Chapter

17
Simulation

17.1 INTRODUCTION

book, we have discussed mathematical models to describe and analyse the

In the earlier chapters of this solved analytically to determine the optimal solutions
characteristics of agiven system. Such models could be decision-making
For example, we considered in Chapters 2 and 3how a given
or to describe the system. problem and then solved using simplex
technique to
be modelled as a linear programming of
problem could
solution. Similarly, in Chapter 10 we saw how a queuing system, that meets a certain setfact
obtain an optimal length, waiting time and so on. However,
despite the
assumptions, can be described in terms of the queue analyse a wide variety of problems, there are many
help to
that mathematical modelling can, and does this way. For instance, a given queuing situation may not meet
situations which are too complex to be handled modelling. In these
assumptions of any of the queuing models and, therefore, not lend itself to only this.
all the given system and study its behaviour. Not
circumstances, it is often possible to simulate the changes when experimentation with the
of system
simulation may also be employed to estimate the impact
real system may not be feasible. phenomenon
to imitate. In general terms, simulation involves developing a model of somereal
To simulate is technique. In
performing experiments on the model evolved. It is adescriptive, and not optimising, in that
and then constants associated with it are manipulated
simulation, a given system is copied and the variables and To illustrate, for aerodynamic testing, scaled
artificial environment to examine the behaviour of the [Link]. Using a wind tunnel, air is blown to
in the wind
down models of aeroplanes are built and placed
model.
examine the aerodynamic properties of the
 864 Quantitative Techniques in Munagement

Using siulation, an analyst can introduce the constants and variables related to the problem, set un s.
possible course of action and establish eriteria which act as measures of cffcctiveness. The benefit
simulation from he viewpointof the analyst stems fromn the fact that the results of taking aparticular course
of action can be estimated prior to its implementation in the real world. Instead of using hunches and
intuition to determine what may happen, the analyst using simulation can test and evaluate Varion
alternatives and sclect thc one that gives the best results.

O.2 PROCESS OF SIMULATION

Broadly. there are four phases of the simulation process. They are:
(a) definition of the problem and statement of objectives,
(b) construction of an appropriate model.
(c) experimentation with the model constructed; and
(d) evaluation of the results of simulation.
Each of the phases calls for the performance of a number of preliminary tasks. Of these, the two
are major tasks
collection of data and selection of means by which the simulation activity would replicate the random
behaviour of the real world.
The first step in problem solving of any situation is to identify and clearly define the
problem and list the
objective(s) that the solution is intended to achieve. This is true of simulation as well. A clear statement not
only facilitates he development of an appropriate model but also provides a basis for evaluation of the
simulation results. In general, simulation aims to determine how the system under
behave under certain conditions. Naturally, the more specific the analyst is about what consideration would
he is looking for, the
greater the chances that the simulation model will be designed to
level of detail of the simulation should be decided upon accomplish that. Thus, the scope and the
carefully.
The nexI step in simultaion is the development of a suitable model. During the course of a
model nmimics the important elements of what is being simulated. A simulation model simulation, the
may be a physical or
mathematical model, a mental conception, or acombination. Many simulations
Examples include a scaled down model of an aeroplane or ship constructed out ofinvolve physical models.
wood or other material.
Since physical models are relatively expensive to build, mathematical models are often
model, mathematical symbols or equations are used to represent the preferred. In such a
relationships in the system.
Collection of data is a significant aspect of model development, and the quantum and type of data
directly governed by the scope and extent of the detailof the simulation. The data are needed bothneeded are
for model
development and evaluation. Obviously, the model for simulation must be so designed that it would enable
evaluation of the key decision alternatives. An ancillary step here is of designing
experiments help answer the 'what if
experiments. The
types of questions in simulation studies. By going through this
process, the analyst is able to learn about the system behaviour.
Once the simulation model is developed, the next step is to run it. If the model is deterministic, with all its
parameters known and constant, then only a single run would suffice. On the other hand, if the simulation
is stochastic in nature, with the parameters subject to random variation, then a number of runs
would be
needed to get aclear picture of the model performance. The probabilistic simulation is akin to the random
sampling where cach run represents one observation. Thus, statistical theory can be used to determine the
optimal sample sizes. Evidently, the greater the variability inherent in the simulation results, the larger
would be the simulation runs needed to obtain a reasonable degree of confidence that the results are truly
indicative of the system behaviour.
 Simulation 865

Thelast step
in the process of simulation is to analyse and interpret the results of the runs. The interpretation
the reality.
is,in alarge mcasure, dependent on the extent to which the simulation model portrays nced for
of results closeri
Obviously, the apprOximation of the rcal systen by the simulation model. lesser will be the
adjustingtheresults and also lesser will be the risk inherent in applying the results.

SIMULATION
17.3 MONTE CARLO
probabilistic simulation usingthe
Although sinulation can be of many types, our discussion will focus onthe numerical technique that
Monte Carlo method. Also called computer simulation, it can be described as a behaviour. Thechance
involves modelling astochastic system with the objective of predicting thc system's
can be used when the given
elenent is a very significant fcature of Monte Carlo simulation and this approach
process has, arandon, or chance, component.
the original data with random
In using the Monte Carlo method. a given problem is solved by simulating carlier. we must have a model.
number generators. Basically, its Use requires two things. First. as mentioned
model refers to the probabílity distribution
that represents and image of the reality of the situation. Here the
may not be known to explicitly follow
of the variable in question. What is significant here is that the variable
any of the theoretical distributions like Poisson, Normal, and so on.
The distribution may be obtained by
bakery keeps a record of the sale of the
direct observation or from past records. To illustrate, suppose that a
number of cakes of a certain type. Information relating to 200 days' sales is,
10 11 12 Total
Demand (No. of cakes) : 5 7

16 50 62 38 12 200
(No. of days) 4 10
demand for the cake, we can derive the
Assuming that this is an adequate representation of the distribution of
of the frequencies in terms of proportions. This is done
probability distribution of demand by expressing each frequency. The resultant distribution follows:
by dividing each one of the values by 200-the total
7 8 10 11 12
Demand (No. of cakes) 5 6

0.08 0.25 0.31 0.19 0.06 0.04

0.02 0.05
Probability
demanded on a day, a 0.05 or 5% chance that the
Thus, there is 0.02 or 2% chance that 5 cakes would be situation under
demand would be for 6 cakes and so on. This distribution would serve as the model of the
consideration.
modelsomething to capture the
The second thing required for simulation is a mechanism to simulate the
procedure that would help us to select.
random nature of the given systenm. Thus, we should have available a
state of the system. Such a
at random, values for the variables which can be used to approximate the
procedure by which random
mechanism can be any random number generator consisting of a device or a
numbers can be determined and/or selected.
There are various ways in which random numbers (or apparently random, but not truly so) may be generated.
numbers, midsguare
These could be: result of some device like coin or die; published tables of random
numbers generated
method, or some other sophisticated method. It may be mentioned here that the `random
called pseudo-random
by some methods may not be really random in nature. In fact such numbers are we shall not
numbers. There are some tests with which numbers may be tested for their randomness but
obtained and used.
Consider them here and consider only briefly how the numbers may be
the arrow is
One way to generate random numbers is to fix up a spinning arrow on a common clock. When
Spun, the number on which it stops would be taken to be random number for that trial. Naturally, anynumber
 866 Quantitative Techniqucs inManagement
In a si
of the arow would result in an cqual number of random numbers. similar way, random
of spinnings
numbers can be gencrated using spinning of aroulette wheel, tossing dice.. . and likke that. Although simple,
these are very slow methods and cannot mcet the practical requirements where a large number of random

numbers may be needed.

Amore fast and convenient method is to make use of the published
tablcs of random numbers. like the
random number table is a
published by the Rand Corporation (of USA): A Million Random Digits. A this table are in rand
efficient way to generatc random data in most situations. The numbers in
arrangements. The underlying theory is that cach number has an equal opportunity of being selected.
Stillon a more sophisticated level, computers are used for generating the random numbers. With computers
it is typically casier to generate random numbers by an arithmetic process as needed rather than to read the
numbers from stored table. An carly. probably the earliest, method proposed for use on digital computers to
generate random numbers is the midsquare. To illustrate this method, suppose that we wish to generate four
digitintegers and the last number generated was 8937. To obtain the next number, in the sequence, we square
the last one and use the middle four digits of the product, In this case the product is 79869969 so that the next
pseudo-number is [Link] nextfew numbers in the sequence are 6726, 2390, 7121and so on. Thus, using
this method. having drawn up a suitable computer programme, a four-igit number may be fed into the
computer and a list of pscudo-random numbers obtained.
Of all the random number generators, we shall make use of randon1 number tables for demonstrating the
simulation process. In particular, we shall use the random number table (Table B7) given in Appendix Bof
the book. To consider how the table can be used for generating data relating to our bakery problem, we
proceed as follows:
Step 1 An assignment has to be worked out so that the intervals of random numbers will correspond to the
probability distribution. Here, since the probabilities have been calculated to two decimal places, which add
up to 1.00, we need 100 numbers of two digits to represent each point of probability. Thus, we take random
numbers 00 through 99 to represent them. Now, as the probability of 5 cakes is equal to 0.02, we assign two
random numbers 00-01 to this demand level; the probability of 6cakes being equal to 0.05, the nextfive
numbers, 02-06 would be assigned to this level. In a similar manner, each of the demand levels would be
assigned appropriate intervals as given here. It may be mentioned that cumulative probabilities shown are
calculated to ease the determination of the random number intervals. The cumulative probabilities column
allows the assigned numbers to correspond to the same probability range for each event.
Demand (No. of cakes) Probability Cumulative Probability Random Number Interval
5 0.02 0.02 00-01
6 0.05 0.07 02-06
7 0.08 0.15 07-14
8 0.25 0.40 15-39
0.31 0.71 40-70
10 0.19 0.90 71-89
11 0.06 0.96 90-95
12 0.04 1.00 96-99
Instead, if probabilities are calculated to three decimal places, then 3-digit random numbers would be
required ... and so on.
 867
Simulation

pattern for drawing random

Step 2 Once the random number intervals are determined, we select atracking of the table and read
the
numbers from the random number table. We may start with any column and row
values in any set manner--horizontally, vertically, or diagonally. Using the pattern, we draw the rando
decide for example. to read every third value
numbers and match them with the assigned events. We mav
of random numbers. The random
horizontally, starting with the fifth column and fourth row of the table
52. 07 and so on. We draW as may
numbers, according to this pattern are 61, 74, 24. 03. 59. 16. 84. 92.
random numbers as the number of day's demand is required to be simulated.
corresponding to the demand level of 9 unitS.
The first of the list of the numbers, 61,lies in the interval 40-70
similar manner, we can obtain the demanO To
Thus, the simulated demand for the first day is 9 cakes. In a
cach of the days. For the 10-day period, we have the following demand:
10
1 2 3 4 5 6 7
Day 07
16 84 92 52
Random Number 61 74 24 03 59
11 9 7
8 6 9 8 10
Demand (cakes) 10
random numbers, we now proceed to illustrate tne
Having considered the method of generating data with
simulation of an inventory and a queuing system. They follow in turn.

17.4 SIMULATION OF AN INVENTORY SYSTEM

was pointed out that when variation is
observed
Chapter 9, it
In our discussion of inventory management in of
in both, the demand and the lead time, then the analytical approach gives way to the technique
inventory
simulation can be employed to model an
simulation. We now consider such a situation and see how
system.
distribution of
certain product for which the probability
Example 17.1 Consider the case of a dealer of a both developed empirically by observations
the probability distribution of the lead time,
daily demand and
made over a long span of period, are as follows:
Probability distribution of daily demand
8 10 11 12
3 4 5 6 7
Units demanded :
0.19 0.13 0.10 0.08 0.07 0.06
0.02 0.08 0.11 0.16
Probability
Probability distribution of lead time
3 4 5
2
Lead time (days)
0.20 0.30 0.35 0.15
Probability Rs 2. while
order, the holding cost per unit per day is estimated at
The ordering cost is known to be Rs 80 per
loss in profits is Rs 20 per unit per day.
the unit shortage cost, representing the the total
re-order levels and re-order quantities, what would be
The dealer is anxious to known, for specific selecting an appropriate
shortage costs), and thereby
inventory costs (made up of the ordering, holding, and
combination of the two.
which calls for a re-order quantity of 40 units and a
For our illustration, we shall evaluate a simulation planbalance of 30 units.
inventory
re-order level of 20 units, with a beginning
relates the 1dentified values of the variable(s) to the
As a first step, we shall assign a coding system that both the given distributions are calculated
random numbers. As discussed before, since the probabilities in
868 Ouantitative Techniques in Management

and given to two decimal places, the coding system to be uscd would be based on
these
consist of two-digit random numbers: 00through99. The random number coding for each ofprobabilities
and
the distributione
shall be as given in Tables 17.1 and 17.2.

Table 17.1 Random Number Coding: Demand Distribution

Daily Dennand Probabiliy Cumulative Probabiliry Random Number Iterval

0.02 0.02 00-01
4 0.08 0.10 02-09
0.11 0.21 10-20
0.16 0.37 21-36
7 0.19 0.56 37-55
0.13 0.69 56-68
0.10 0.79 69-78
10 0.08 0.87 79-86
0.07 0.94 87-93
12 0.06 1.00 94-99

Table 17.2 Random Number Coding: Lead Time Distribution

Lead Time (days)Probabiliny Ciumulative Probabiliny Random Number Interval
2 0.20 0.20 00-19
3 0.30 0.50 20 49
4 0.35 0.85 50-84
5 0.15 1.00 85-99
Having determined the random number intervals, we select an appropriate random number table to draw
numbers from. Also, we decide the tracking pattern for identifying the random numbers. For this illustration,
again we shall use the random number table (Table B7) and simulate the data on demand using
random
number read column-wise from the top of the sixth column. The simulation of the lead time would be
similarly done by reading numbers column-wise from the last column of the table. The results of the 40-day
simulation are given in Table 17.3.
The second columnof the table contains the random numbers drawn in respect of the daily demand while the
corresponding demand for each day, determined in accordance with the random number coding given in
Table 17.1. For instance, the random number 68, for the day 1, lies. in the interval 56-68 indicating a
demand of 8 units. With an initial balance of 30 units as indicated in column 7 of the table, and a demand of
8 units on the first day, the number of units in the stock would be 22 for this day, involving a holding cost of
Rs 44 given in column 9. The balance would reduce to 17 units when the 5 units of demand simulated for
the day 2 are sold. On this day, an order for 40 units be given for an ordering cost of Rs 80 (column 8). The
 Simulation 869

Table 17.3 Simulation Worksheet

R No. Demand RNo. L Time H
Day Receipts Balance
(10)
2) (4) (5) (6) (7) (8) (9)
30
68 22 44
13 47 3 34
(09 4 26
|3
4 20 5 J6
73 40 39 78
07 4 70
35
7 92 11 24 48
Q0 12 74 4 12 80 24
1 1 2
18 5 80
10

24 120
12 6 40 34 68
07 4 30 60
14 29 24 48
15 57 8 25 16 80 32

33 20
16
49 7 3 6
17
65 8 40 35 70
18
92 11 24 48
19
20 98 12 21 3 12 80 24

21 00 3
57 2
22
23 12 40 36 72
24 31 6 30 60
25 96 12 47 3 18 80 36

85 10 16
26
60
27 92 11
40 29 58
28 91 11
29 77 69 4 20 80 40
30 37 7 13 26

31 6 7 14
34
2 4
32 11 5
33 27 6 40 36 72
34 10 31 62
35 8 23 46
59
2 17 80 34
36 33 6 09
37 6 12
87
38 40 37 74
72 9
39 28 56
73
40 80 4 18 80 36
79 1C
Total 640 1,456 260

0:Ordering cost H: Holding cost S: Shortage cost

870 Quantitative Techniques in Management

random number chosen for simulating the lead time being 47, the lead time would be 3 days, and the 40 units
ordered would, thcrefore, be received on day 5, as entered in column 6. We proceed in this manner to
discover that on day 9 there is one unit in the stock. The demand of 5 units on day 10 indicates that it would
be an out-of-stock situation. The shortagc of 4 units would result in a stock out cost of Rs 80, given in the last
column of the table. The conditionof bcing out-of-stock would again prevail on the next day.
the
Compleing a 40-day period, we find that the total ordering cost = Rs 640, the holding cost = Rs 1,456 and
out-of-stock cost = Rs 260, the three adding upto Rs 2,356.
Performing the same feat with re-oder level of 15 units and the same re-order quantity of 40 units (the
random numbers being the same) we obtain a total cost egqual to Rs 2,358 comprising: ordering cost Rs 560,
having
holding cost Rs 1,198, and out-of-stock cost Rs 600. In a similar manner, the total cost obtained from
a policy of re-order level cqual to 20 units and the re-order quantity of 30 units, equals Rs 2,730 (ordering
cost: Rs 720, holding cost: Rs 1,090 and stock-out cost: Rs 920).

In a problem like this, which requires the determination of two parameters (order quantity and re-order level),
values. We
the approach is to set up a two-way table of costs with various combinations of the two parameter
would simulate cach combination a large number of times, compute the total cost corresponding to each one
and enter it into the table. This procedure will enable us to locate the neighborhood of the optimum and select
the plan with approximately the lowest cost.
It may be observed here that the cost estimates for a given combination of order quantity, re-order level and
starting inventory level depend upon the tracking pattern adopted and the starting point chosen. It is desirable,
therefore, that a number of cost estimates be made by varying the starting point and the tracking pattern, and
then and average value may be determined. Alternatively, a large number of days' data may be simulated in
order to ensure more dependable cost estimates.

17.5 SIMULATION OF QUEUING SYSTEM

In Chapter 10 we considered the queuing situations. The characteristics of the situation described were based
on the assumptions, among others, that the arrivals are Poisson distributed while the service times follow an
exponential distribution. Although such assumptions are often appropriate for the real life queuing systems,
they are not invariably so. Thus, where such assumptions fail to apply, formal mathematical analysis can be
extremely difficult and even impossible. Simulation provides the answer in such cases.
We shall illustrate the use of simulation in the study of queues through the following example.

Example 17.2 In a large workshop undertaking servicing jobs, the mechanics obtain their tools and
equirements from a central store. The manager, perturbed about the waiting time of mechanics, is in the
process of determining whether more attendants be hired for the store for raising the level of service. The idle
time cost for the mechanics and the wages required to be paid to the attendants being known, he wishes to
ascertain how many attendants may be employed to minimise the total cost involved. For helping the manager
solve his problem using simulation, we proceed as follows. The following dataon the times between successive
arrivals and the service times for the mechanics have been obtained from the past 200 observations made on
the system presently in operation.
 Simulation 871

time:
Distribution of inter-arrival

Time (Minutes) Frequency Probability

12 0.06
3 18 0.09

6 50 0.25

74 0.37

12 32 0.16

15 14 0.07

Total 200

Distribution of service time:

Time (Minutes) Frequency Probability

4 0.04

6 20 0.10

36 0.18

10 88 0.44

12 48 0.24

Total 200

based on the
As usual, in the first step we assign random numbers to each observed arrival time interval
likelihood of the occurrence of each time interval as shown in Table 17.4. For instance, the probability of
arrival within 0 minutes of the last arrival (implying simultaneous arrivals) is 0.06. Therefore, we assign the
two-digit random numbers 00 through 05 to the event of 0 minutes between the arrivals. To facilitate this
process, as before, cumulative probabilities have been determined as given in the third column of the table.

Table 17.4 Random Number Coding: Inter-arrival Times

Time (minutes) Probability Cumulatiye Probability Random Nunber Interval

0.06 0.06 00-05

3 0.09 0.15 06-14

6 0.25 0.40 15-39

9 0.37 0.77 40-76

12 0.16 0.93 77-92

15 0.07 1.00 93-99

The same procedure is followed for the service time distribution. Table 17.5 depicts the cumulative
probabilities and the random number coding for cach of the observed service times. The table shows. for
872 Quantitative Techniques in Management

instance, that the probability of service time of 4 minutes is 0.04. Thus the random numbers o0 02
assigned to the event that the service time is 4 minuts. If a random number in the range of 00-03 is dra
from the random number table, the event that the service time is 4 minutes' is assumed to have
occurest
Table 17.5 Random Number Coding: Service Times

Time (minutes) Probability Cunulative Probabiliby Random Number Interval

4 0.04 0.04 00-03
0.10 0.14 04-13
0.18 0.32 14-31
10 0.44 0.76 32-75
12 0.24 1.00 76-99
Now wc are rcady to simulate the operation. To
determine the arrival times of the successive mechanics, we
start with the fifth major column of the table of random
numbers, and for the service times, we begin with the
fourth major column, and read vertically down in each case.
decided to be made, we draw 25 random numbers for cach, Supposing that a total of 25 observations are
the inter-arrival times and the service times.
These numbers are given in columns 2 and 4, respectively of
random numbers, as determined from the Tables 17.4 and 17.5, areTable 17.6. The times corresponding to the
contained in columns 3 and 5, respectively.
Now, we start with the first mechanic and observe that the
random number for the arrival is 58. This lies in
the interval 40-76 and therefore corresponds to the
time
starts at 8.00 anm, the first arrival at the store would be at period of 9 minutes. Assuming that the workshop
8.09
being 87, which lies on the interval 76-99 in the Table 17.5, a am. Further, the random number for service
the service to him begins at 8.09 am and ends at 8.21 service time of 12 minutes is suggested. Thus
am. Obviously. there would be no waiting time for the
mechanic. The second of the arrivals' random number is 47 which also
minutes. Thus, he would come to the store at 8.18 am9 indicates an inter-arrival time of 9
minutes after the first arrival. Since the store
be free at 8.21 am from the first
mechanic, the service to the second would start at that time. Thus, the would
time for him would be 3 minutes. Corresponding to the waiting
the time needed is 10 minutes. This means that the random number 39 for service time for this mechanic,
service to this mechanic would go from 8.21 to 8.31 am.
Notice that when the first mechanic is being serviced
would have to wait. Thus, the queue length over this between 8.09 and 8.21 am, then the second mechanic
interval would be taken to be 1.
In the same way we proceed further and
determine the
contained in the columns 4 and 7 of the Table 17.6. Queuearrival length
and waiting times for the mechanics, as
obtained by considering the number of mechanics reaching upto the for each interval of the service times is
being currently serviced. upper limit of that interval after the one
From the column showing the pattern of the
queue length, we can obtain the average waiting mechanics' waiting time, and the last column containing the
time
each one by the number of observations, 25. We get, and the average queue length by dividing the total of
the
the average queue length=73/25=2.92 mechanics. average waiting time = 623/25 = 24.92 minutes, and
Using this information, the manager can determine the total
10. For examining other alternatives, he can do so cost in the manner as we considered in Chapter
by estimating the new service time
additional attendants. He would have to simulate the new alternative to patterns based on
new service patterns and choose the best one. determine the outcome based on the
 Simulation 873

Table 17.6 Simulation Worksheet

Arrivals Service Service Oueue
S. No Waiting
R No. Time Oclock R No, Time Time (mts) Begins Ends Lengtl
() (3) (4) (S) (6) (7) (8) 9

58 8.09 am 87 12 00 8.09 am 8.21 am 1

47 8.18 39 10 03 8.21 8.31

23 6 8.24 28 07 8.31 8.39 2
3
69 9 8.33 97 12 06 8.39 8.51 2
4
35 6 8.39 69 10 12 8.51 9.01 2
5

55 8.48 87 12 1| 9.01 9.13 2

69 9 8.57 52 10 16 9.13 9.23 2

7
8 90 12 9.09 52 10 14 9.23 9.33
86 12 9.21 15 8 12 9.33 9.41
74 9 9.30 85 12 11 9.41 9.53 2
10

6 9.36 41 10 17 9.53 10.03 2

11 39
9.42 82 12 21 10.03 10.15 3
12 15
12 9.54 98 12 21 10.15 10.27 4
13 90
15 10.09 99 12 18 10.27 10.39 4
14 98
6 10.15 23 24 10.39 10.47 4
15 39

16 3 10.18 77 12 29 10.47 10.59

16
52 9 10.27 42 10 32 10.59 11.09 6
17
18 56 9 10.36 60 10 33 11.09 11.19 6
6 10.42 22 37 11.19 11.27
19 21
20 23 6 10.48 91 12 39 11.27 11.39

21 00 10.48 68 10 51 11.39 11.49 4

22 87 12 11.00 36 10 49 11.49 11.59 3

23 20 6 11.06 22 53 11.59 12.07 pm 2

24 40 9 11.15 92 12 52 12.07 pm 12.19 1

25 73 9 11.24 34 10 55 12.19 12.29

Total 623 mts 73

then sinmulation can be used to test has to develop the proposed solution:
the relative desirability of those solutions.
17.7
APPLICATIONS OF SIMULATION
In spite of its limitations, simulation is a
very
quantitative tool for solving decision problems. Itpotent., flexible, and, therefore, a commonly employed
problenms. To count a few, its use extends to areashaslikebeen applied successfully to a broad spectrum of
emergency vehicles like ambulances; making inventory policy police dispatching and beat design: location of
at airports; in financial decisions; evaluation of operating alternatives
planning-both portfolio selection and capital budgeting, scheduling the production
processes; large scale military battles as well as individual
both the weapon systems and of the strategicand weapons systems for aiding in the designing of
where the experimentation with the real situationtactical operations. Simulation is of paramount importance
is risky. The use of
likely bchaviour of the nuclear reactors during computer simulation for studying the
or even scaled down reactor models accidents case in point. Obviously, testing actual reactors
is a
under emergency conditions would involve
Thus, its application in situations which are too excessive risks.
complex permit mathematical solution, its
to
use in varied fields, its power in
enabling the decision-maker to conduct experimentssimplicity
that
and
help in
understanding the system's behaviour while avoiding the risks associated with
model's real-life counterpart, and also availability of extensive conducting tests on the
factors which account for the popularity of simulation computer software packages, are the man
technique.
Revieuw Illustrations

Example 17.3 The Tit-Fit Scientific Laboratories is engaged in producing

different types of high class
equipment for use in science laboratories. The company has two different assembly lines to produce its most
popular product 'Pressurex'. The processing timne for each of the assembly lines is regarded as a
random
variable and is described by the following distributions.
 Simulation 875

Process Time (minutes) Assembly A, Assembly Ag

10 0.10 0.20

11 0.15 0.40

12 0.40 0.20

13 0.25 0.15

14 0.10 0.05

process times for 15 units of the item and compute

Using the following random numbers, generate data on the
purpose, read the numbers vertically taking the first two
the expected process time for the product. For the
processing time on assembly A2
diaits for the processing time on assembly A, and the last two digits for
4134 8343 3602 7505 7428

7476 1183 9445 0089 3424

4943 1915 5415 0880 9309

assign random number intervals to the processing times on each of the assemblies. This
In the first stage, we
is shown in Table 17.7.
Times
Table 17.7 Random Number Coding for Process
Assembly A Assenbly A
Time (nts.) RW Interval
Prob. Cum. Prob. RNInterval Prob. Cum, Prob.

00-09 0.20 0.20 00-19

10 0.10 0.10
0.60 20-59
0.15 0.25 10-24 0.40
11
25-64 0.20 0.80 60-79
12 0.40 0.65
65-89 0.15 0.95 80-94
13 0.25 0.90

90-99 0.05 1.00 95-99

14 0.10 1.00

Thus, the
shows the simulated data on the process times.
The simulation worksheet given in Table 17.8
respectively, for the assemblies A, and A,. From the
random numbers for the first unit are 41 and 34,
are 12 and 11 minutes respectively. Thus the
Table 17.7 we observe that the times corresponding to these determined
times for the other 14 units are
total time required for the unit is 23 minutes. In the same way, the
time for a unit works out to be 23.27
and shown in the last column of the table. The expected completion
minutes.
876 Quantitative Techniques in Management

Table 17.8 Simulation Worksheet

Unit Assembly Aj AssemblA Total Time

R. Number Time R Number Time (mts.)
41 12 34 23
2 74 13 76 12 25
49 12 43 23

4 83 13 43 11 24

5 11 11 83 13 24

19 11 15 10 21

7 36 12 02 10 22

8 94 14 45 11 25

12 15 10 22
9 54
13 05 10 23
10 75
10 89 13 23
11 00
80 13 23
12 08 10
13 28 11 24
13 74
24 11 23
14 34 12
09 10 24
15 93 14
Expected time = 349/15
= 23.27