the Esc key at any time to abort. If SimQuick is running after 30 seconds, a window will open asking you how you want to proceed. If you made some typing mistakes, SimQuick will indicate where the problem is with an error message. You should correct the problem and then hit “Run Simulation(s)” again. A common mistake is entering an “Output destination” for an element that is not exactly the same as the “Name” of the element where you want the outputs to go. (Watch for extra spaces at the end, and make sure uppercase and lowercase match.) Clicking the “View Model” button on the Control Panel brings up a worksheet containing copies of all the element tables defining your model. This provides a convenient way to check the logic of your model and to print it. You cannot make changes to your model on this page. Interpreting SimQuick results When the simulations are complete, click on the “View Results” button. You should see a table similar to the following; the exact numbers will probably be different because SimQuick generated random numbers during the simulations. Your table will have 30 numbered columns; we only display the first five here.

Simulation Results

Element Element Statistics Overall Simulation number(s) types names means 1 2 3 4 5

Entrance(s) Door Objects entering process 53.88 58 56 55 57 48

Objects unable to enter 6.70 7 3 6 18 0

Service level 0.90 0.89 0.95 0.90 0.76 1.00

Work Station(s) Teller Final status NA Working Working Working Working Not Working

Final inventory (int. buff.) 0.00 0 0 0 0 0

Mean inventory (int. buff.) 0.00 0.00 0.00 0.00 0.00 0.00

Mean cycle time (int. buff.) 0.00 0.00 0.00 0.00 0.00 0.00

Work cycles started 48.43 51 49 49 50 48

Fraction time working 0.96 0.98 1.00 0.97 0.97 0.96

Fraction time blocked 0.00 0.00 0.00 0.00 0.00 0.00

Buffer(s) Line Objects leaving 48.43 51 49 49 50 48

Final inventory 5.45 7 7 6 7 0

Minimum inventory 0.00 0 0 0 0 0

Maximum inventory 7.77 8 8 8 8 7

Mean inventory 4.47 5.03 4.64 3.05 6.43 3.42

Mean cycle time 11.04 11.83 11.36 7.46 15.43 8.54

Served Customers Objects leaving 0.00 0 0 0 0 0

Final inventory 47.44 50 48 48 49 48

Minimum inventory 0.00 0 0 0 0 0

Maximum inventory 47.44 50 48 48 49 48

Mean inventory 22.75 22.84 24.19 22.71 23.29 22.89

Mean cycle time Infinite Infinite Infinite Infinite Infinite Infinite

Although, for the most part, the meaning of these results may be obvious, it’s worth discussing them in detail one time. In addition to summarizing the behavior of the model over time, the results also allow you to track down every object that was present during each simulation. The first two columns contain the element types and names in the SimQuick model. The third column contains the types of statistics collected during the simulations. Slightly different statistics are collected for each type of element. The columns labeled 1 through 5 contain the statistics collected for the first five of the 30 simulated 2-hour time periods. Each number in the column labeled “Overall means” is the mean of the 30 numbers to its right in the complete worksheet. For the first simulation, let’s consider the element Door, which is an Entrance. We learn that 58 objects entered the model, whereas 7 were unable to enter the model. This means that during our first simulated two hours, 65 (= 58 + 7) objects arrived at the model, 58 of these entered Line, and 7 left the process (or balked) because Line was full. As we noted above, these 7 objects did not return to the model during the simulation (if we wanted them to, then we would have to alter the model). The simulated service level is computed by SimQuick using the following formula:

enter tounable Objects process entering Objects process entering Objects level Service

+ =

Or, in this case, for simulation 1:

89. 758

58 level Service =

+ =

In words, the service level for each simulation is the fraction of demand that is satisfied. Again for the first simulation, let’s next consider the element Line, which is a Buffer. First, we notice that 51 objects left the Line during the simulation and went to the Teller. We see that at the end of the simulated two hours there were 7 objects left in the Line. Observe that 51 + 7 = 58 is the number of objects that entered the model at Door. You can also read the maximum, minimum, and mean number of objects in the Line during the simulation (referred to by the general term “inventory”). Finally, we see that the “Mean cycle time” was 11.04. The mean cycle time of an object at a Buffer during a simulation is the mean number of time units during the simulation that an object spent in the Line. In terms of our model, this represents the mean waiting time of a person in line. Next, let’s consider the statistics collected for the element Teller, which is a Work Station: “Final status” can be either Working or Not working, and simply describes what the Work Station is doing at the end of the simulation. “Final inventory (int. buff.)” refers to a small internal buffer that Work Stations have. A Work Station has enough room to hold one object on which it has finished working (if it outputs more than one object per work cycle, then this number of objects is the capacity of this internal buffer). The Teller’s final inventory (int. buff.) is always 0 because it can always pass finished objects to the Buffer called Served Customers. A Work Station will retain a finished object in its internal buffer when it cannot be passed to an output destination (e.g., a Buffer that is full or another Work Station that is working). If a Work Station has some objects in its internal buffer, then it cannot start working on a new object until this inventory drops to zero. Hence, a Work Station with some inventory in its internal buffer is called blocked. “Mean inventory (int. buff.)” and “Mean cycle time (int. buff.)” are statistics collected for the internal buffer and are analogous to the statistics we collect for Buffers. Note that “Mean cycle time (int. buff.)” does not include the working time at the Work Station. These two statistics are equal to zero in all the simulations in this example. “Work cycles started” indicates how many times the Work Station has begun working on an object, and “Fraction time working” indicates the fraction of the time of the entire simulation that the Work Station spent working (this is an important performance measure that is often called utilization in the context of machines in a factory). We see, for example, that the Teller is working 98% of the time during simulation 1 and all of the time during simulation 2. “Fraction time blocked” indicates the fraction of the time of the entire simulation that the Work Station was blocked. In the bank example, the Teller is never blocked. Interpreting the results for the Buffer called Served Customers is the same as for the Line. Note that the “Mean cycle time” is “Infinite” because no objects ever leave this Buffer.

Let’s focus on those statistics that are probably of most interest to the management of this bank. These will typically be numbers in the “Overall means” column, each of which is the mean of the 30 numbers to its right in the complete spreadsheet. At this bank, management is probably very interested in the statistic “Overall mean mean cycle time” of the Line (let’s just refer to this as the “overall mean cycle time”). This represents the mean amount of time a customer is waiting in line. In this example, the overall mean cycle time is 11.04 time units. A related statistic of interest is the “Overall mean cycle time” of the Line plus the Teller. This represents the mean amount of time a customer is both waiting and being served. This is easy to compute from the Results using the following formula:

Overall mean cycle time of Line plus Teller =

Overall mean cycle time of Line + Overall mean working time of Teller The overall mean working time of Teller is simply 2.4 time units (from Nor(2.4,.5)). Hence, the overall mean cycle time of the Line plus Teller equals 13.44 (= 11.04 + 2.4) time units. Another important statistic to management is the “Overall mean service level,” which represents the fraction of customers that enter the bank. In this example, the overall mean service level is .90. Also of interest is the “Overall mean mean inventory” (let’s just refer to this as the “overall mean inventory”) This represents the mean number of people in line and, in this example, is 4.47. We can see how busy our simulated teller was by looking at the “Overall mean fraction time working,” which in this example is .96. A final statistic to consider is the “Overall mean final inventory” of Served Customers. This represents the total number of customers served during the simulated two hours and is sometimes referred to as the throughput of the process. In this example, the overall mean final inventory is 47.44. Note: Consider any row of the Results worksheet. The number in the “Overall means” column of the Results worksheet is a summary statistic of all the numbers to its right. Another important statistical calculation with these numbers is a confidence interval, which contains the true overall mean (with a specified probability, under some reasonable assumptions). Finding a confidence interval (as well as other important statistics, like the standard deviation) is easy to do in Excel and is discussed in Appendix 2. Before using this model to experiment with changes in the process, management should make sure the results are close to what is observed in the real bank; this is called checking the validity of the model. The most important statistic to check in this case is probably the overall mean inventory or the overall mean cycle time of Line. If the model is not valid, then the structure and

statistical distributions in the model need to be checked against the real process. Let’s assume our model gives a good approximation of the real bank. Improving the process, Variation 1 Let’s see how much the overall mean cycle time of the Line can be reduced by adding a second teller to our model. Here is our new process flow map: Process Flow Map for Bank, Variation 1

Work Station Teller 1

Entrance Buffer Buffer Door Line Served Customers

Work Station Teller 2

The two Work Stations operate as follows: Suppose during a simulation there is an object in Line. If Teller 1 is not working, then this object will go to Teller 1. If Teller 1 is working and Teller 2 is not working, then this object will go to Teller 2. If both Teller 1 and 2 are working, then the object remains in Line until one of them finishes. Assume both tellers can serve customers at the same rate. Can you guess by how much the overall mean inventory and the overall mean cycle time of Line will drop from the one-teller case? Here is the new SimQuick model (note that Line now has two output destinations): Entrance:

Name → Door Time between arrivals → Exp(2) Num. Objects per arrival → 1 Output Destination(s) ↓

Line

Buffers:

1 2

Teller 1 1 Teller 2 1

Work Stations:

Name → Teller 1 Working time → Nor(2.4,.5)

Output # of output Resource Resource destination(s) ↓ objects ↓ name(s) ↓ # units needed ↓ Served Customers 1

Name → Teller 2 Working time → Nor(2.4,.5)

Output # of output Resource Resource destination(s) ↓ objects ↓ name(s) ↓ # units needed ↓ Served Customers 1

Here are the results from running 30 simulations with SimQuick:

Simulation Results

Element Element Statistics Overall Simulation number(s) types names means 1 2 3 4 5

Entrance(s) Door Objects entering process 60.32 64 57 60 65 55

Objects unable to enter 0.10 0 0 0 0 0

Service level 1.00 1.00 1.00 1.00 1.00 1.00

Work Station(s) Teller 1 Final status NA Working Not Working Working Working Working

Final inventory (int. buff.) 0.00 0 0 0 0 0

Mean inventory (int. buff.) 0.00 0.00 0.00 0.00 0.00 0.00

Mean cycle time (int. buff.) 0.00 0.00 0.00 0.00 0.00 0.00

Work cycles started 33.76 36 32 34 37 32

Fraction time working 0.67 0.70 0.68 0.69 0.73 0.66

Fraction time blocked 0.00 0.00 0.00 0.00 0.00 0.00

Teller 2 Final status NA Working Not Working Working Working

Not Working

Final inventory (int. buff.) 0.00 0 0 0 0 0

Mean inventory (int. buff.) 0.00 0.00 0.00 0.00 0.00 0.00

Mean cycle time (int. buff.) 0.00 0.00 0.00 0.00 0.00 0.00

Work cycles started 26.29 28 25 25 27 23

Fraction time working 0.52 0.56 0.51 0.50 0.50 0.50

Fraction time blocked 0.00 0.00 0.00 0.00 0.00 0.00

Buffer(s) Line Objects leaving 60.05 64 57 59 64 55

Final inventory 0.27 0 0 1 1 0

Minimum inventory 0.00 0 0 0 0 0

Maximum inventory 3.57 7 3 5 4 4

Mean inventory 0.37 0.82 0.45 0.39 0.36 0.58

Mean cycle time 0.71 1.53 0.95 0.80 0.67 1.27

Served Customers Objects leaving 0.00 0 0 0 0 0

Final inventory 58.93 62 57 57 62 54

Minimum inventory 0.00 0 0 0 0 0

Maximum inventory 58.93 62 57 57 62 54

Mean inventory 28.85 31.69 29.35 26.27 26.52 28.97

Mean cycle time Infinite Infinite Infinite Infinite Infinite Infinite

As before, a key statistic of interest is the overall mean cycle time for Line. The overall mean is now .71 time units. This is a clear improvement. We could use a t-test to prove that it is significant (see Appendix 2), but it is not necessary in this case. The overall mean cycle time of the Line plus the Tellers has decreased to .71 + 2.4 = 3.11. Observe that the overall mean service level has significantly increased to 1. Hence, all our simulated customers were able to enter the bank in each simulation. Also notice that the overall mean inventory of Line has significantly dropped to .37 objects. We see that the overall mean fraction time working of each teller (.67 and .52 time units) has dropped quite a bit. Finally, the overall mean final inventory of Served Customers, or the throughput, has increased to 58.93.

The post SimQuick Process Simulation with Excel appeared first on My Assignment Online.

Get Instant Help on WhatsApp