Is SD better than DES for Manufacturing Simulation

[Bruce Campbell]

Check out http://www.imaginethatinc.com/
for an easily understood comparison, and differences, between DES and
SD. They are used for different types of problems.

There was also a discussion on this list some time ago about the
differences between SD and DES. Let me know if you cant get hold of the
digest of this particular thread as I kept a record of the discussion.

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

[A.Oyarbide-Zubillaga.1998@Cranfi]

Hello everyone.

I am a posgraduate student in Cranfield University (UK). At this
moment, I am doing a thesis, in which I am applying the System
Dynamics methodology to elaborate a simulation program in visual basic
to simulate an automotive asembly line.

I have read some usefull papers about SD (Road maps, etc.), but still
now, my biggest problem is the SD concept.

In my opinion, Discrete Event Simulation is good, and there are many
commercial packages oriented to simulate manufacturing systems. I
dont know really which are the main advantages to simulate
manufacturing systems using SD.

I know that many people says that SD biggest advantage is the
simplicity and that it needs less resources, but with actual
computers, is it an advantage? In which areas can SD be better than
DES in manufacturing simulation?

I dont know if the question is clear, but I would appreciate any
answer which helps me to solve this doubt.

Thanks

___________________________________________
Aitor Oyarbide (Cranfield University)
A.Oyarbide-Zubillaga.1998@cranfield.ac.uk

[Stephen Shervais]

Aitor
The answer to this is the standard systems answer..."it depends."

In simulations coursework here at PSU the students are encouraged to try
both approaches, to demonstrate that overall systems dynamics can be
portrayed equally well by either SD or DES.

Therefore, your choice depends on your purpose. SD treats the membership
of each flow/stock as a "molecule" -- everyone is the same and you dont
care what happens in individual cases, like molecules of water or packets
of money. In DES, the individual matters, and you can collect statistics
on longest/average/least time spent in a queue.

So, in a production line simulation, SD would show you the fluctuations in
the stock called "work in progress" and allow you to identify bottlenecks,
while DES would let you measure "maximum processing time" for a production
item and allow you to determine if you had met your goals for delivering
special orders to customers.

Regards,
Steve

Steve Shervais shervais@acm.org
Graduate Student shervais@sysc.pdx.edu
Systems Science PhD Program psu00872@odin.cc.pdx.edu
Portland State University http://www.sysc.pdx.edu
Portland, Oregon http://www.ee.pdx.edu/~shervais
(503) 725-7344 / 725-4997

["Lukiantseva, Natasha "]

If you are simulating the production itself - assembly line - to evaluate
cycles and productivity, etc. then SD might not be the best choice. There is
a number of packages for manufacturing descrete event simulation (Arena from
System Modeling is one of them).
Sincerely,
Natasha Lukiantseva
From: "Lukiantseva, Natasha " <Natasha_Lukiantseva@csx.com>
~~~~~~~~~~~
RNX 322-4408
Operations Research
CSX Transportation

["Jim Hines"]

Aitor,

Please correct me if Im wrong, but I think that...

There is a difference in practice in that System Dynamics Models are almost
always centrally interested in feedback processes. In contrast discrete
event models are almost always centrally concerned with impact of stochastic
processes on things like though-put, lines and waiting times.

Also in practice, system dynamics models are usually at a managerial level
looking at things like pricing policies or company expansion or hiring and
firing, while discrete event models are often focussed at a "shop-floor"
level on things like machine placement or the progress of packages through a
transport system.

On a technical level, I think that analyzing feedback is easier in SD models
where "events" usually are continuous (i.e. the variables tracking events
are continuous). Discrete event models present simulation results in terms
of statistics (e.g. average throughput); whereas SD models present output as
plots of the continuous curves - a format that is well suited for seeing
behavior patterns and finding the causal loops responsible.

Finally on a model level, I think that discrete-event models are often more
intuitive, because the model actually represents concrete items like an
individual machine , separate packages, or a single person. System dynamics
models use a lot of aggregation (e.g. a number representing the level of
inventory instead of the individual items in the inventory) which SDers
eventually come to see as being, but which I think is actually a bit more
abstract.

On a theoretical level, I dont see any reason that you couldnt use
discrete event tools to examine some (and perhaps most) feedback issues; and
I know that system dynamics tools like iThink are frequently used to model
the impact of stochastic processes on lines and waiting times.

Regards,
Jim Hines JHines@mit.edu