Discrete Event Simulation v. System Dynamics

[Rod MacDonald]

Hello Aitor,

I would recommend Pidd, M. (1992) Computer Simulation in Management
Science. Chichester: John Wiley and Sons.

Sincerely,
Rod MacDonald
From: Rod MacDonald <Rod@mindwalkconsulting.com>

["Jay W. Forrester"]

Perhaps you will find something of use below.

Information on System Dynamics
Jay W. Forrester
January 28I, 1999

System Dynamics Bibliography:

To order the system dynamics bibliography of over 4100 entries,
specify IBM type PC, or Macintosh

Send $35 in US$ drawn on a US bank to:

System Dynamics Society
Roberta Spencer, Executive Director
Milne 300--Rockefeller College
State University at Albany
Albany, NY 12222 USA

tel: 1-518-442-3865=DD
fax: 518-442-3398
email: System.Dynamics@albany.edu

Three formats are available:

1. For Endnote, a very effective bibliography software available for either
Macintosh or PC from:

Niles & Associates, Inc
800 Jones St.
Berkeley, CA 94710 USA

Tel: 510-559-8592
Fax: 510-559-8683
Internet: nilesinc@well.sf.ca.us

I use Endnote and recommend it and use it to search for the references.

2. An exported version with field delimiters that presumably can be loaded
into some other kind of database.

3. A listing that one can look at in a word processor and do some simple
finding operations.

The bibliography can also be downloaded from:
http://www.vensim.com/sdmail/sdbib.html
-----------------------------------------------

The publications list of the System Dynamics Group at MIT is available on
the web as an Adobe Acrobat document from:
=46TP://Sysdyn.MIT.edu/Ftp/sdep/papers/D-3059-43.pdf

---------------------------------------
Membership in the System Dynamics Society and subscription to the System
Dynamics Review
are US$80 per year for regular members
and US$40 for students.
Send application to:

John Wiley & Sons
Periodicals Division, System Dynamics Review
P.O. Box 7247 8491
Philadelphia, PA 19170

or to

John Wiley & Sons
Journals Administration, System Dynamics Review
1 Oldlands Way
Bognor Regis
West Sussex PO22 9SA

To contact the office of the System Dynamics Society and to order copies of
the "Beer Game" group simulation exercise:
tel: 1-518-442-3865
fax: 518-442-3398
email: System.Dynamics@albany.edu
email: rspencer@cnsvax.albany.edu

-------------------------------------
There is a system dynamics discussion group on the Internet.
To join, send email to: majordomo@world.std.com
In the body of the message, enter the following two lines:

Subscribe system-dynamics
End

------------------------------------
The next annual international conference of the System Dynamics Society
will be in New Zealand, July 20-23, 1999. Write to the System Dyanmics
Society,

System Dynamics Society
Roberta Spencer, Executive Director
Milne 300--Rockefeller College
State University at Albany
Albany, NY 12222 USA
tel: 1-518-442-3865=DD
fax: 518-442-3398
email: System.Dynamics@albany.edu
----------------------------------------

The publisher for books in the following block has changed from
Productivity Press to Pegasus Communications.

Pegasus Communications, Inc.
One Moody Street
Waltham, MA 02453-5339

Within the U.S
tel:1-800-272-0945
fax: 1-800-701-7083
Outside the U.S.
tel: 781-398-9700
fax: 781-894-7175
Web page: www.pegasuscom.com

Alfeld, Louis Edward, and Alan K. Graham. 1976. Introduction to Urban
Dynamics. Waltham, MA: Pegasus Communications. 333 pp.

=46orrester, Jay W. 1961. Industrial Dynamics. Waltham, MA: Pegasus
Communications. 464 pp.

=46orrester, Jay W. 1968. Principles of Systems. (2nd ed.). Waltham, MA:
Pegasus Communications. 391 pp.

=46orrester, Jay W. 1969. Urban Dynamics. Waltham, MA: Pegasus
Communications. 285 pp.

=46orrester, Jay W. 1971. World Dynamics. (1973 second ed.). Waltham, MA:
Pegasus Communications. 144 pp. Second edition has an added chapter on
physical vs. social limits.

=46orrester, Jay W. 1975. Collected Papers of Jay W. Forrester. Waltham, MA:
Pegasus Communications. 284 pp
=2E
=46orrester, Nathan B. 1973. The Life Cycle of Economic Development. Waltham=
,
MA: Pegasus Communications. 194 pp.

Goodman, Michael R. 1974. Study Notes in System Dynamics. Waltham, MA:
Pegasus Communications. 388 pp.

Lyneis, James M. 1980. Corporate Planning and Policy Design: A System
Dynamics Approach. Waltham, MA: Pegasus Communications. 520 pp.

Mass, Nathaniel J., ed., 1974. Readings in Urban Dynamics: Volume I,
Waltham, MA: Pegasus Communications, 303 pp.

Mass, Nathaniel J. 1975. Economic Cycles: An Analysis of Underlying
Causes. Waltham, MA: Pegasus Communications. 185 pp.

Meadows, Dennis L. 1970. Dynamics of Commodity Production Cycles. Waltham,
MA: Pegasus Communications. 104 pp.

Meadows, Dennis L., et al. 1974. Dynamics of Growth in a Finite World.
Waltham, MA: Pegasus Communications. 637 pp.

Meadows, Dennis L., and Donella H. Meadows, ed., 1973. Toward Global
Equilibrium: Collected Papers, Waltham, MA: Pegasus Communications, 358
pp.

Morecroft, John D. W., and John D. Sterman, ed., (1994). Modeling for
Learning Organizationa, Waltham, MA: Pegasus Communications, 400 pp.

Randers, Jorgen, ed., 1980. Elements of the System Dynamics Method,
Waltham, MA: Pegasus Communications, 488 pp.

Richardson, George P., and Alexander L. Pugh III. 1981. Introduction to
System Dynamics Modeling with DYNAMO. Waltham, MA: Pegasus Communications.
413 pp.

Roberts, Edward B. 1978. Managerial Applications of System Dynamics.
Waltham, MA: Pegasus Communications. 562 pp.

Roberts, Nancy, David Andersen, Ralph Deal, Michael Garet, William Shaffer.
1983. Introduction to Computer Simulation: A System Dynamics Modeling
Approach. Waltham, MA: Pegasus Communications, 562 pages
=2E
Schroeder, Walter W., III, Robert E. Sweeney, and Louis Edward Alfeld, ed.,
1975. Readings in Urban Dynamics: Volume 2, Waltham, MA: Pegasus
Communications, 305 pp.

---------------------------------------------------
Books from other publishers include:

Coyle, R. G., 1996. System Dynamics Modelling--A Practical Approach,
London: Chapman & Hall. 413 pp.

Mandinach, Ellen B., and Hugh F. Cline, 1994. Classroom Dynamics:
Implementing a Technology-Based Learning Environment, Hillsdale, NJ:
Lawrence Erlbaum Associates. 211 pp.

Richardson, George P., 1991. Feedback Thought in Social Science and Systems
Theory, Philadelphia, PA: University of Pennsylvania Press. 374 pp.

Richardson, George P., 1996. Modelling for Management: Simulation in
Support of Systems Thinking, Brookfield, Vt.: Dartmouth Publishing. 493 &
447 pp.
----------------------------------
A self-study guide to system dynamics, called "Road Maps," is available for
downloading from:

http://sysdyn.mit.edu

or in paper copy from:

Creative Learning Exchange
Ms. Lees Stuntz, Director
1 Keefe Road
Acton, MA 01720, USA
tel: 1-508-287-0070
fax: 1-508-287-0080
email: stuntzln@tiac.net

---------------------------
=46or those wanting information on introducing system dynamics in
kindergarten through 12th grade education:

1. The Creative Learning Exchange is a nonprofit foundation that acts as a
clearinghouse to provide information on system dynamics in precollege
education and to help teachers share their experiences. They can be
reached at:

Creative Learning Exchange
Ms. Lees Stuntz, Director
1 Keefe Road
Acton, MA 01720, USA
tel: 1-978-287-0070
fax: 1-978-287-0080
email: stuntzln@tiac.net

2. The System Dynamics in Education Project at MIT has a web page with
links to pages at other K-12 activities:
http://sysdyn.mit.edu/

3. Many papers on curriculum can be found at
sysdyn.mit.edu

4. An internet discussion group on K-12 issues related to system dynamics i=
s:
k-12sd@sysdyn.mit.edu
To subscribe, send a note to:
mailmasters@sysdyn.mit.edu.
Please provide the following information:
First Name:
Last Name:
E-mail:
Title:
Organization:
Address:
City:
State or Province:
ZIP or Postal Code:
Country:
Day Phone Number:
Evening Phone Number:
Fax Number:

5. The summer 93 issue of the System Dynamics Review, vol 9 no. 2, was a
special issue on "Systems thinking in education" It contains many
interesting pieces including reports from the field by teachers.

----------------------------------------------

There are now three good software packages for system dynamics. You can
request information:
--------------------------------------------
STELLA for Macintosh or PC:

High Performance Systems
45 Lyme Road, Suite #300
Hanover, NH 03755, USA

Phone: 1-603-643-9636 customer support
tel: 1-800-332-1202 product inquiries
fax: 1-603-643-9502
email: support@hps-inc.com
http://www.hps-inc.com/

--------------------------------------

Powersim for PC:

Powersim Corporation
1175 Herndon Parkway Suite 600
Herndon, VA 20170
Phone: (703) 481-1270
=46ax: (703) 481-1271
Email: powersim@powersim.com
http://www.powersim.com


Norway Address:

Powersim AS
PO Box 206
N-5100 Isdalst=F8
Phone: +47 56 34 24 00
=46ax: +47 56 34 24 01
Email: powersim@powersim.no
http://www.powersim.no
-------------------------------------------

Vensim for PC or Macintosh:

Ventana Systems, Inc.
60 Jacob Gates Road
Harvard MA 01451, USA

tel: 1-617-489-5249
fax: 1-617-489-53316
email: vensim@vensim.com
http://www.vensim.com

A "Personal Learning Edition" of Vensim and its manual can be downloaded
free from:
http://www.vensim.com/download.html



---------------------------------------------------------
Jay W. Forrester
Professor of Management, Emeritus
and Senior Lecturer, Sloan School
Massachusetts Institute of Technology
Room E60-389
Cambridge, MA 02139
tel: 617-253-1571
fax: 617-258-9405

[ann.s.chen@ac.com]

Hello Aitor!

I dont have info about the use of SD in manufacturing applications, but I
just wanted to let you know of an article comparing SD and Discrete Event.
On the homepage for ImagineThat Inc. (http://www.imaginethatinc.com/ ),
there is a fairly insightful article comparing SD and Discrete Event
Simulation. When you get to the page, go to Simulation, and then Simulation
Concepts Paper(at the very bottom of the page).

Hope you find it helpful!

Ann
From: ann.s.chen@ac.com

["Raymond T. Joseph"]

Aitor,

I have seen no single location for an extensive comparison. Most of my
experiences with literature is from the IEEE Transactions on Automatic
Control.

DES optimization methods require exhaustive searches. For industrial
problems, the solution set is too large for such methods. One usually ends
up using heuristics to choose subsets of the solution set to search.

SD techniques have many advantages except they cant track individual
events, only statistics of events. Thus, SD may not resolve differences in
neighboring solutions.

A combination of the two may be extremely advantageous: SD techniques may
not be able to differentiate between two neighboring candidate solutions as
DES methods, but it can give a relative comparison across a broad solution
set. Thus, one could use SD to localize a search or at least partition a
solution space. DES could then be used to search those subsets.

Now, what would be cool from a user perspective is to be able to describe a
model at some high level and then have an application that would decompose
this into both SD and DES formats. These could then be feed to optimization
applications.

Raymond Joseph, PE
Aarden Control Engineering and Science
r.joseph@ieee.org

[John Sterman]

Bill Harris is right - the essence of system dynamics is the feedback
perspective. The assumption of continuous flows is a modeling
approximation, useful for many but far from all problems and model
purposes. It is perfectly possible and often done to incorporate feedback
processes in discrete simulations. There is in principle no constraint on
incorporating feedback in discrete simulation models. In practice,
however, many such models have rather narrow boundaries and exclude
feedbacks known to be important (e.g. queuing models assuming exogenous
arrivals when in reality people balk based on their judgment of the
expected wait time). Note that many queuing models do include these loops.
The key, as always, whatever the modeling method chosen, is to evaluate
whether or not the simplifying assumptions, including aggregation and model
boundary - are appropriate relative to the model purpose.

John Sterman
From: John Sterman <jsterman@MIT.EDU>

["Phil Odence"]

I think the feedback issue is often related to a timeframe component as
well.

Queuing models may run over a hours or days, but if it takes weeks or months
for potential queue-ees to build up expectations for queuing times, that
model would not/could not take that into account except as an exogenous
scenario. So the queuing model that incorporates enough feedback to know
that a car cant pull ahead until the next one does, may not take into
account that people whove had to wait too long on their last couple of
trips to McDonalds could head to Burger King next time.

At the risk of over-generalizing: DES models tend to run over shorter
timeframes; SD models tend to run over longer timeframes and incorporate
longer term feedbacks.

Phil Odence
High Performance Systems, Inc.
45 Lyme Road, Suite 200
Hanover, NH 03755
603.643.9636.x107, fx 9502
www.hps-inc.com
From: "Phil Odence" <podence@hps-inc.com>

[A.Oyarbide-Zubillaga.1998@Cranfi]

Hello everyone.

It is my first message in this group. Let me present myself.=20

I am a spanish guy, but in this moment I am in UK studying.=20

I just have start a research program, which its aim is to apply System
Dynamics methodology in manufacturing systems, and compare it with
Discrete Event Simulation.

I know that SD technique is quite old, and I have been in Forrester
web page, but I couldnt find relevant information about the
aplication of SD in manufacturing systems, and its strengths and
weakness against Discrete Event Simulation.

Can someone help me and tell me where can I found this kind of
information? And to finish, can someone recommend me references?

Thanks for your attention.


___________________________________________
/ Aitor Oyarbide (Cranfield University) r ,-~~-.___. _|A.Oyarbide-Zubillaga.1998@cranfield.ac.uk|
/ | | =
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D |
( ) 0 | Estar preparados es importante, |
_/-, ,---- | saber esperar lo es aun mas, |
=3D=3D=3D=3D // | pero aprovechar el momento adecuado =
|
/ -~; /~~~(O) | es la clave de la vida. |
/ __/~| / | | ______________________________________|
=3D( _____|(_________| _/_____ICQ:29224736_____________TTLPK___/

[JandJpugh@aol.com]

Which is better depends on the questions that you wish to answer. If you are
worried about manufacturing floor layout, machine scheduling, machine downtown
and the like use discrete modeling. Alternatively, if you want to relate
manufacturing to product development and marketing use SD.

Jack Pugh
From: JandJpugh@aol.com

[Bruce Campbell]

Aitor,

One place to check out is http://www.imaginethatinc.com/

This does not compare SD with discrete event specifically, but gives a
good overview of continuous (eg SD) modelling vs discrete event
modelling and some of the strengths and weaknesses of both.

As far as Im aware no-one has compared SD with a specific discrete
event method which is good for your study. It is also worthwhile as both
tend to model similar situations. I suspect that both are possibly used
in the wrong situations at times. ie SD used when discrete event would
be warranted and vice versa.

There is a bit of literature around describing the above situation -
modellers make the problem fit the method they are familiar with.

All the best,

Bruce Campbell



--
Bruce Campbell
Joint Research Centre for Advanced Systems Engineering
Macquarie University 2109
Australia

E-mail: Bruce.Campbell@mq.edu.au
Ph: +61 2 9850 9107
Fax: +61 2 9850 9102

["Bernhard Angerhofer"]

Aitor,

there is a paper in the proceedings of the International System Dynamics =
Conference 1996, that might be of interest to you:

"System Dynamics and Discrete Simulation in a Constant Work-in-Process =
System: A Comparative Study".
Authors: Ruiz Usano, R., Manuel Framinan Torres, J., Crespo Marquez, A., =
Zubiria de Castro, R.
Universidad de Sevilla, Dpto. Organizacion Industrial y Gestion de =
Empresas Escuela Superior de Ingenieros.

The authors compare System Dynamics and discrete event simulation =
applied to a production line.

Bernhard

*******************************************************
Bernhard Angerhofer
School of Computing, Information Systems & Mathematics
South Bank University, 103 Borough Road, London SE1 0AA
e-mail: b-angerhofer@sbu.ac.uk

["Lane,DC"]

I think that it important to understand that System Dynamics and DES are
rather different approaches, with different styles of formulation, modelling
languages and, most importantly, aims for the results of the modelling.
Together, these mean that the two approaches are suitable for rather
different types of problem.
I think that both deserve respect, in the same way that a good, sharp saw
and a finely balanced hammer deserve respect for their ability to do
particularly jobs well.
For an analysis if the relevance and benefits of the two approaches to
healthcare problems you might like to look at:

Taylor, K. S. and D. C. Lane. 1998. Simulation Applied to Health Services:
Opportunities for applying the system dynamics approach. Journal of Health
Services Research and Policy 3(4), 226-232.

Copies are also available from me by send your address.


Regards,

David C. Lane
_______________________________________________________________

Dr. David C. Lane
Operational Research Department
London School of Economics and Political Science
Houghton Street
London WC2A 2AE
United Kingdom
e-mail: d.c.lane@lse.ac.uk
Tel. (44) (0) 171-955-7336
Fax. (44) (0) 171-955-6885

[Bill Harris]

> Which is better depends on the questions that you wish to answer. If you are
> worried about manufacturing floor layout, machine scheduling, machine downtown
> and the like use discrete modeling. Alternatively, if you want to relate
> manufacturing to product development and marketing use SD.
>
> Jack Pugh
> From: JandJpugh@aol.com

For most cases, that is true. I always wonder about a case I rarely see
described here (although it may be in some of the articles referenced
here). The key pieces of an SD approach seem to be (overly simplified) the
notion of feedback and the notion of temporal state (or, alternatively, the
notion of stocks and flows, differential equations with time as the
independent variable, etc.). There doesnt seem to be anything that says
SD wouldnt work with feedback applied to discrete systems (difference
equations instead of differential equations, etc.) While many DES
environments may not facilitate the easy creation of such formulations,
especially including the feedback of information to control earlier flows,
some of the more powerful colored Petri Net formalisms seem to offer
promise. For certain types of problems which have attributes of discrete
systems _and_ feedback, that may be a more direct approach.

Admittedly, a continuous approximation often works quite well to indicate
the dynamics one will experience, but, much as some feel SD is more
accepted (or acceptable) when a simulation can exactly reproduce historical
results because it lends credibility, perhaps modeling discrete problems
with discrete simulations but using feedback and stocks and flows would be
useful in getting results and acceptance.

Comments?

Bill
--
Bill Harris Year 2000 Program Office
mailto: bill_harris@am.exch.hp.com Hewlett-Packard Company
phone: (425) 335-2200 M/S 330C
fax: (425) 335-2483 8600 Soper Hill Road
web: http://hpweb.lsid.hp.com:8080/~billh/ Everett, WA 98205-1298
cupertino: (408) 447-0452 (no voicemail)

[JandJpugh@aol.com]

Comments in reply to Bill Harriss comments!

SD and DES are quite different paradigms although both involve feedback. SD
builds models using ordinary (not partial) differential equations; DES builds
models with difference equations that generally (always?) incorporate
probabilistic behavior. Both methods of necessity incorporate stocks, states,
levels (call them what you will) and flows (physical and information).
Combining the techniques is not necessary to achieve this.

To build an SD model one must carefully identify the system boundaries and
include the significant subsystems; DES models require a detailed
understanding at a very low level of how the system works. Frequently, the DES
model is much larger and requires more computer time to simulate than an SD
model that probably incorporates much more of the total system but in FAR less
detail. As I originally said, SD is good at some problems; DES is good at
others. Comparing them by attacking one problem with both techniques will only
prove which technique was best for that problem (given a quality job of model
construction) while throwing little light on which technique is better.

Reproducing historical results builds confidence, but is NOT a guarantee of a
good model. I know of one instance where a model reproduced historic data
well, but omitted a change in the system that was occurring as the model was
being built. Fortunately, the omission was repaired before the model was used
for policy design.

Jack Pugh
From: JandJpugh@aol.com

["Eric Melse"]

Dear Bill Harris,

Although at this stage I cannot submit a more substantial contribution
to the discussion, I certainly can confirm that your suggestion holds
an important promise to innovate the Systems approach to real world problems
people have to face (and in my field of research, that is the business
environment).

Personally, after following the debate ref. SD vs. DES, or, discrete vs.
continuous simulation,
I remain somewhat puzzled about the difference people seem to make between
these two approaches.
Indeed, technically they are marketed very differently (especially by the
software tool
companies and there application). But, I do not think they are conceptually,
mathematically or
even technically very different. Both approaches are discrete state space
simulation environments
anyway (even Prof. Forrester explains that in his Industrial Dynamics, p.
74). There is no such
thing as a continuous simulation in a digital computer. Continuity itself is
simulated! For those
who would like to hear a proof of that: play a digital music compact disc...

It is the selection of the number of solution intervals (i.e. time steps)
and the way model equations
are designed to operate that make the technical & functional distinction
(and, I think, it is here
where the fundamental difference is between Forresters System Dynamics
approach/methodology and the
rest of the cybernetic academic community, see for example, the already
mentioned: M. Pidd Computer
simulation in management science, ISBN 0471934623, p. 252).

Except for the option to be able to track the state of a discrete object
through the model, which is
in my opinion nothing more than an implementation of an object-based
algorithm within the state space
system itself-and a necessary requirement for discrete optimization models
for problems like queue
simulation-discrete simulators are just as continuous as the continuous ones
are discrete.
Colored Petri Nets are indeed one of the approaches that make that point (a
bit) more clear.

Having said that, working from either end of the modeling spectrum indeed
leads to different
methodological, conceptual and functional approaches. And I wholeheartedly
agree that the problem
a modeler wants to address will drive the solution to either end.

>From my work with SD-tools like Powersim and Vensim, it is very clear to me
that your suggestion:
discrete simulations but using feedback and stocks and flows would be
useful
indeed confirms that there is a new approach on the horizon to getting
results and acceptance;
especially in the business world.

It is that area where the focus is of my current Ph.D. research.

With best regards,
Eric Melse
From: "Eric Melse" <emelse@casema.net>

Drs E. Melse MBA
Van Vredenburchweg 30-B
2282SJ RIJSWIJK
The Netherlands

home phone +31 (0) 70 414 07 67
home fax +31 (0) 70 414 07 56
mobile +31 (0) 6 532 779 33