DISCRETE SIMULATIONS PLUS SYSTEM DYNAMICS

[Jose Antonio Dominguez Machuca <]

Did someone work mixing discrete-event simulation (eg: ARENA software) with
System Dynamics? or do someone know some work about it? If the answer is
yes, was it on the Production/Operations Management field?

Best regards. Jose Antonio Dominguez Machuca

******************************************
Dr. Jose Antonio Dominguez Machuca
Universidad de Sevilla
Facultad de Ciencias Economicas y Empresariales
Departamento de Economia Financiera y Direccion de Operaciones
Avda. Ramon y Cajal, 1
41018 - Sevilla
ESPAÑA
E-mail: jmachuca@cica.es
Tlfn(s): 95 4557627-95 4557610
Fax: 95 4557570
For foreign countries:
Phones: 34954557627-34954557610
Fax: 34954557570
*****************************************

["geoff coyle"]

Jose,

Yes, of course, this has been done but not by mixing two different types of
software, it has been done perfectly well in DYNAMO format languages. It has
been done in several different domains, not just production.

References (some of them) are:

Modelling the discrete ordering of hydro-electric projects RGC and J Rego,
DYNAMICA, Vol 8, Part 1, 1982, pp 36-48.

Representing discrete events in system dynamics models, RGC, Journal of the
Operational Research Society, Vol 36, No 4, pp 307-318, 1985

A system dynamics model of aircraft carrier survivability RGC, System
Dynamics Review, Vol 8, No 3, Fall 1992.

The techniques, and more examples, are described in my book, System Dynamics
Modelling: A Practical Approach, Chapman and Hall/CRC Press, 1996, ISBN 0
412 61710 2. No, its not on Amazon for some weird reason but it is on
www.crcpress.com or from orders@crcpress.com.

Hope that helps.

Regards,

Geoff

geoff.coyle@btinternet.com
Professor Geoff Coyle
Consultant in System Dynamics and Strategic Analysis
Tel: (44) 01793 782817 Fax: 01793 783188

[ularoch@ibm.net]

yes i did,
pls refer to utrecht SDC 92 paper titled
system dynamic analysis, a prototyping tool for production system design

// yours sincerely ulrich la roche
From: ularoch@ibm.net
fast focus consulting
heilighuesli 18, CH-8053 Zuerich,
switzerland
fax +411 382 1349

["Phil Odence"]

>Did someone work mixing discrete-event simulation (eg: ARENA software) with
>System Dynamics? or do someone know some work about it? If the answer is
>yes, was it on the Production/Operations Management field?

The ithink and STELLA software have some limited discrete funtionality. It
tends to be used in doing detailed process modeling (producting/OR kind of
stuff), and can be freely mixed with continuous simulation structures.

You might try downloading an ithink demo from http://www.hps-inc.com and
looking at the emergency room planning model and the manufacturing model.

From: "Phil Odence" <podence@hps-inc.com>

["Bob Walker"]

Jose and others: I suggest you look at the excellent article by Jack Homer
in the most recent (Vol 15, No. 2) issue of SDR. "Macro- and micro-modeling
of field service dynamics". In this work Jack used a combination of
traditional continuous flow modeling techniques and a subsidiary "queuing"
or discrete-event model which helped to bring out some valuable learning for
the team. This application certainly seems to have the operational flavor
you are looking for.

The ARENA product, from Systems Modeling Corp is perhaps the newest but far
from the only discrete-event modeling package. In my look at it I have seen
no obvious way to interface what it does to (in my view) the far more
powerful SD tools such as Vensim. Perhaps through a spreadsheet with DDE
etc. but anyone who has tried this will testify to how messy it can get.

You might also look at the discrete event capabilities in the Stella/iThink
products.

A caution however; it is very easy to get trapped by the "apparent" richness
of discrete event simulation technologies in dealing with the myriad of
details which always surface in a typical modeling project. Such details
seldom enrich the reliability , confidence or robustness that is possible
through a really good continuous dynamic simulation. The usual difficulty is
in convincing the client that this hypothesis is true.

Bob

Robert J. Walker
Condor Associates
Business Dynamics for Management
24 Condor Drive
Ottawa, ON, K1V 9B9
CANADA
Bus +1 613-733-7084
Fax +1 613-733-0883
condora@home.com

[Wayne Wakeland]

Arenas VBA interface would allow its powerful discrete functionality to be
linked with any SD package that also provides a VBA interface (I dont know
if any do).

As others have mentioned, STELLA provides some discrete capability that
allows the construction of combined models. In my experience, while the
use of some discrete logic often improves user acceptance due to better
apparent realism (certainly an important benefit), the discrete logic
rarely changes the fundamental behavior pattern of an inherently
feedback-driven system. My students are always drawn to the discrete
constructs despite my insistence that they avoid them initially. Some
years ago, I challenged my students to replace all their discrete logic
with [roughly] equivalent (and simpler) continuous logic. To their
amazement, in not one case did it have an appreciable impact on the
fundamental behavior PATTERN of their model.

Am I saying that discrete logic is not useful? Not at all!! What I am
saying is that in those cases where the behavior of the system under study
is predominantly FEEDBACK-driven, the incorporation of discrete logic tends
to serve more as a fine-tuning mechanism.

On the other hand, for systems where the fundamental driving concern is the
UNCERTAINTY in the various processes in the system, I believe that a
discrete formulation is essential. For relatively simple processes, I have
found STELLA to work reasonably well. However, it does not allow one to
model and track individual entities with multiple attributes; and its
statistical data collection and analysis capabilities are limited. Hence
the need for discrete system simulation packages such as Arena, SLAM,
ProModel, Simscript, GPSS, and others. I do not agree with Bob Walkers
assertion that SD packages are inherently "far more powerful" than DSS
packages. I believe that both are very powerful and useful...but for
different problem situations.

In situations where feedback AND uncertainty both play a major role in
determining the behavior of the system, one would want a simulation package
that does both equally well. Of the packages I am familiar with, the one
that comes the closest to doing this is Extend from Imagine That, Inc.
But, at a price. In my opinion, Extend is not nearly as good for
visualizing and modeling pure feedback systems as the other SD packages,
nor is it as good for pure discrete system simulation as the best of the
DSS packages. The other problem is that although the basic module
libraries provided with Extend do allow one to construct complex, combined
discrete-continuous models, the execution speed is very slow (one of our
graduate students is addressing this second issue as part of his
dissertation research).

I have room in my toolkit for a variety of tools: a good SD package
(STELLA, Vensim, or Powersim), a good general-purpose DSS package (Arena is
my choice), a good process-oriented package (Extend or ProcessModel), and a
manufacturing-oriented DSS package (ProModel or Arena).

Wayne Wakeland
Adjunct Professor, Portland State University
Portland, Oregon
wwakeland@uswest.net