जाहिरात
जाहिरात
AN INTRODUCTION TO MODELING WITH INS
ABSTRACT
INS (the Integrated Network Simulation Tan-
guage) facilitates modeling by freeing the mod-
eler from the mechanics of programming a simula-
tion model. Instead, the modeler characterizes
the system as a network containing easily defined
nodes and branches. The network is easily trans-
lated to INS statements from which INS constructs
and executes the simulation model. Output is
automatically provided to reflect various per-
spectives of system behavior. Features of INS
allow easy enhancement of statistics and embel-
lishment of the model. INS is easy to learn and
potentially applicable to a variety of environ-
ments.
MODELING WITH INS
Simulation modeling is a high level task in
which the modeler seeks to represent the behavior
of a system mathematically and statistically.
The simulation model will usually require the
construction of a computer program, often utiliz-
ing a simulation language. But if the simula-
tion language requires the modeler to code
events, manipulate files, generate random devi-
ates, monitor status, collect statistics, and
format reports, then modeling can become a sec-
ondary task to programming. For the novice or
infrequent simulation user, the programming re-
quirement may impede modeling because it can re-
quire considerable effort to translate the state-
ment of the simulation problem into a simulation
program ready for computer processing.
INS, the Integrated Network Simulation lan-
guage, was created to provide a high level simu-
lation language that would facilitate modeling.
The INS modeler conceptualizes the simulation
problem graphically as a network, employing a set
of predefined modeling symbols or nodes. The
modeling symbols represent such real world pro-
cesses as queues and activities enabling the mod-
eler to abstract directly from the real system to
the INS network. The difficulties in translating
the network into computer-readable form are mini-
mized since the INS symbols correspond almost
directly to INS statements which INS interprets.
INS automatically executes the simulation and
provides output. Thus the INS modeler can focus
his attention on the important problems of mod-
eling and ignore the mechanics of simulation pro-
gramming.
Stephen D. Roberts.
James Schier
Thom Sulanke
Because modeling is a process of successive
refinement, there is a need to be able to quickly
produce a working “first cut" model and to embel-
lish this model as greater insight is generated.
INS facilitates embellishments because modifica-
tions are easily incorporated by changing speci-
fications at a node or altering segments of the
network. In this way the network becomes a mod-
eling vehicle, useful not only in the creation of
the simulation but also in giving structure to
the modeling process.
Since INS is concise, easy to document, and
visually appealing, INS models are also excellent
vehicles for communication between modelers and
their clients. Such communication can encourage
clients and decision makers to participate more
directly in the modeling process and consequently
improves the chances of eventual implementation.
To illustrate the fundamentals of modeling
with INS, we will consider an example. It will
be enriched to demonstrate the ease of modifying
INS models. Some discussion will also focus on
INS output and its enhancement. Space does not
permit a detailed discussion of many of these
aga ne you are referred to the INS User's Man-
ua .
A TV INSPECTION AND ADJUSTMENT EXAMPLE
Television sets which have ¿just been assem-
bled are routed to an inspection station for
final testing. If they fail the inspection, they
are sent to an adjustor who, after fixing them,
routes them back for inspection again. It has.
been observed that 15% of the sets fail inspec-
tion, regardless of their previous adjustment.
Suppose TVs arrive every 12 minutes and it takes
10 minutes to inspect, and 30 minutes to adjust a
TV, if adjustment is required. Conceptually, we
might depict the problem as the flow diagram in
Figure 1.
The flow diagram is valuable because it pro-
vides a visual interpretation of the process. In
fact, it is natural to define a problem by de-
scribing its processes or procedures. The dia-
gram serves to interrelate the processes and re-
mind us of the entire system. But the flow dia-
gram is too ill-defined. Its expression is high-
ly dependent on the individual and his choice of
symbols to represent the problem.
343
MODELING WITH INS . . . Continued
INS NETWORK
INS builds on the notion that a graphic model
is a natural way to describe a problem. “To
standardize the format of expression, INS, pro-
vides the modeler with a set of parameterized
symbols that specify élemental processes. The
modeler selects appropriate symbols and inte-
grates them into a network by connecting the
nodes with branches. For example, an INS network
corresponding to the previous problem is con-
tained in Figure 2.
Several features should be noted. The INS
network corresponds closely to the conceptual
flow diagram. The INS symbols represent elemen-
tal but familiar processes. Node numbers are
printed in bold letters and are ¡located on the
left side of the node symbol. Node 1 is a “Source
node and depicts the arrival of TVs. Nodes 2 and
4 are Activity nodes representing the delay of
TVs in inspection and adjustment respectively.
Node 3 is a Sink node and it depicts the depar-
ture of TVs. Notice that the data relevant to
“the problem is contained in the network, an ‘im-
provement over the flow diagram. For example,
TVs. arrive with a 12 minute interarrival time at
Source 1, adjustment takes 30 minutes at Activity
4, etc. Branches connect the nodes and represent
the possible paths for the TVs. Branching in INS
js determined by a branching mode specification
located where the branch leaves the node. Only
Activity 2 requires a branching decision and here
branching is performed PRObabilistically. Speci-
fications of the branching mode are listed on the
branches, so we see that 85% of the TVs flow: to
Sink 3 and 15% flow to Activity 4. |
The INS network translates immediately to INS
statements as shown in Figure 3. It is these
statements which INS interprets and checks for
errors. By providing a little additional infor-
mation on the creation of TVs and how to termi-
nate the simulation, the above model can be simu-
lated as it is written and we would obtain sta-
tistics which we could use to make further en-
hancements. Notice that the INS statements are
themselves quite readable and the input is
sparse. Employing the documentation potential,
the INS model can be communicated easily to
others.
Without any additional statements, INS pro-
vides default output, which in this case would be
the average time TVs are in the network, the av-
erage time TVs are in each activity, and the
average number of TVs in each activity. Also, the
number of times TVs encounter each node in the
network would be printed. INS supplies values
for most network specifications by default,
printing all of these in an echo and generating a
full array of output without special instruc-
tions. Our design philosophy is that INS .-should
do as much for a user as possible and should al-
low the user to easily alter the model and output
after examining a working simulation. |
TRANSACTIONS
In a more generic sense, the units of traffic
which flow through an INS network are called
344
transactions. Transactions can be many and var-
ied, all flowing in the same network. For exam-
ple, the production system may manufacture -dif-
ferent types of TVs.
To permit the distinction between trans-
actions, each transaction has a set of system-
defined attributes. Recognized by their first
three characters, CREation time, and transaction
TYPe are examples. These attributes become
available to the modeler when the transaction is
created. Furthermore, TYPe can be assigned by
the modeler and can be changed during the trans-
action's flow through the network. For example,
if two types of TVs arrive to the inspection sta-
tion, we might alter the arrival of the TVs so
that one Source node generates TVs of type 1 and
the other TVs of type 2. Figure 4 contains these
changes.
Now Source 1 generates TVs of TYPe = 1 and
Source 5 generates TVs of TYPe = 2 both with in-
terarrival times of 24 minutes. We have further
shown how to route transactions -based on type.
At Route 6, the TVs are routed based on their
ATtribute Value (ATV) so that transactions which
have TYPe equal to I are sent to Sink 3 and those
which have TYPe equal to 2 are sent to Sink 7.
Attributes have many other uses. In addition,
the user may assign and manipulate his own attri-
butes and interpret them uniquely.
RESOURCES
We have shown how transactions flow through
the INS network, being created, routed and de-
stroyed, and possibly delayed in activities. But
often the delay in an activity also represents
the use of some resource. For example, TVs being
inspected may involve the service of an inspec-
tor. In general, a transaction serviced at an
activity may require several resources.
Resources in INS can be identified individ-
ually (such as resource number 6) and collective-
ly (for example, resources of type 4). The user
may also assign the arrival time of the resource,
perhaps stochastically. The resources can be
utilized at activities. to respond to the demands
of transactions. For example, at the inspection
activity, we can model thé use of one inspector
as seen in Figure 5. We slightly modify the ac-
tivity node by placing a resource symbol above
it. In the resource - symbol, we have specified
that resource number 3 will service the ‘nspec-
tion activity for 10 minutes. The specification
of activity time changed from the activity node
to the resource symbol for several reasons.
It is not uncommon for resources to be sub-
stitutable at an activity. For example, two in-
spectors may be available to perform the ‘inspec-
tion. One inspector is new and takes 15 minutes
while the other is experienced and requires 10
minutes. If both inspectors are available, a
choice is needed since they are not identical.
Hence INS employs a means of selecting among al-
ternative resources, called the resource selec-
tion mode. Several selection modes are available
in INS, the simplest being PRIority. For the
sake of illustration, we will prefer the inexpe-
rienced inspector (resource number 4) to the ex-
perienced one (resource number 3). The revisions
at the inspection activity are shown in Figure
6. The two resource symbols in one column above
the inspection activity depict. the two alter-
native resources that can satisfy the requirement
for one resource. The preference for resource 4
(the top symbol) is noted by a priority number
(Tower numbers mean higher priority) and the PRI
in the activity stipulates that resource selec-
tion is based on PRIority.
MULTIPLE RESOURCES AT AN ACTIVITY
If more than one resource is required at an
activity, additional columns of resource symbols
above the activity would be specified. Each
column corresponds to the selection of a resource
and therefore, a resource selection mode for each
resource requirement is needed. If a testing
machine (resource 1) is needed at the inspection
activity, then the inspection activity would Took
like Figure 7. It is important to recognize that
INS considers multiple resource requirements at
activities simultaneously. If all resources are
not available, the activity cannot be started.
The capturing or holding of resources ‘individu-
ally can be specified, but in general, multiple
resource requirements will be considered together.
This unique feature of INS is particularly
useful in the study of systems where several re-
sources can constrain the processing of trans-
actions. In production systems, typically both
men and machines are required, and the simulation
will reveal which are more critical. This pro-
motes the realism of the simulation and expands
the modelers ability to represent very complex
decision processes, ones which, nevertheless, are
frequently encountered.
QUEUING AT ACTIVITIES
Since a resource may not be available every
time a TV arrives for inspection, some TVs must
wait for resources to become available. Repre-
senting queues in INS merely means the addition
of queue nodes to the network. The inspection
activity with a queue appears in Figure 8.
Queue 8 allows for queuing of TVs before Ac-
tivity 2 to await an available resource. By de-
fault, the queue ranking is FIFO, but this is
easily modified by specifying a ranking method
(either HIGh or LOW) and a. ranking attribute.
For example, if transactions of TYPe 1 are to be
prefered, Queue 8 would be specified as shown in
Figure 9.
- Figure 9 illustrates the multi-server single
queue ‘with a special queue discipline and server
preference. Muiti-queue, multi-server activities
can be as simply represented in INS by adding
additional queues at the activity. For example,
we may want to separate the queues for the TVs
being inspected for the first time and those
which have been adjusted (Figure 10). Other fea-
tures of the queuing capability in INS permit
resources at an activity to serve only selected
queues, permit reneging from the queue, and allow
transactions to capture resources in the queue.
These and other abilities are illustrated in the
User's Manual.
USING RANDOM DEVIATES
To change from constant activity times and
deterministic arrivals to stochastic variables in
INS involves referencing any of eleven built-in
distributions. User-specified probability den-
sity's can also be employed. Parameter sets are
created using these distributions and each para-
meter set is identified by a number.
.To obtain a sample from a distribution, the
modeler simply uses the negative parameter set
number. The PARAMETER statement defines the dis-
tribution but it is not an explicit part of the
network. For example, the following statement
defines an exponential whose mean is 12.
PARAMETER SET, 1, EXPONENTIAL, , 12
The parameter set number. is 1 and it might be
referenced on the Source node as in Figure. 11 so
that the interarrival of TVs from Source 1 will
follow an exponential of mean 12.
SYSTEMS WITH ACTIVE RESOURCES
A particularly important and unique feature
of INS is its ability to model systems where re-
sources are themselves mobile in serving trans-
actions at several activities within the net-
work. Suppose our more experienced inspector can
service transactions from either of two queues
(numbers 8 and 9) at the inspection activity and
furthermore can also perform adjustment (we will
assume that Queue 10 exists before the adjustment
Activity 4). Thus he is mobile, performing at
several activities, serving several queues. This
means that if a transaction enters any of these
queues and this inspector is idle, then he can be
used to begin their service. However, when this
inspector finishes either inspecting or adjust-
ing, there may be TVs in all three queues and he
needs to decide which to service. By default,. he
will simply inspect whichever TV has waited the
longest. But in many instances this may be in-
accurate, for the inspector's job description may
call for him to exercise choice. To model this
choice process, INS permits the user to create a
queue selector tree, composed of selector nodes
and queue node references, arranged to depict a
resource's decision-making process. For example,
Figure 12 might be the selector tree for the ex-
perienced inspector (resource number 4). By re-
ferencing Selector 20 as the primary selector
node for resource 4, the resource will employ
this tree in determining which transaction to
service. The selector node contains a selector
rule that describes the selection process. Sev-
eral rules are available in INS and the one in
Selector 20 is called PREferred order while the
one for Selector 21 is Longest Queue Wait. The
selector tree is read left to right and top to
bottom. Hence, the inspector will prefer (PRE)
Selector 21 to a transaction in Queue 10, but at
Selector 21, he will choose the transaction in
Queues 8 and 9 which has been waiting the
longest. If no transactions satisfy the algo-
rithm, the resource becomes idle.
Notice that by referencing queues rather than
activities in the selector tree, the modeling of
activities with multiple queues is enhanced.
Resources can then give priorties to queues which
they serve. Combining this feature with the
ability to rank transactions in the queue adds
345
MODELING WITH INS . . . Continued
great Flexibility to modeling complex queuing
processes.
MORE ADVANCED ‘INS CONCEPTS
Space does not permit the discussion of addi-
tional INS concepts and our purpose here is
simply to convey some fundamentals of the INS
modeling philosophy. However, three features
merit mentioning.
One is user-assigned attributes. INS offers
the convenient capability for the assignment of
several types of attributes. These attributes
can be used as variables for introducing coun-
ters, switches and totals into the network. Fur-
thermore, extensive arithmetic is available to
permit complex manipulation of attribute values.
Attributes can be used for indirect specification
within the network. For example, attributes can
refer to parameter sets and resources. Attri-
butes can be used to route transactions in the
network, assign resource requirements at an acti-
vity, rank transactions in a queue, and serve as
the basis. for the assignment of other attri-
butes. In addition, statistics can be collected
on the attributes. Attributes are changed within
the network as the situation dictates. |
A second important feature of INS is its
ability to model preemptive relationships among
activities using resources. This feature allows
the modeler to say that certain activities can
preempt less important activities for resources.
Thus, if adjustment of TVs is given absolute pre-
ference to inspection, the adjustment activity
can be specified to preempt resources from: in-
spection when needed. Such a modeling capability
further enhances the active nature of resources
within the network.
The third feature that merits mentioning is
the ability to model processing dependencies
among transactions. Such dependencies occur when
the continued processing of one transaction de-
pends upon another transaction. If TVs were
loaded into a truck rather than leaving at Sink
3, a truck-full of TVs would need to be collected
before the truck could leave. Hence a trans-
action may wait because of a synchronization ‚ ге-
quirement. This added complexity can be handled
with the addition of one statement, thus encour-
aging the modeling of a broader variety of dqueu-
ing circumstances in addition to those that arise
because of resource unavailability.
REMODELING THE TV INSPECTION AND
ADJUSTMENT EXAMPLE
To cap our discussion of modeling with INS,
we will reconsider the TV Inspection and Adjust-
ment example. In addition to our previous de-
scription of the problem, we add the following
embellishments. Interarrival times of the TVs
follow an exponential distribution, the inspec-
tion activity times are taken from Erlangs of
order 3, and the adjustment activity time is Tlog-
normal. Four resources are available; one is a
testing machine to be used at inspection; one is
an adjustor; and the other two are inspectors,
one is new and the other experienced. The new
346
inspector arrives on the job 30 minutes after the
start of the simulation. The experienced inspec-
tor can inspect and help out at the adjustment
station, but should give preferencé to ‘inspec-
tion. At the inspection activity, the newer in-
spector should be used first if both inspectors
are available, while at the adjustment activity
the adjustor should be prefered. Since resources
may not always be available at the activities,
queues are needed. Two queues exist at the in-
spection station, one for newly arriving TVs and
a second for TVs being reinspected. Those TVs
which have been in the inspect-adjust system the
longest should be given priority at the point of
reinspection. Also, with the addition of the
testing machine, the inspection activity is di-
vided into two parts. In the first part, the
testing machine is used to diagnose circuit mai-
functions. In the second part, whoever is in-
specting the TV continues to inspect without us-
ing the machine. The complete INS network for
the embellished problem is given in Figure 13.
The network illustrates how the embellishments
which have been described are easily integrated
into a new INS network. The ‘inspection is mod-
eled with two activity nodes, the second of which
has a resource selection mode that calls for the
resource CURrently in use in the inspection oper-
ation. This specification frees the testing ma-
chine for diagnosis on another TV while retaining
the inspector. Since we continue to utilize a
resource, no queue is needed as the TV wiil never
need to wait for the resource to become avail-
able.
Even though the problem has been greatly ex-
tended, the network retains its intuitive appeal
with the nodes and the branches providing a close
analogue to the physical system. The depiction
is concise, and sufficient alphabetic and numeric
information exists so that the simulation model
is fully documented from information contained in
the INS network.
TRANSLATING THE NETWORK TO INS STATEMENTS
The network of Figure 13 has been translatéd
into a complete set of INS statements in Figure
14. This is the information from which INS con-
structs and executes the simulation. The free
form nature of the statements when accompanied by
suitable comments (comments in INS can be placed
in any line after the $ symbol) render the input
very readable, even by persons unfamiliar with
INS. This further enhances model documentation,
and structures the model, so that even the mod-
eler will find it helpful as modifications and
enhancements of the model are made.
The input statements can be separated into
two parts. The first part serves as a preamble
to the description of the graphic network. The
preamble usually begins with a GENERAL statement
(statements are defined by their first word) that
contains the simulation project title and other
overall information. In this case, five runs of
the simulation are requested, each run lasting
480 minutes (or one day). Other preamble state-
ments define the structure of the simulation
model. For this problem, parameter sets and
resources are needed. Notice that the resource's
queue selector trees, as defined by the SELECTOR
statements, are also included in the preamble.
The second portion of the INS statements de-
scribes the network of transaction flow. Here
the correspondence of the graphic network to the
set of statements is virtually identical where
each symbol has its own statement. In general,
the rule is that each node corresponds to a
statement, therefore changes to nodes or even the
deletion or addition of nodes are isolated in one
statement. This means that working models are
easily changed.
Notice that the input involves no FORTRAN,
ALGOL, or other general purpose programming.
Furthermore, the modeler need not be concerned
about the mechanics of the simulation, i.e. event
handling, file manipulation, statistics collec-
tion, and random deviate generation. This lack
of mechanical involvement encourages the modeler
to focus his creative attention on the model
building by placing very few requirements on him
to translate and program the simulation model
once the problem is formulated through the INS
network. He simply works at the problem defini-
tion stage, leaving to INS the details of the
simulation. These features are especially impor-
tant to the naive or infrequent simulation mod-
eler.
INS OUTPUT
Without any special instructions, INS auto-
matically provides output on the construction and
execution of the simulation model and output
which statistically summarizes the behavior of
the model. The output will include a listing of
the input statements and an echo of the simula-
tion variables both user- and system-supplied. If
any errors are found by INS in construction or
execution, INS prints an appropriate error mes-
sage. ‘In general, INS attempts to execute the
model whenever possible, supplying variables by
default, and warning the user of suspicious cir-
cumstances. The echo, therefore, totally de-
fines the simulation problem as INS understands
it and is particularly useful to the modeler for
debugging.
The summary output is the statistical de-
scription of the model results. Four points of
view of the simulation are given by default.
First, overall statistics about the run are
given, such as the number of runs included in the
report and the time during which statistics are
collected.
Second, statistics relating to nodes are pro-
vided. Node counts indicate the number of times
each node has been encountered by transactions.
Node statistics are generally related to occupan-
cy of each node by transactions. For example,
queue nodes and activity nodes provide statisti-
cal data (such as mean, standard deviation, mini-
mum, and maximum) on the average number and aver-
age time transactions spent in each queue and in
each activity. Queue time statistics are sub-
divided-to include and exclude those transactions
which do not wait but pass through the queue in
zero time. Sink node statistics provide total
time in the system for transactions which exit
the system at the particular sink node and this
time is divided into the total time spent in
queues and activities.
The third view reflected by the output is the
transaction statistics. These statistics indi-
cate the number and time of transactions in the
network and also state the number of transactions
created.
The fourth portion of the summary output
gives resource utilization statistics. For each
resource, resource utilization by resource state
is given. Resource states include the time the
resource is not in the network, the time it is
idle, and the time it is busy at an activity.
Results include the percentage of their utiliza-
tion in each of thé states as well as the average
time they occupied that state.
Thus, the INS modeler will be able from the
INS output to extract several perspectives on the
results. The diagnostics will aid in verifying
the model while the summary should reinforce the
focus on modeling by generating data that should
validate the modeler's view of the system's oper-
ation. This insight will hopefully lead to prob-
lem solutions and the testing of alternatives.
ENHANCING THE INS OUTPUT
INS has a number of provisions for expanding
or even constraining the default output so more
or less detailed information can be obtained.
The execution of the simulation can be monitored
in several ways. A system state trace can be
activated to produce a motion picture-like pre-
sentation of the motion of transactions as they
flow from node to node and the behavior of re-
sources as they change their states. The trace
can be constrained to display only certain re-
sources or only particular transactions. Static
pictures of the network during execution are also
available, triggered either by the flow of trans-
actions in the network or by time. These pic-
tures may range from portions of the summary re-
port to information on the current status of
transactions and resources.
Summary report statistics can be enhanced in
several ways. Additional statistical nodes can
be added to the network to compute transit times
between points in the network or interarrival
times at some point in the network. Resource
utilization statistics can be extended to yield
data by individual activity and queue nodes.
Furthermore, node statistics and resource utili-
zation can be enhanced by the display of a his-.
togram of the statistical distribution or a table
of statistics broken down by transaction type.
The table and histogram in Figure 15 were pro-
duced by the TABLE statemént in Figure 14. It
required only one statement to produce a table
combining the statistics on the number of TVs
waiting in Queues 8 and 9. Tables and histograms
for resources, attributes, activities, sinks, and
other nodes can be equally simply created.
The extended output thus gives the modeler a
broad, but convenient range of outputs. He can
examine the detailed processing to debug or to
otherwise verify his model and he can create so-
phisticated statistics to gain additional insight
into the system's behavior.
APPLICATION AND AVAILABILITY OF INS
INS has been applied in several environments.
347
MODELING WITH INS . . . Continued
The thirty examples in the User's Manual (1)
illustrate the range of applications for INS
including queuing, inventory, computer, manufac-
turing, and service systems.
INS has been successfully used in simulation
courses for industrial engineering students.
Graduate students utilized it as a new network-
based language while it has been used as the
principle simulation methodology in an under-
graduate course in systems analysis. The lan-
guage has also been used as the basis of a senior
elective in medical school. Medical students
have used INS to create models for clinical in-
vestigation. These instances of its use in a
teaching environment have demonstrated that, not
only can the language be quickly learned and ap-
plied by persons with minimal technical exper-
tise, but it can also be valuable to those with
deeper appreciation of simulation.
INS is written in ANSI FORTRAN making it
portable and has been implemented on a number of
machines. The User's Manual and INS are avail-
able from the Regenstrief Institute, 1001 West
Tenth Street, Indianapolis, Indiana, ' 46202. Be-
cause the language has been developed in a
non-profit institute, INS is available at the
current price of $400. User's Manuals are avail-
able separately. INS is actively being main-
tained by the Institute and is the subject of
continuing improvement. both to increase the ef-
ficiency of its execution and to extend its ap-
plications.
CONCLUSION
Simulation modeling involves constructing a
model by abstracting data and processes from the
system being studied and implementing the model
on a computer. INS facilitates model —con-
struction by providing a convenient ¡set of sym-
bols representing processes such as queues and
activities which enable the modeler to formulate
the model quickly yet precisely. ‘Ву a simple
translation from the symbols to INS statements,
the model is implemented without effort. INS
automatically handles all details of the simula-
tion including file handling, random deviate gen-
eration, and statistics collection and reporting,
thus relieving the modeler of these tasks. This
dual approach in a high level language allows the
modeler to focus attention on the important tasks
of model creation and validation. INS is easy to
learn, to use, and to modify as has been shown by
the example we have developed. Since INS is vi-
sually appealing and easily documented, it is an
excellent vehicle for communication between the
modeler and the client, encouraging cooperative
participation. Last but not least, INS is writ-
ten in ANSI standard FORTRAN and is available for
a variety of computers.
REFERENCES
1. Roberts, S.D., and J.S. Schier, "INS User's
Manual for Version 3", Regenstrief Institute
for Health Care, Indianapolis, Indiana, 1978.
348
Arriving
TVs
Figure 1: Flow Diagram of TV Inspection and Adjustment Problem
NN Y
Lal] (Cal) | Inspection > Good
ese ase же. TVs
Faulty
TVs
М Nr
Га] Adjustment
Adjusted TVs
Figure 2: Initial INS Network
TVs
Arrive
12
inspection
PRO
10.
Adjustment
4
SOURCE, 1, , 12
BRANCH TO, 2
ACTIVITY, 2, PRO, (8)10.
BRANCH TO, 3, .85
BRANCH TO, 4, .15
ACTIVITY, 4, (8)30.
BRANCH TO, 2
SINK, 3
Figure 3: INS Statements
$ Arrival of TVs
$ Go to Inspect
$ Inspect TVs
$ 85% pass
$ 15% fail
$ Adjust TVs
$ Reinspect TVs
$ Good TVs leave
349
MODELING WITH INS . . . Continued
ny, da “
Figure 4: Network Modification for TV Types
inspection
4
Adjustment
Figure 5: Using an Inspector
3
10
Inspection
2 | |
[ra
Figure 8: Queuing for Inspection
Y
e
Figure 6: Alternative Inspectors
4
15,
71
\3
10,
2
Inspection
2
| PRI
Figure 7: Tw
o Resource Requirements
15
4
1
3 1
10,
2
Inspection
2 |
| PRI,PRI
3
1
1
4 1
119,
Inspection
2 | |
| PRI,PRI
Figure 9: Queuing in Order of TYPe
3
15,
1
4 1
10
2
7 Inspection
2 | |
Za | PAILPRI
350
Arriving
TVs
Adjusted
TVs
Figure 10:
4 1
10.
2
SS
v
e
Multiple Queues
3
15.
1
Inspection
|
| PRt!,PRI
Figure 12:
Queue Selector Tree
Figure 11: Stochastic Interarrivai Times
Arriving
TVs
|
1
ï
/
Figure 13: INS Network for TV Inspect and Adjust
3 3
-2 -5
TV Arrival }
| -1 4 a
-3 | -5
2
1
TIM | 480 Insbact , Inspect
with Machine without Machine
(.85) ol 3
11
| PRO
CUR
(.15)
Adjustment
|
E
RESOURCES AVAILABLE
Primary Arrival
Description Number Selector Node Time
Testing Machine 1 -- 0.
Adjustor 2 -- 0.
Inspector 3 - 30.
inspector 4 20 0
Inspector
QUEUE SELECTOR TREES
20
PRE
A.
2.
-3.
-D,
PARAMETER SETS
Exponential Interarrival
time of TVs.
Erlang inspection time by
new inspector.
Erlang inspection time by
experienced inspector.
Lognormal adjustment
time.
Normal inspection time
without tester.
351
an
PRE 6 A = pea ay
4 !
1
MODELING WITH INS . . . Continued
Figure 14
t
352
|
AMA A. |
PARAMETER, 2, ERLANG ‚‚ 45, 5, 25,
TV INSPECTION AND ADJUSTMENT EXAMPLE
1 GENERAL, INSPECT AND ADJUST TVS,1:(7)TIME TO STOP 2,482, 5 § RUN FOR 5 DAYS
junecantesPARAMETER SETS UNE
SET DISTRIBUTION -MEAN- MIN MAX
PARAMETER, 1, EXPONENTIAL », 12: TV INTERARRIVAL TIME
$ INSPECTION TIME W/ MACHINE - NEW
PARAMETER, 3, ERLANG ,, 12, 6, ;
PARAMETER, 4, LOGNORMAL »» 38, 15, 50;
„PARAMETER, 5, NORMAL a. 5, 2, 18,
TABLE: 1. NUMBER OF TVS WAITING, a 174 1, 1
ADJUSTMENT TIME
: INSPECT WITHOUT MACHINE
nN
œ
DO IC
$ .
SotsantrasRESQURGES + à 8 nu 4 0x
$
$ NUMBER TYPE SELECTOR ARRIVAL
RESOURCE, 1. 1 $ TESTING MACHINE
RESOURCE; ‘ 2, 2 $ ADJUSTOR
RESOURCE. 3: 3. ‚ 38, $ NEW INSPECTOR
RESOURCE, 4, 3e 20 $ OLD INSPECTOR
$
SELECTOR TREE FOR OLD INSPECTOR, 28, PREFERRED QROER, 21, 19
SELECTOR, 21, LaW, B, 9 |
$ |
Gutannttnsotantaadatatsatnstutaues THE NETWORK +64 0005 dde 06000 00 06 00 50 M Le ae 55 9 5 0 ma dc masia
$
SOURCE »1, 20017 va TIME TO STOP =,48% $ ARRIVING TV SETS
BRANCH TO, 8
$
QUEUE 8,1 QUEUE OF ARRIVING Tvs
QUEUE 9,1 QUEUE oF ADJUSTER Tvs
INSPECTION OF TVS WITH MACHINE
BY AN INSPECTOR
USING THE MACHINE
ACTIVITY,2,{8)PRIORITY;,PRIORITY
GROUP, 1, Sri 32a 4,,2,-3
GROUP,2, 1
BRANCH TO, 11
$ -
ACTIVITY,11,PRO, (S)CURRENT
HAAN
$ INSPECTION OF TvS WITHOUT MACKINE
GROUP, 1, 31,,-5, 4,,,-5 $ BY THE SAME INSPECTOR
BRANCH TO, 3, .85 $ 85 PERCENT FASS
; BRANCH TO, 16, ‚15 $ 15 PERCENT FAIL
QUEUE, 1D $ QUEUE FOR ADJUSTMENT
ACTIVITY,4,(8)PRIORITY $ TV ADJUSTMENT °
GROUP,L, 2,,1,~4, 4,,2,-4 $ BY THE ADJUSTOR OR OLD INSPECTOR”
N BRANCH To; 9 $ RETURN FOR INSPECTION
SINK,3 $ GOOD TVS LEAVE
FINISH
$
$
$ INSPECTION TIME и/ MACHINE - Lo
3
Y
- Figure 15: Output From TV Inspection and Adjustment Example
*# TABLE NUMBER 1 ws
NUMBER OF TVS. WAITING
STANDARD TIME OF SUM OF THE
ME AN DEVIATION MINIMUM MAXIMUM OBSERVATION SUN SQUARES
3.62416 4,1779 GATOGAE +2 17,2262 2405, 40 8697,99 70265 ,2
#* HISTOGRAM NUMBER 1 +4
NUMBER IF TVS WAITING
"UPPER LIMIT INCLUSIVE
RELATIVE CUMULATIVE =““CELL LIN[T==.
FREQUENCY FREQUENCY LOWER UPPER р, й 0.2 0,4 Bab 0.8 1,0
+ + + + + + + + + + û
7.309 2,309 “INFINITY 0, OOBLEE DE MA LL ; ; +
2.107 7,416 Й, ЙО ООЙЕ +0 1.006% HNOS Y
0,799 ‘84515 1.0809 2.2502 +4 8042 *
6,091 0,6935 ‚2.0009 3,0002 +=. #
8,8077 7,682 3.2008 4.0008 +44 #0 »
6.951 0,733 4,2949 5,0500 eun +
2,049 9,782 5,0040 6, LUPE +## С *
0,050 7,832 6,578 7,0005 +44 4 С. %
й. йЗ9 0,871 7,0008 8,2207 +. с +
¢.014 6,885 8,9080 9,0900 en Co»
0,23? в, 917 9.49000 18,900 +. 6 $
2,025 9,542 19,079 11,00% ++ C' # |
0,076 0,958 411228 12,500 + C a
2.815 0,964 12.282 13,597 +“ Ca
@, 217 7 2,981 13.020 14,000 | + 6 - Ce.
0.299 2,998 14,000 15,000 + ‚6 |
: 8.910 1.008 15.000 16.006 + ¢
2,206 1.200 16,000 17,608 + С
2.90 1.620 171008 *INFINITY + ` Cr
+ + % » Û
, 0 0,8 a и
ESE
a
su
MO)
">
/
डाउनलोड
फक्त एक मैत्रीपूर्ण स्मरणपत्र. आपण येथे दस्तऐवज पाहू शकता. परंतु सर्वात महत्त्वाचे म्हणजे, आमच्या एआयने ते आधीच वाचले आहे. हे गुंतागुंतीच्या गोष्टी सोप्या भाषेत समजावून सांगू शकते, आपल्या प्रश्नांची उत्तरे कोणत्याही भाषेत देऊ शकते आणि अगदी लांब किंवा सर्वात गुंतागुंतीच्या कागदपत्रांमध्ये द्रुतपणे नेव्हिगेट करण्यात मदत करू शकते.
जाहिरात