Title: Method of determining level of effect of system entity on system performance, by use of active time of same entity
Abstract: A method of determining the level of the effect of each system entity on the overall system performance, wherein the status of each entity at each one of discrete points of time is determined as the corresponding one of an active and an inactive status; the length of the active period in which each entity is situated without interruption in the active status is determined as an active time; and, on the basis of the relationship in magnitude between the system entities with respect to the determined active time, the level at which each entity affects the performance of the system is determined.
Patent Number: 6,978,222 Issued on 12/20/2005 to Roser, et al.
| Inventors:
|
Roser; Christoph Hermann (Aichi-gun, JP);
Nakano; Masaru (Aichi-gun, JP);
Tanaka; Minoru (Aichi-gun, JP)
|
| Assignee:
|
Kabushiki Kaisha Toyota Chuo Kenkyusho (Aichi-gun, JP)
|
| Appl. No.:
|
103809 |
| Filed:
|
March 25, 2002 |
Foreign Application Priority Data
| Mar 29, 2001[JP] | 2001-097640 |
| Oct 10, 2001[JP] | 2001-313133 |
| Current U.S. Class: |
702/182; 702/186; 703/2; 700/99; 700/108 |
| Intern'l Class: |
G06F 011/30 |
| Field of Search: |
702/186,182
703/2
700/99,108
|
References Cited [Referenced By]
U.S. Patent Documents
| 5229948 | Jul., 1993 | Wei et al.
| |
| 5351202 | Sep., 1994 | Kurtzberg et al.
| |
| 5446671 | Aug., 1995 | Weaver et al.
| |
| 5479361 | Dec., 1995 | Kurtzberg et al.
| |
| 5546329 | Aug., 1996 | Kurtzberg et al.
| |
| 5636144 | Jun., 1997 | Kurtzberg et al.
| |
| 5838565 | Nov., 1998 | Hsieh et al.
| |
| 5880960 | Mar., 1999 | Lin et al.
| |
| 5946661 | Aug., 1999 | Rothschild et al.
| |
| 5966694 | Oct., 1999 | Rothschild et al.
| |
| 6259959 | Jul., 2001 | Martin.
| |
| 6473721 | Oct., 2002 | Chacon et al.
| |
| 6564113 | May., 2003 | Barto et al.
| |
Primary Examiner: Barlow; John
Assistant Examiner: Dougherty; A.
Attorney, Agent or Firm: Oblon, Spivak, McClelland, Maier & Neustadt, P.C.
Parent Case Text
CROSS-REFERENCE TO RELATED APPLICATIONS
Not Applicable.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a technology of determining the level of the
effect of each one of a plurality of entities constructing a system on the performance
thereof, wherein the status of each entity changes with time.
2. Description of the Related Art
In many systems, it is of interest to control the throughput of each system,
usually
in order to maximize the throughput. For example, in a manufacturing system, it
may be of interest to maximize the number of products or parts produced within
a certain time.
The throughput of a real system, however, is always finite. While many factors
in a system affect the throughput, it is usually only a few entities (e.g., processing
entities such as machines, transporting devices, processors of a computer, etc.)
in a system limiting the throughput.
These limiting entities in a system are commonly called bottlenecks or constraints.
These bottlenecks or constraints limit the overall flow of a discrete event system
constructed with entities. Each entity constructing a system is a material, a human,
or an abstract element, as for example as a machine, a worker, an order, information, etc.
Subsequently, in order to change the system throughput, it is necessary
to change the throughput of the one or more bottlenecks. Adjusting non-bottleneck
entities will usually have little or no effect on the throughput.
Consequently, it is important to determine the effect-level at which
each entity constructing a system affects the performance or throughput of the
system, and to detect, on the basis of the thus determined effect-level, one or
more entities of the system as one or more bottlenecks.
Currently, a number of different methods for detecting the bottleneck
are in use. One commonly used method is to determine the busiest entity, i.e.,
the entity that has the largest percentage working time, and eventually has the
smallest percentage idle time.
In this conventional method, as illustrated with a flow chart in FIG. 27, step
S101 is first implemented to collect data required for determining the bottleneck
of a system, and step S102 is then implemented to measure the percentage
working time of each one of entities constructing the system. Step S103
is thereafter implemented to order the entities by the measured percentage working
time, and step S104 is then implemented to determine that one or more entities
with the largest percentage working time is one or more bottlenecks of the system.
BRIEF SUMMARY OF THE INVENTION
Our detailed analysis of the meaning of the bottleneck of a system has revealed
that the bottleneck is one of the entities constructing the system, which hinders
the potential throughput of another entity, and which is kept active for the longest
time because the present entity itself does not have a sufficient processing capacity.
Contrary to the above findings, the conventional method described above
does not permit a direct consideration of an active time in which each entity is
kept active without interruption, and unavoidably permit an indirect consideration
of the above active time in the form of a percentage working time, as calculated
by dividing the sum of working times of each entity by the overall operation time
of the system.
For this reason, the above conventional method would provide a large percentage
working time for an entity, even in the case where the entity fails to be deemed
as such an entity that has been kept active without interruption for a long time,
such as the case where the entity shows a frequently iterated shifting between
an active and an inactive status alternatively.
As a result, the above conventional method has difficulties in clearly distinguishing
between the entity truly being the bottleneck, and the entity which is not truly
the bottleneck and which shows a frequently iterated shifting between an active
and an inactive status alternatively, because this conventional method employs
the percentage working time that fails to sensitively reflect the strength level
at which each entity functions is the bottleneck.
Thus, using the above conventional method it is essentially difficult to detect
an entity truly functioning as the bottleneck.
Another conventional method to detect the bottleneck is the analysis of the
queue in front of a processing entity (e.g., a machine), wherein the queue is a
line of processed objects waiting for them to be processed by the processing entity.
A first variant of the another conventional method is implemented, such that
the
queue length, i.e., the average of the number of processed objects waiting is calculated
for each one of entities constructing a system, and subsequently the maximum of
the thus calculated averages for these entities is detected as the longest average
queue length.
In the first variant described above, the entity with the longest average queue
length is deemed to be the bottleneck of the system. This, however, is not always
true. If, for example, an entity is processing parts, i.e., processed objects at
a faster rate than other entities, a longer queue (that is, a larger number of
processed objects waiting) does not necessarily mean that this entity is the bottleneck.
A second variant of the another conventional method, which is to detect the bottleneck
by means of the analysis of the queue, as described before, is implemented, such
that the entity with the maximum waiting time, i.e., the entity where the processed
objects have to wait for the longest time before they are processed, is detected
as the bottleneck. However, this second variant also suffers problems.
A first one of the problems results from the structure of the system. In many
cases,
a buffer for storing processed objects in front of an entity is limited, setting
the maximum queue length to the buffer size and subsequently limiting the waiting
time. This, of course, invalidates the measures for the bottleneck detection.
Also, in many cases, one buffer stores part for more than one processing entity,
or more than one buffer stores part for only one processing entity. In general,
if one buffer is not uniquely assigned to only one entity, it is extremely difficult
to determine which waiting time applies to which entity. For this reason, subsequent
detection of the bottleneck is difficult to justify.
Another problem that the above second variant suffers is that usually the
input of the system, i.e., the supply, does not have a queue. Yet, it is possible
that the supply is actually the bottleneck of the system, where the system could
process more objects if there would be more objects delivered. However, if the
bottleneck is detected by analyzing the queues, a supply bottleneck will not be detected.
Furthermore, the supply may exceed the capacity of the system. In this
case, the queue length or waiting time of queues may increase towards infinity.
Therefore, subsequent detection of the bottleneck is difficult to justify.
Still another conventional method for bottleneck detection is a theoretical
analysis of the system. Possible examples are a logical structure analysis and
a queuing theory analysis. However, the applications of these approaches are usually
complex. Most examples are limited to simple systems analyzed for academic research
purposes. The practical use is limited due to the complexity of those methods,
even for simple systems. In most cases, it is not economical to analyze real life
systems using theoretical methods.
As will be readily understood from the above explanation, any one of those conventional
methods described above is insufficient to accurately detect the bottleneck, even
in a steady state system where the same one of the entities constructing the system
is kept as the bottleneck irrespective of an elapse of time.
Further, these conventional methods are also insufficient to accurately
detect the bottleneck in a non-steady state system where the bottleneck shifts
with time between entities constructing a system. The reason will be described below.
The above non-steady state system is a system where its parameter changes over
time. Here, the parameter include for example in the case of the system being a
manufacturing line, a time required for manufacturing one product or part, and/or
includes the kind of products or parts to be manufactured. A change in the parameter
may accompany a shifting of the bottleneck from an entity to another entity in
the system.
There exists a conventional method for detecting the bottleneck in such a non-steady
state system. This method measures the aforementioned queue length in front of
a processing entity, or alternatively measures the waiting time in front of the
processing entity. The processing entity with the longest queue is considered to
be the bottleneck. However, this method has many flaws.
First of all, many entities in a system do not have a queue, and therefore
a queue length or waiting time cannot be measured. Even if there is a queue, the
queue capacity is often limited and the resulting queue lengths or waiting times
do not represent the real bottleneck.
Also, queue lengths fluctuate significantly over time, and the bottleneck detected
according to the queue length may change very frequently for only very short periods
of time. For this reason, this leads to difficulties in tracking the shifting bottleneck
that qualifies as one of the entities of a system showing a shifting of the bottleneck,
and accordingly this also causes occasionally incorrect results.
There is a system of the type where some processed objects, such as materials,
parts, products, are bundled to form a batch (i.e., a group or a bundle) for thereby
altogether processing these processed objects, and where a batch size meaning the
number of processed objects belonging to each batch fluctuates over time. For example,
a manufacturing system receives parts in large batch sizes (i.e., lot sizes), but
then feeds parts on one by one through the system. In such a manufacturing system,
the entity with a larger batch size is likely to have a temporary longer queue
length even if the entity is not the bottleneck.
Overall, the measurement of the queue length or waiting time can be used
only with limitations, and the resulting bottleneck may not be the true bottleneck.
It is therefore an object of the present invention to permit an accurate determination
of the effect-level at which each one of entities constructing a system affects
the performance of the system, wherein the status of each entity changes with time.
The object may be achieved according to any one of the following modes of the
present invention. Each of these modes of the present invention is numbered, and
depends from other mode or modes, where appropriate. This explanation about the
present invention is for better understanding of some instances of a plurality
of technological features and a plurality of combinations thereof disclosed in
this specification, and does not mean that the plurality of technological features
and the plurality of combinations in this specification are interpreted not to
include ones other than the following modes of the present invention:
(1) A method of determining a level at which each one of a plurality of entities
together constituting a system in which each one of the plurality of entities changes
in status with time, affects a performance of the system, comprising:
a collecting step of collecting data representative of change in status of the
each entity:
a qualifying step of qualifying, on the basis of the collected data, the status
of the each entity at each one of discrete points of time, as a corresponding one
of an active and an inactive status;
an active-time determining step of determining as an active time a length of
an
active period in which the each entity is situated without interruption in the
active status; and
an effect-level determining step of determining, on the basis of a relationship
in magnitude between the plurality of entities with respect to the determined active
time, the level at which the each entity affects the performance of the system.
The method according to this mode (1) is implemented, such that the level at
which each entity affects the performance of the system is determined on the basis
of the relationship in magnitude between the plurality of entities with respect
to the active time during which each entity is situated without interruption in
the active status.
In addition, as will be evident from the aforementioned definition of the bottleneck,
employment of the length of the uninterrupted active period of each entity would
make it easier to accurately determine the effect-level of each entity on the performance
of the system.
On the other hand, as will be below described in detail, the above relationship
in magnitude between the plurality of entities with respect to the active time
represents more clearly the order in which the effect-levels of the plurality of
entities on the performance of the system are ranked, than the relationship in
magnitude between the plurality of entities with respect to the percentage working
time which was referred to in the aforementioned conventional methods.
Consequently, the method according to this mode (1) would make it easier
to accurately determine the effect-level of each entity on the performance of the system.
Additionally, the method according to this mode (1) would make it possible
to avoid a strong dependency of the above relationship in magnitude between the
plurality of entities with respect to the active time, upon random variation of
data used for determining the effect-level of each entity on the performance of
the system.
As a result, the method according to this mode (1), making it inessential to
collect
an increased amount of data required for mitigating the effects of the random variation
of data required, would facilitate an accelerated determination of the effect-level
of each entity on the performance of the system, with a reduced amount of data required.
The "collecting step" in this mode (1) may be, for example, of the type to collect
from the system in operation data required, or the type to collect data required
as a result of simulating the system with a computer.
The "effect-level of each entity on the performance of the system" in this mode
(1) may be interpreted to mean, for example, the level at which each entity impedes
or limits an improvement in performance of the system.
The method according to this mode (1) may be practiced in such a manner that
all or part of the steps included in the method are implemented with a computer,
or may be practiced in such a manner that all the steps are implemented by a worker
without using a computer.
The "relationship in magnitude" in this mode (1) may be defined as one obtained
by locally observing the status of each entity in association with a given point
in time, or may be defined as one obtained by globally observing the statuses of
each entity in association with a given period of time.
(2) The method according to the above mode (1), wherein the effect-level determining
step is implemented to compare the plurality of entities with respect to the active
time determined in the active-time determining step, and to determine, on the basis
of results from the comparison, the effect-level of the each entity on the performance
of the system.
(3) The method according to the above mode (1) or (2), wherein the system is
a discrete event system.
(4) The method according to any one of the above modes (1) to (3), wherein the
system is utilized to receive a plurality of processed objects, to manufacture
a plurality of products by processing the received plurality of processed objects,
and to deliver the manufactured plurality of products, and
the plurality of entities comprise at least one of the following entities: a
processing entity for processing each one of the plurality of processed objects;
a transport entity for transporting the each processed object; a service entity
for servicing another entity; a maintenance entity for maintaining another entity;
and a storage entity for storing the each processed object.
(5) The method according to the above mode (4), wherein the processing entity
is constructed as a machine.
(6) The method according to the above mode (5), wherein the machine is utilized
for processing the each processed object.
(7) The method according to any one of the above modes (4) to (6), wherein the
performance is defined with a number of products manufactured in the system within
a given time.
(8) The method according to any one of the above modes (1) to (7), wherein the
effect-level determining step comprises a bottleneck determining step of determining
at least one of the plurality of entities as at least one bottleneck which affects
the performance of the system at a higher level than other entities.
The method according to this mode (8) would determine the effect-level of each
entity on the performance of the system, according to the standard for determining
whether or not each entity functions as the bottleneck of the system.
(9) The method according to the above mode (8), wherein the bottleneck determining
step comprises a step of determining one of the plurality of entities that has
the longest active time among the plurality of entities, as one bottleneck affecting
the performance of the system more strongly than other entities.
(10) The method according to the above mode (8), wherein the bottleneck determining
step comprises a step of determining at least one of the plurality of entities
that has the active time in the vicinity of the longest active time among the plurality
of entities, as at least one bottleneck affecting the performance of the system
more strongly than other entities.
A system has not always only one bottleneck. In addition, at least one of a plurality
of entities constituting a system that has the active time in the vicinity of the
longest active time among the plurality of entities is all likely to qualify as
at least one bottleneck of the system.
Based on the above findings, the method according to this mode (10) is implemented,
such that at least one of the plurality of entities that has the active time in
the vicinity of the longest active time among the plurality of entities are determined
as at least one bottleneck affecting the performance of the system more strongly
than other entities.
Consequently, the method according to this mode (10) would facilitate
detection of all the at least one of the plurality of entities that is likely to
qualify as at least one real bottleneck of the system.
(11) The method according to any one of the above modes (8) to (10), wherein
the system is of a steady type in which the same at least one of the plurality
of entities is kept qualifying as the at least one bottleneck of the system as
time elapses.
The method according to this mode (11) would make it possible to determine at
least one bottlenecks in a steady state system where the same at least one of the
plurality of entities is kept qualifying as the at least one bottleneck of the
system as time elapses.
(12) The method according to any one of the above modes (8) to (10), wherein
the system is of a type in which, while the system is in a steady state in which
the same at least one of the plurality of entities is kept qualifying as the at
least one bottleneck as time elapses, as long as no disturbance is applied to the
system, the system is brought into a non steady state and is then temporally situated
in the non steady state, in which the at least one bottleneck of the system shifts
from a part of the plurality of entities to another part of the plurality of entities
as time elapses, after a disturbance is applied into the system, and
the bottleneck determining step is implemented to determine at least one of the
plurality of entities as the at least one bottleneck while the system is in the
steady state.
The method according to this mode (12) would make it possible to determine at
least one bottleneck, while the system is in a steady state, where the system is
of the type in which the system shifts from a steady state into a non steady state
and is then temporally situated in the non steady state, in response to a disturbance
applied to the system, and in which at least one bottleneck of the system shifts
from a part of the plurality of entities to another part of the plurality of entities
as time elapses. This will be below described in more detail.
In the case of the aforementioned conventional method, large sets of data are
needed to determine the bottleneck with reasonable accuracy. Unfortunately, large
sets of data are often time-consuming to obtain. If it takes a long time to determine
the bottleneck, implementation of possible bottleneck improvements may have to
wait until the bottleneck is detected. Thus, a less than optimal system has to
be run for a long time, wasting resources in the process of the system.
Even a steady state system in which the bottleneck basically does not change
from an entity to another entity as time elapses may be brought into a non steady
state in which the bottleneck changes from an entity to another entity as time
elapses, due to a disturbance applied into the system. For example, in a manufacturing
line, new products are likely to be added to the schedule, and old products are
likely to be removed from the schedule, possibly changing the bottleneck. In both
of these situations, the bottleneck of the manufacturing line may change over time.
In such a system that may bear the above non steady state, large sets of data
collected from the system irrespective of whether it is in the steady state or
the non steady state are inappropriate to be used for accurately detecting the
bottleneck. The reason is that the collected large sets of data contain different
kinds of sets of data representative of different bottlenecks. This would create
the concept of collecting data from the system only in the steady state, for thereby
detecting the bottleneck.
However, with this concept, the amount of data available for detecting the
bottleneck is likely to be less than data that could be obtained from a steady
state system without non steady state.
On the other hand, the aforementioned conventional methods require large sets
of data to mitigate variation in the percentage working time for an ensured accuracy
in detecting the bottleneck.
For the above reasons, the conventional methods, requiring large sets of data,
may either give no valid results at all, or the results may be obsolete by the
time they become valid.
Alternately, the method according to this mode (12) would not greatly
require a careful consideration of mitigation of variation in the active time,
while the above conventional methods would greatly require a careful consideration
of mitigation of variation in the percentage working time. Hence, the method according
to this mode (12) would make it easier to accurately determine the bottleneck,
even where available sets of data are smaller than ones of data required in the
above conventional methods.
In addition, such a system that is brought into a temporary non steady state
upon
application of a disturbance into the system requires separated determinations
of the bottleneck between in a steady state and in a non steady state. Data that
could be obtained in the steady state is less than data that could be obtained
in the non steady state. However, as described before, the method according to
this mode (12) would make it easier to accurately determine the bottleneck even
where available data is less in this method than data available in the aforementioned
conventional methods.
Consequently, the method according to this mode (12) would facilitate
an accurate determination of the bottleneck of such a system that is brought into
a temporary non steady state upon application of a disturbance into the system.
It is added that such a system that is brought into a temporary non steady state
upon application of a disturbance into the system, if the fact that the system
possibly bears a steady state is focused on, may be categorized as a steady state
system, and on the other hand, if the fact that the system possibly bears a non
steady state is focused on, may be categorized as a non steady state system.
(13) The method according to any one of the above modes (8) to (12), wherein
the active-time determining step comprises a representative-active-time determining
step of determining, for the each entity, as a representative active time, a representative
value of a plurality of active times of a plurality of active periods which are
discrete in time, and
the bottleneck determining step is implemented to determine, on the basis of
a relationship in magnitude between the plurality of entities with respect to the
determined representative active time, at least one of the plurality of entities
which functions as the at least one bottleneck.
The method according to this mode (13) would permit the bottleneck to be determined
by taking account of a representative value of the plurality of active times determined
for a plurality of discrete active periods as a representative active time.
Consequently, the method according to this mode (13), in the case where
the active time of each entity is varied between the plurality of active periods
of the same entity, would prevent the variation in active time from affecting the
level of accuracy at which the bottleneck is determined. The reason is that the
representative value of the plurality of active times functions to absorb active
time variation between different active periods of the same entity.
It follows that the method according to this mode (13) would make it easier to
accurately determine the bottleneck irrespective of variation in active time.
(14) The method according to the above mode (13), wherein the representative
active time comprises at least one of an arithmetic mean; a harmonic mean; and
a median, of the plurality of active times.
(15) The method according to the above mode (13) or (14), wherein the effect-level
determining step comprises an accuracy determining step of determining, for the
each entity, an accuracy of the determined representative active time, and the
bottleneck determining step is implemented to determine at least one of the plurality
of entities which functions as the at least one bottleneck, on the basis of the
determined accuracy and the determined representative active time.
Where a representative active time is determined for an entity, the accuracy
of the determined representative active time for the same entity is not always
equal to that for another entity.
In addition, where all the ones of the plurality of entities which are to be
compared
with each other with respect to the representative active time are adequately high
in accuracy of the representative active time, it is effective to determine the
bottleneck by focusing only on the representative active time.
However, where all the above ones are not adequately high in accuracy of
the representative active time, if the bottleneck is determined by focusing only
on the representative active time, it is possible to fail to detect the real bottleneck
with certainty.
In view of the above findings, the method according to this mode (14) is implemented
to determine the bottleneck of the system on the basis of the representative active
time determined per entity and its accuracy.
Thus, the method according to this mode (14) would facilitate an ensured determination
of the real bottleneck, by considering not only the length of the representative
active time but also its accuracy.
(16) The method according to the above mode (15), wherein the accuracy comprises
at least one of a confidence interval of the representative active time; and a
standard deviation of the plurality of active times.
(17) The method according to any one of the above modes (1) to (16), wherein
the effect-level determining step comprises a bottleneck determining step of determining
at least one of the plurality of entities as at least one bottleneck which affects
the performance of the system at a higher level than other entities at a given
point in time.
Where the bottleneck of a system is likely to change from one of a plurality
of entities constituting the system into another entity as time elapses, detection
of a time-dependent change in the bottleneck, i.e., the shifting of the bottleneck
is possibly needed. The shifting of the bottleneck can be detected by detecting
the bottleneck not in association with a period of time but in association with
a point in time.
In light of the above finding, the method according to this mode (17) is implemented
to determine at least one of the plurality of entities which functions as the bottleneck
at a give point in time.
Although it is needless to say that this method is effective where the system
is a non steady state system bearing a non steady state in which the bottleneck
changes over time, this method may be practiced where the system is a steady state
system without bearing such a non steady state.
(18) The method according to any one of the above modes (1) to (16), wherein
the effect-level determining step comprises a bottleneck determining step of determining,
on the basis of a relationship in magnitude between the plurality of active times
determined in the active-time determining step for the plurality of entities at
each one of a plurality of discrete points of time, at least one of the plurality
of entities which affects the performance of the system at a higher level than
other entities, as at least one bottleneck.
The method according to this mode (18) would permit the bottleneck to be determined
in association with each point in time.
(19) The method according to the above mode (17) or (18), wherein the bottleneck
determining step is implemented, for at least one of the plurality of entities
which is situated in the active status at the given point in time, such that at
least one of the plurality of entities which functions as the at least one bottleneck
is determined on the basis of a relationship in magnitude between the at least
on active time determined in the active-time determining step.
Since it is not always that only one entity is in an active status at a given
point in time, two or more entities may be in an active status at the same time.
Where only one entity is active at a given point in time, it is automatically determined
that the one entity functions as the bottleneck at the given point in time. However,
where two or more entities are in active at a given point in time, it is possibly
needed to determine the order in which these entities are ranked by the strength
level at which each of these entities shows the qualification for the bottleneck.
In addition, it can be recognized that the more strongly each entity shows the
qualification for the bottleneck, the longer the active time of each entity is.
In view of the above findings, the method according to this mode (19) is implemented,
such that at least one of the plurality of entities which functions as at least
one bottleneck is determined on the basis of a relationship in magnitude between
the at least on active time determined for at least one of the plurality of entities
which is situated in the active status at the given point in time.
This method may be practiced in such a manner that an analysis reference time
required to be specified for determining the bottleneck in association with a point
in time, is set as an arbitrary point in time. Here, the "analysis reference time"
may be defined, for example, as one substantially equal to the "given point in
time" set forth in this mode (19), or as the "starting time" referred to in a second
embodiment of the present invention as described below. The above manner would
permit the bottleneck to be detected at any given point in time.
Further, the method according to this mode (19) may be practiced in such
a manner that a determination of the bottleneck is iterated over the range of different
points in time. This manner would permit monitoring of the shifting of the bottleneck
in the system in which the bottleneck changes over time, wherein the shifting is
defined as an event in which the bottleneck changes between entities of the system.
(20) The method according to the above mode (19), wherein the bottleneck determining
step is implemented to determine an with-maximum-active-time entity which is one
of the at least one of the plurality of entities which at least one is situated
in the active status at the given point in time, as the bottleneck, wherein the
with-maximum-active-time entity has substantially the longest active time among
the at least one of the plurality of entities.
The method according to this mode (20) is implemented to determine the bottleneck
of a system, behind the findings that it is common that an with-maximum-active-time
entity which is one of at least one of the plurality of entities which at least
one is situated in the active status at a given point in time, functions as the
bottleneck. Wherein, the with-maximum-active-time entity has substantially the
longest active time among the at least one the plurality of entities.
In implementing this method, if, for example, the number of the "at least one
of the plurality of entities which at least one is situated in the active status
at a given point in time" is single, no entity is capable of being compared with
the at least one of the plurality of entities, and therefore, the at least one
automatically qualifies as the "with-maximum-active-time entity."
To the contrary, if the number of the "at least one of the plurality of entities
which at least one is situated in the active status at a given point in time" is
plural, plural entities together situated in an active status at a given point
in time are compared with each other with respect to the active time, and the "with-maximum-active-time
entity" is determined depending on the results from the comparison.
The "with-maximum-active-time entity" in this mode (20) may be interpreted to
mean the entity with the truly longest active time, or the entity with an active
time in the vicinity of the longest active time, for example.
(21) The method according to the above mode (19) or (20), wherein the bottleneck
determining step is implemented to further determine a bottleneck period which
is a duration of the at least one bottleneck, on the basis of the collected data.
The method according to this mode (21) would determine the period during which
each entity is kept functioning as the bottleneck. Here, the period may be identified
with its location on a time base and its length.
According to an arrangement of this method, at least a part of the active
period of the with-longest-active-time entity set forth in the previous mode (20)
is determined as the bottleneck period. It is not always that the overall of the
active period qualifies as the bottleneck period. There can arise a case where
the active period of the with-longest-active-time entity overlaps with the active
period of another entity, and in this case, during the overlapping period, plural
entities are together in an active status, and therefore, it is difficult to identify
the bottleneck in the form of a unitary entity.
(22) The method according to any one of the above modes (17) to (21), further
comprising a shifting bottleneck determining step of determining, on the basis
of an overlap with respect to the active period between ones of the plurality of
entities which ones have been determined as a plurality of bottlenecks, at least
one of the ones which at least one performs a shifting of the bottleneck, as at
least one shifting bottleneck.
There exists a case where ones of the plurality of entities which ones have
been determined as bottlenecks overlap with each other with respect to the active
period. In this case, during the overlapping period, it is recognized that the
bottleneck shifts from the entity which was previously determined as the bottleneck
into another entity which was subsequently determined as the bottleneck.
Based on the above findings, the method according to this mode (22) is implemented
to determine, on the basis of an overlap between plural entities with respect to
the active period, at least one of the plurality of entities which at least one
performs a shifting of the bottleneck, as at least one shifting bottleneck.
(23) The method according to the above mode (22), wherein the shifting bottleneck
determining step is implemented to determine at least one of the plurality of entities
which at least one has an overlap with another entity with respect to the active
period, as the at least one shifting bottleneck, for at least a part of a period
permitting the overlap of the at least one with the another entity, wherein the
at least one entity and the another entity each have been determined as the bottleneck.
(24) The method according to any one of the above modes (17) to (23), wherein
the bottleneck determining step is implemented to determine in real time at least
one of the plurality of entities which functions as the at least one bottleneck
at a current point in time, during operation of the system,
the collected data is updated as the operation of the system progresses, whereby
the collected data is representative of a manner in which the each entity changes
in status with time, during a past period and the current point in time, not during
a future period, and
the active-time determining step comprises an active-period determining step,
operable when the each entity is situated in the active status at the current point
in time, of determining on the basis of the collected data a period from a starting
point in time at which the active status starts up to the current point in time,
as the active period of the each entity.
For an accurate determination as to whether or not each entity of a system functions
as the bottleneck of the system during each active period, an accurate determination
of an active time representative of the length of the active period to be considered
for determining the bottleneck. For this end, the starting time and the termination
time of the active period are required to be already known at the time of determining
the bottleneck.
In addition, a determination as to whether or not each entity functions as the
bottleneck is needed to be made in real time during operation of the system, in
some cases. One example is where it is necessary to quickly improve the performance,
i.e., the throughput (including the manufacturing capacity), by quickly detecting
the entity which functions as the bottleneck during operation of the system, and
by taking suitable countermeasures (e.g., resources, such as human, materials,
money, etc.) for the entity.
However, during operation of the system, there can be assumed the case where
the active period to be considered for determining the bottleneck has not been
terminated, leading to an incapability of obtaining the true active time. During
operation of the system, a future change in status of each entity could not be
obtained without any special idea.
On the other hand, if, where there exists an active period which is not yet terminated
at the current time defined as an analysis time, it allow to deem the period from
the starting time of the existing active period up to the current time, as an active
period, for the sake of convenience, on the basis of data indicative of changes
in status of each entity at the current time and at points in time during the past,
it becomes possible to detect the bottleneck in real time during operation of the
system, although it involves a more or less sacrifice of an accuracy in determining
the bottleneck.
In view of the above findings, the method according to this mode (24) is implemented,
such that, where each entity is currently in an active status, the period from
the starting time of the active status up to the current time is determined as
an active period, on the basis of data which represents time-dependent changes
in status of each entity at the current time and at points in time during the past,
but which fails to represent these changes at points in time during the future,
by being sequentially updated as an operation of the system progresses.
Further, this method is implemented, such that at least one of the plurality
of entities of the system which at least one functions as the bottleneck at the
current time is determined in real time during operation of the system, on the
basis of the length of the active period as described above.
Consequently, this method would permit, during operation of the system,
a real time detection of the entity currently functioning as the bottleneck, and
a real time detection of the shifting in which the bottleneck shifts from an entity
to another entity.
Therefore, this method would make it possible to, for example, add available
resources into the system in association with the shifting of the bottleneck, thereby
improving the performance of the overall system.
(25) The method according to the above mode (24), wherein the bottleneck determining
step is implemented, such that, after at least one of the plurality of entities
has been determined as the at least one bottleneck, a bottleneck period which is
a duration of the at least one bottleneck is not subject to any later correction
based on the data collected after determination of the at least one bottleneck.
For fulfilling the need to determine in real time the bottleneck of a system
with the method according to the above mode (24), once an entity of the system
has been determined as functioning the bottleneck during a given period, irrespective
of whether or not the determination has been made by ignoring the future during
which changes in status of the entity is unknown at the current time, it is not
allowed to later correct the content of the determination after termination thereof.
In light of the above findings, the method according to this mode (25) is implemented,
such that, after at least one of the plurality of entities has been determined
as the at least one bottleneck, the bottleneck period which is the duration of
the at least one bottleneck is not subject to any later correction based on the
data collected after determination of the at least one bottleneck.
It is noted that the method according to the above mode (24) may be practiced,
such that, after at least one of the plurality of entities has been determined
as the at least one bottleneck, the bottleneck period which is the duration of
the at least one bottleneck is corrected, on the basis of data representative of
the behavior of the system as collected after determination of the at least one bottleneck.
(26) The method according to the above mode (24) or (25), wherein the bottleneck
determining step comprises a step of determining, after an duration of at least
one of the plurality of entities which has been determined as the at least one
bottleneck has been determined as a precedent bottleneck period, the active period
of each one of at least one of the plurality of entities exclusive of a precedent
bottleneck having the precedent bottleneck, which the at least one is in the active
status at a reference time after-termination which is after a time at which the
precedent bottleneck period is terminated, wherein the active period has an overlap
with the precedent bottleneck period, as a subsequent bottleneck period exclusive
of the overlap.
There exists a case where the precedent bottleneck period for the precedent
bottleneck overlaps with the active period for the subsequent bottleneck having
the subsequent bottleneck period. In this case, under an environment where the
bottleneck is scheduled to be determined in real time, the precedent bottleneck
is previously determined as functioning the bottleneck during the overlap of the
active period for the subsequent bottleneck with the precedent bottleneck period,
and any correction to the previous determination may not be allowed. Consequently,
in this case, the overlap of the active period for the subsequent bottleneck with
the precedent bottleneck period is incapable of being treated as a part of the
subsequent bottleneck period.
Based on the above findings, the method according to this mode (26) is implemented,
such that the active period for the at least one entity which is in an active status
at the reference time after-termination which is after the time at which the precedent
bottleneck period is terminated, wherein the active period has an overlap with
the precedent bottleneck period, is determined as a subsequent bottleneck period
exclusive of the overlap.
(27) The method according to the above mode (26), wherein the bottleneck determining
step further comprises a step of determining, immediately after determination of
the precedent bottleneck period, that a shifting in which the bottleneck instantaneously
shifts from one of the plurality of entities having the determined precedent bottleneck
period to another entity having the determined subsequent bottleneck period.
(28) The method according to any one of the above modes (17) to (23), wherein
the bottleneck determining step is implemented to determine at least one of the
plurality of entities which functions as the least one bottleneck at a given point
in time, for an actual past operation, or an analyzed past operation by simulation,
of the system,
the collected data is representative of a manner in which the each entity changes
in status with time, not only during a past period before the given point in time,
but also during a future period after the given point in time, and
the active-time determining step comprises an active-period determining step
of determining, on the basis of the collected data, a period from a starting time
at which the active status starts up to a termination time at which the active
status is terminated, as the active period for the each entity.
There exists a case where a determination of an entity functioning as the bottleneck
at a give point in time is needed for an actual past operation, or a virtual past
operation analyzed by simulation, of the system. In this case, data which is collected
until the given point in time is representative of a manner in which each entity
changes in status with time, not only during the past period before the given point
in time, but also during the future period after the given point in time.
Accordingly, the method according to this mode (28) is implemented,
such that the period from the starting time of an active status up to the termination
time of the same active status is determined, per entity, as its active period,
on the basis of the collected data, for thereby determining the bottleneck for
the actual past operation, or the analyzed past operation by simulation, of the system.
Consequently, this method would permit an active period that is to
be considered for determining the bottleneck, to be obtained as the true active
period, resulting in an improved accuracy in determining the bottleneck.
(29) The method according to any one of the above modes (17) to (23), wherein
the bottleneck determining step is implemented to determine at least one of the
plurality of entities which functions as the at least one bottleneck at a current
point in time, during operation of the system,
the collected data is sequentially updated as the operation of the system progresses,
thereby representing a manner in which the each entity changes in status with time,
during a past period and the current point in time, not during a future period,
the active-time determining step comprises an active-period predicting step of
predicting by simulation the active time for at least one of the plurality of entities
which is situated in the active status at the current point in time, on the basis
of the collected data, and
the bottleneck determining step is implemented to determine the at least one
bottleneck, on the basis of the collected data and the predicted active time.
Where data indicative of the status of each entity is obtained during operation
of the system, the data represents the behavior of the system during the past and
the current point in time, but not during the future. However, it is possible to
predict by simulation the future behavior of the system, from the past and current
behavior of the system. One example of the reasons is that it is possible to formulate
a specified causal relation between the past and current behavior and the future
behavior, of the system.
The prediction of the future behavior of the system from the past and current
behavior of the system would make it possible to estimate the active time at a
certain adequate accuracy, in an attempt to determine the bottleneck during operation
of the system, even where the active period has not yet terminated, which is to
be considered for determining the bottleneck. It is recognized that the thus estimated
active time reflects the true active time more accurately than the active time
determined as a result of completely ignoring the future behavior of the system.
Based on the above findings, the method according to this mode (29) is implemented
to determine the bottleneck, in operation of the system, while predicting the future
behavior of the system.
(30) The method according to the above mode (28) or (29), wherein the bottleneck
determining step is implemented to determine at least one of the plurality of entities
as the at least one bottleneck, thereby determining at a determination time a duration
of the at least one bottleneck as a bottleneck period, and to subsequently correct
the determined bottleneck period, on the basis of data representative of a behavior
of the system shown after the determination time.
Where it is necessary to determine the entity functioning as the bottleneck
at a give point in time, for the actual past behavior, or the analyzed past behavior
by simulation, of the system, later correction to the previous determination of
the bottleneck is permitted. In addition, even where it is necessary to detect
the bottleneck in real time in operation of the system, later correction to the
previous determination of the bottleneck is possibly permitted.
In light of the above findings, the method according to this mode (30) is implemented,
such that, after the bottleneck period is determined for an entity, the determined
bottleneck period is corrected on the basis of data representative of the operational
behavior of the system shown after the previous determination of the bottleneck period.
Consequently, this method would readily improve the accuracy of the
resulting determination of the bottleneck.
(31) The method according to any one of the above modes (28) to (30), wherein
the bottleneck determining step comprises a first step of determining at least
one of the plurality of entities as the at least one bottleneck, thereby determining
an duration of the at least one bottleneck as a precedent bottleneck period; and
of subsequently the active period of at least one of the plurality of entities
exclusive of a precedent bottleneck having the precedent bottleneck period, which
at least one is in the active status at a reference time after-termination which
is a time after the precedent bottleneck period is terminated, as a subsequent
bottleneck period resulting from a beforehand or afterward exclusion of at least
one portion of an overlap of the active period with the precedent bottleneck period,
from the active period.
There exists a case where the active period of the at least one entity which
is in an active status at the time after the precedent bottleneck period of another
entity, i.e., the precedent bottleneck is terminated. In this case, it is normally
unreasonable to consider that the at least one entity functions as the bottleneck
over the overlap of the active period of the at least one entity with the precedent
bottleneck period.
In view of the above findings, the method according to this mode (31) is implemented
to determine the active period for at least one entity other than the precedent
bottleneck, which at least one is in an active status at the reference time after-termination
which is the time after the precedent bottleneck period is terminated, wherein
the active period has an overlap with the precedent bottleneck period, as a subsequent
bottleneck period resulting from a beforehand or afterward exclusion of at least
one portion of the overlap, from the active period.
In this mode (31), the phrase "beforehand exclusion of at least one portion"
means
that the determined bottleneck period does not include the at least one portion,
at the beginning, while the phrase "afterward exclusion of at least one portion"
means that, although the determined bottleneck period originally had the at least
one portion, the determined bottleneck period is thereafter corrected so as not
to include the at least one portion.
(32) The method according to the above mode (31), wherein the bottleneck determining
steps further comprises a second step of determining, after termination of implementation
of the first step, that there is performed between the precedent and the subsequent
bottleneck period a shifting in which the bottleneck shifts from an entity which
has the precedent bottleneck period to another entity which has the subsequent
bottleneck period.
The method according to this mode (32) would make it possible to detect a shifting
in which the bottleneck shifts from an entity to another entity.
(33) The method according to the above mode (32), wherein the second step comprises
a step of determining as a shifting period during the shifting is continuously
performed, an overlap of the active period of one of the plurality of entities
which has the subsequent bottleneck period, with the active period of one of the
plurality of entities which has the precedent bottleneck period.
The method according to this mode (33) would define the shifting of bottleneck
as a continuous event.
(34) The method according to the above mode (
