Figure 1. New Users
of Simulation Software
Abstract
Computer hardware continues to decrease in
cost and increase in power. Simulation vendors have hypothesized
that this situation, together with new requirements for system
integration, has produced a new business environment for discrete-event
simulation all over the world. This paper, which provides an
update to an earlier report, describes current trends in Japanese
industry regarding the use of discrete-event simulation.
The information contained in the paper was
derived from interviews with many representatives of Japanese
industry. It shows that, in general, this hypothesis is true
in Japan.
Recent advances in a number of technologies
have provided industrial users with high performance computer
hardware and powerful Graphical User Interfaces (GUI). This,
in turn, has led to significant advances in computer performance
and visualization. The result is a collection of very powerful
software tools for solving problems across the manufacturing enterprise
[1-4]. Part designers, manufacturing engineers, and production
managers can now compute and visualize the impact of their
individual decisions right at their desktop. They can assess,
to a limited degree, the impact of their decisions on both product
and process performance and company profitability. To compute
the real impact of those decisions, however, they must work together.
That is, their decisions must be made in a coordinated way.
This requires communication between the people, and integration
of the software tools that they use to make those decisions [5-8].
One of those tools is simulation. Recently, many
simulation tools (both continuous and discrete-event simulation
tools) have become available on the market. [9-15]. The advances
in computing mentioned above have made it possible to run these
tools on desktop computers. This, in addition to improvements
in the tools themselves, is the main reason that simulation is
used heavily in manufacturing plants around the world.
Nowhere is this more evident than in Japan, where
simulation use is on the rise. This paper, which extends our
earlier work [16-18], summarizes the current use of simulation
in Japanese industry. It also discusses some of the criteria
used to make purchasing decisions. Finally, it presents some
preliminary views of Japanese industry toward the use of simulation
as an enabling technology for both system integration and virtual
manufacturing.
2. Simulation practices in Japanese industries
Figure 1 displays the evolution of simulation use in Japanese industry. It shows the number of new manufacturing firms using simulation in their company. The graph shows very little interest after its introduction into Japan more than 25 years ago. However, starting in 1985, a gradual increase occurred.
Figure 1. New Users
of Simulation Software
Around 1989 or 1990, a dramatic expansion occurred
when many manufacturing enterprises made major investments in
computer technologies and manufacturing equipment. Following
this, a sharp reduction took place. The number of new users declined
to pre-89 levels and remained relatively constant for a few years.
The main reason for this, we believe, was the downturn in the
Japanese economy. Then, remarkably, another dramatic increase
took place in 1994 or 1995. We think that the main reason for
this was the rapid advances in computing technologies mentioned
above. These changes included a decrease in price, growth in
processing power, more robust operating systems, and better user
interfaces. Many vendors took advantage of these changes to release
desktop simulation tools. These new tools had graphical user
interfaces that were integrated with existing windows-based operating
systems (see Table 1). This was very attractive to simulation
users.
2.2 Target applications of simulation analysis
The two main users of simulation tools are communications companies and manufacturing enterprises, with the latter being the most popular. The target applications within these companies that are the subject of simulation studies are shown in Figure 2.
The most popular applications deal with the movement
of raw materials and finished goods. These include factory material
handling systems, logistics systems including automated warehousing,
and transportation to and from those warehouses.
To date, the least popular applications areas are
computer and communication systems. However, we believe that
Certain commercial software and hardware products are identified in this paper. This does not imply approval or endorsement by NIST, nor does it imply that the identified products are necessarily the best available for the purpose.
this is likely to change since many companies have
introduced Local Area Networks that are connected to the INTERNET.
As these systems grow in complexity and importance, industrial
users will require tools to analyze various design alternatives
and to predict performance.
2.3 Simulation project characteristics
Figure 3 shows simulation projects organized by department. The people who actually carry out the project can be classified into two groups: the developers' group and the users' group. People from the developers' group analyze potential problems, design simulation models to solve those problems, and implement those models on target computing environments. People from the users' group run the developed models to give a solution to specific instantiations of those problems. That is, they provide the actual data to run the models. For example, suppose the plant manager decided that a new simulation-based scheduler was needed. The developers' group would develop the model with parameters for all of the required variables. Each time a new schedule was required, a user would input that data, run the simulation, and get a new schedule.
Figure 3. Simulation
developers and users
The level of effort required to complete a simulation
project varies between 0.2 worker-months and 24 worker-months,
with an average of a little more than 2 worker-months (Fig. 4).
This is difficult to estimate because 1) the workload changes
as the project evolves, and 2) larger projects require more effort
than smaller projects. Typically, the model development and data
collection phases of the project account for approximately 60%
of the entire effort. The remainder is split equally among the
other phases. Table 2 also shows the range for the various phases.
| Data Collection /Gathering | |
| Model design | |
| Animation | |
| Model modification | |
| Simulation experiments | |
| Summary of result |
2.4 Experimental design and output analysis
Depending on the complexity of the project, a statistical design might be required to determine 1) how many simulation experiments to run, 2) what kind of data to collect, and 3) how to vary the inputs parameters to minimize the overall variance in that data. Experimental design methods like orthogonal arrays and Taguchi methods [19] are very popular and powerful methodologies used in Japan. These methodologies can support multiple views into an enterprise including financial, accounting, sales, marketing, management, information, and manufacturing. Table 3 provides some summary information on the use of experimental design in simulation projects today. Clearly, there is no widespread use of these techniques in industry at this time. These results have not changed substantially from those reported in our earlier work.
.
| Use experimental design method | 15 | utilize orthogonal array methods do not use orthogonal array methods | 9 6 |
| Do not use experimental design | 68 |
The final step in a simulation project is the analysis
of the output data [20-23]. This will include statistical calculations
to compute estimates of performance measures, a comparison of
one set of measures against another, or both. The knowledge required
to carry out these calculations is very specialized. Since most
users will not have this knowledge, it is important that simulation
software packages support a variety of statistical data analysis
routines. These routines should allow users to generate the required
results easily and with confidence. Table 4 summarizes the use
of statistical analysis routines.
| Do statistical analysis | 36 | implement own analysis routines use supported analysis routines use statistical data analysis packages | 6 24 8 |
| Not do statistical analysis | 45 |
The users gave the following reasons for not using
statistical analysis.
The implication of this is that most users still
perform deterministic simulations that include no stochastic variables.
Motivations for using this approach include:
Many of these are what is commonly called what
if simulations, because they show what happens in the
system if particular plans and schedules are executed.
They will identify problems, workload, lead-time, bottleneck
processes, bottleneck queues and others. They will also provide
some insight as to why these things happen. The interactive parameter
setting function will help such studies. More than half of users
have implemented such a function (Table 5).
| Interactive parameter setting functions | |||
| Support the functions | Frequent use
Occasionally None |
| |
| Not support the functions |
2.5 Model maintenance
About 40 % of simulation models are reused
(Fig. 5). Some of them are copied into new modeling code as
is, and some are modified for use in new simulation models.
Figure 6 shows the various methods that are used to accomplish
this. By far, the most effective of these methods is a model
library. By using a model library, model builders can use models
that have been built and tested previously to address the problem
at hand. The ability to do this reduces the time and cost associated
with the development of a new model. The goal is have models
of each component in enterprise system [24]. This way, models
of different systems can be built from these components quickly
and easily. The model library can also include development tools
that will support the integration of simulation tools with other
management tools.
Other problems associated with model reuse, and possible
methods for addressing those problems are described below:
Figure 5. Frequency
of model re-utilization
3. Reasons For Buying Particular Software Packages
3.1 Current Market shares
As noted above, there are a large number of simulation software packages on the market today. During the
last five years, many vendors have developed relationships
with Japanese firms who sell their products and provide consulting
services. Even though the simulation software market in Japan
is not as mature as in the United States, these efforts are beginning
to pay off (see Figure 7). The number of different simulation
products in use in Japan has increased sharply since our earlier
reports [16-18]. Even so, there is
still a requirement to implement models in packages based on general
programming languages. This
can be seen from the relatively large usage of packages like SLAMSYSTEM
and WITNESS. 
3.2 User Interface Capabilities
Table 6 shows some of the user interface
factors which users consider when they purchase simulation software.
The top two, modeling support and output analysis tools, have
remained unchanged from our earlier reports. In those reports,
we combined graphical displays and animation into a single factor.
In this report, we have put them into two separate categories.
Since animation appears to be a very important consideration in
the United States, it is perhaps surprising that it is Japanese
users rated it only fourth. We believe that this reflects the
current decision making style of management people in Japanese
firms - it is very rare that these people make important decisions
based on the business reviews and presentations. Consequently,
the role of animation will become gradually more important only
if this style changes.
| Modeling support tool | ||
| Statistical simulation output analysis | ||
| Graphical display routine for simulation output | ||
| Animation facility | ||
| Interactive parameter handling facility in simulation execution | ||
| Others |
3.3 Modeling Capabilities
Figure 8 summarizes users' views on the different capabilities offered by various packages.
Modeling support tools are considered most important
when choosing simulation software for particular applications.
User-friendly, icon-based modeling tools reduce the time and
cost associated with building and maintaining simulation models.
Some of these tools are aimed at specific application domains
only, like manufacturing or communications. Others provide a
collection of templates for various domains. The system definition
language for these tools is contained in the domain expertise
of the user. This is accomplished by providing:
As a result, these tools can generate the simulation
model directly from system configuration data. This gives the
user the ability to automatically model changes in dynamic systems
at various levels of complexity and detail.
3.4 Animation Capabilities
Animation facilities enable the user to visualize
the results of simulation directly. Users can see which process
is the bottleneck or which process is not in tune with other ones.
Many users indicated the importance of animation capabilities
of simulation software. Currently, only 60% of users run animation
programs every time; while 20% say that they run them often. Users
have summarized their reasons for not using animation more frequently.
They are given below:
As we said earlier, this is a little surprising,
but we expect this number to grow in the coming years.
Figure 8. Capabilities
of simulation software 3.5 Other reasons
Several other reasons were given for selecting
particular packages. They include: the ability to address problems
in a specific target domain; the ability to handle models of
a particular scale; the ability to provide a particular level
of accuracy; the suitability to a user's modeling skills; and,
the ability to finish the project on time. Some examples are
given below
- Use SIMTOOL for complex and large-scale simulation models.
- Use FACTOR/AIM to mode manufacturing line or when rapid solution is required.
- Use AUTOMOD when an exact layout or 3 dimensional simulation is required.
- Use WITNESS when designer use simulation or when rapid solution is required
- Use AUTOMOD or EXTEND when the target system is material handling or AS/RS.
- Use SIMSCRIPT when the target system is traffic flow simulation, and use AUTOMOD.
- Use SIMAN when the target system is
a lot-based production system.
4. Simulation and System Integration
The requirements for the integration of simulation
with other manufacturing software applications and manufacturing
databases are increasing. The subsequent comments from simulation
users provide support for this view:
Fig. 9 shows the current implementation of system
interfaces with simulation. About 30% of the cases connect simulation
with other application systems. Most of these cases combine simulation
with production planning system, floor shop control systems, and
engineering support systems. Implementations are currently realized
by using file transfer mechanisms, but the majority of them are
moving toward real-time data exchange using some sort of messaging
protocol or database access. Some examples are:
We asked users to comment on application systems
they planned to integrate with simulation in the near future.
The systems they included were
Virtual Production Systems are by far the most ambitious
of these plans. The objective is the synchronization of material-flow
with information-flow across all factory operations. This is
achieved through the integration of manufacturing automation
processes, business processes, and data management systems (Fig
10). The implementation will require both advances in optimization
technologies [25-29] and distributed simulation techniques [30-32].
Figure 9. Current
supporting system interfaces
The use of simulation in Japanese industry
is still modest compared to the United States, but it is on the
rise. With the growing emphasis in Japan on virtual manufacturing,
particularly through the Intelligent Manufacturing System program
[33], we expect this trend to continue. Some of the most important
areas of future interest among Japanese users are:
6. References
[ 1] A.M. Law and W.D. Kelton, Simulation Modeling & Analysis 2nd Ed, McGraw-Hill Inc., 1991
[ 2] W. Kreutzer, System Simulation Programming Style and Languages, Addison-Wesley, 1986
[ 3] Michael Pidd, Computer Simulation in Management Science 3rd Edition, John Wiley & Sons
Lid, 1992
[ 4] A. Carrie, Simulation of Manufacturing System, John Wiley and Sons, 1988
[ 5] ISO, Framework for Enterprise Modeling, ISO/TC184/SC5/WG1, 1993
[ 6] Williams, T.J., The Purdue Enterprise Reference Architecture, Instrument Society of America, Research Triangle Park, NC, 1992
[ 7] Williams, T.J., Architecture for Integrating manufacturing activities and enterprises, Proc. of Workshop on the design of Information Infrastructure Systems for Manufacturing, 1-17,1993
[ 8] Pels, H. J. and J. C. Wortmann. (ed.), Integration in Production Management Systems, IFIP Transactions B-7, IFIP, 1992
[ 9] A. Mullermey, SIMSCRIPT II.5 Programming Language, CACI, 1983
[10] J.O. Henriksen, The GPSS/H User's Manual, Wolverine Software, Falls church, Va, 1979
[11] A.A.B. Pritsker, Introduction Simulation and SLAM II, John Wiley and Sons, 1986
[12] G.M. Birtwistle et al, SIMULA begin, student literature, Sweden, 1985
[13] K. Nygaard and O. Dahl, The Development of the SIMULA language, ACM SIGPLAN
Notices, 13,8,245-272,1978
[14] D.F. Geuder, Object Oriented Modeling with SIMPLE++, Proc. of Winter Simulation
Conference, 534-540, 1995
[15] AESOP GMDH, SIMPLE++ Users Manual, AESOP, 1994
[16] S. Umeda, A state-of-the-art: Discrete Event Simulation in Japan , New Directions in
Simulation for manufacturing and Communications, ORSJ, 1994
[17] S. Umeda, Industrial practices of discrete-event simulation in Japan, ORSJ seminar
textbook, 1996 (in Japanese)
[18] A. Jones (eds.) Simulation in Japan: A State-of-the-Art Report, NISTIR 5614, 1994
[19] G. Taguchi, The experiment design method, Maruzen, 1990 (in Japanese)
[20] J. Kleijnen, Statistical Tools for Simulation Practitioners, Marcel Dekker Inc., 1987
[21] J. Kleijnen and W.V. Groenendaal: Simulation A Statistical Perspective, 1989
[22] G.E.P. Box and N.R. Draper: Empirical Model Building and Response Surfaces, John Wiley
& Sons, New York, 1987
[23] J. Kleijnen and W. Groenendaal, Simulation: A statistical Perspective, Wiley, 1992
[24] C.A. Roberts et al. (eds.) Object-Oriented Simulation, Proc. of 1995 Western Multi-
Conference, 1995
[25] S. Umeda, Simulation-based Production Control System in CIM environment, Proc. of
ICPR, 353-354,1993, 8
[26] S. Umeda, "Virtual Plant System" for Real Time Scheduling and Resource Management,
Proc of the 3rd Int. Conference on Automation Technology, 279-284,1994, 7
[27] M.S. Meketon, Optimization of Simulation, Proc. of WSC, 1987
[28] Y.C. Ho and X.R. Cao: Perturbation Analysis and Optimization of Queue in Networks, J. of
Optimization Theory and Applications, 40, 559-582, 1983
[29] S. Takakuwa, Simulation Optimization in CIM production systems, Corona Co. 1994, (in
Japanese)
[30] D.K. Arvind, et al.(eds.) Proc. of the 7th Workshop on Parallel and Distributed Simulation
(PADS'93),1993, ACM SIGSIM
[31] R. Bagrodia, et al. (eds.) Proc. of the 8th Workshop on Parallel and Distributed Simulation
(PADS'94), 1994, ACM SIGSIM
[32] S. Fujii et al. A Study on Distributed Simulation for Flexible Manufacturing System, Proc. of
IFAC, 27-32, 1989
[33] U. S. Department of Commerce, Technology Administration, Office of Technology Policy.
IMS Intelligent Manufacturing Systems - A Program for International Cooperation in Advanced
Manufacturing: Final Report of the International Steering Committee adopted at ISC6, Hawaii,
24 to 26 January, 1994