linking flow to flow directly?

["Richard G. Dudley"]

I appreciate all the suggestions and comments and suggested sources of
modeling ideas.

regarding:
> "Try linking the inflow
> directly to the outflow
> (bypassing the stock) ...

It seems that if we link inflow to outflow then the stock becomes
meaningless. The actual value of the stock can be anything, minus
values included, and has no effect on remainder of the model! Therefore,
it might be better to just leave out that stock, and have the "inflow"
go directly to a delayed inflow of the next stock.


ford stone asked:
> is this an actual (i.e., physical) shrimp farm you are
> analysing?

This is a natural shrimp population on the south coast of Java near
Cilacap. The ocean fishery is dominated by 5 to 7 species of about 15,
but one important species in paticular uses an endangered large coastal
lagoon as a nursery area. This species is caught both in the nursery
area as small juveniles and in the ocean as adults (much more
valuable). This is a common problem with shrimp fisheries in developing
countries. We are interested in looking at the effects on ocean catches
of limiting the lagoon fishing. We have only very limited information on
the actual mortality rates.

If there are no reproductive pulses, and "new juviniles arriving" is
constant over time then a two stock model is sufficient (shrimp in
lagoon) (Shrimp in ocean) with both natural and fishing mortality
draining both. For a general idea as to what is going on this is
sufficient.

If there are pulses in "new juveniles arriving" then if time in lagoon
is correct pulses in the ocean occur too soon in relation to the lagoon
pulse. A simple way to handle this is to have "shrimp leaving lagoon" go
to a cloud and then connect to "shimp moving into the ocean" as a fixed
delay. This allows the shrimp to be in the lagoon long enough for the
mortality rates to have the correct effect and has them arriving in the
ocean at the right time.

However, during the delay time the shrimp are essentially in
hyperspace... they have already spent three months in the lagoon. This
solution only works if there are no month-dependent changes in the
mortality rates. That is there may be specific times when mortality
rates are higher or lower (fishing or mortality patterns related to
weather for example).

Otherwise the best approach is to make a more detailed model of the
lagoon stocks using multiple stocks of shorter duration, though in this
case the ocean pulses appear too attenuated.

--
Richard G. Dudley
rdudley@indo.net.id

http://home.indo.net.id/~rdudley
http://ourworld.compuserve.com/homepages/drrdudley

["geoff coyle"]

Id really be careful about making the model more detailed as a substitute
for making it smart. In my philosophy of SD (borne out by a certain amount
of experience) there are sophisticated models and there are large models but
there are no large sophisticated models. World Dynamics is only about 100
equations!

I bet this brings a storm of protest!

Regards,

Geoff

Professor R G Coyle,
Consultant in System Dynamics and Strategic Modelling,
Telephone +44 (0) 1793 782817, Fax ... 783188
email geoff.coyle@btinternet.com


----- Original Message -----

[Ford Stone]

X-UIDL: 10366-938505322
Status: O
X-Status:


Greetings:

Bill Braun wrote:

"Try linking the inflow
directly to the outflow
(bypassing the stock) and use
the delay as such:

flow: shrimp entering lagoon =
some input
flow: shrimp leaving lagoon =
delay(shrimp entering lagoon,
time in lagoon)"


Perhaps Im missing something here, but
isnt it a violation of the fundamental
concepts of the System Dynamics approach
to do this? Of course, algebra lets one
do anything, -- but should one? I
thought that the whole thing about the
System Dynamics way is to foster clear
and explicit thinking about the system
one is attempting to model. If the
dynamics of the situation and/or the
needs of the modelmaker require a more
complicated model, then one needs the
additional structure.

Why not have additional drains from the
lagoon stock representing predation,
human and otherwise, and other kinds of
mortality?

Of course, pragmatism usually gets the
job done, theres no denying that.

Ford Stone

[Tom Fiddaman]

X-Mozilla-Status: 8011
X-Mozilla-Status2: 00000000
X-UIDL: 10440-938505322
Status: O
X-Status:


> "Try linking the inflow
> directly to the outflow
> (bypassing the stock) ...

> Perhaps Im missing something here, but
> isnt it a violation of the fundamental
> concepts of the System Dynamics approach
> to do this?

This idea is so important that Forrester mentions it three times in
Industrial Dynamics, e.g. "Rates are not determined by other rates. This is
always true in principle."
"Rate equations in principle depend only on the values of levels...."
The motivation for this is that its not possible to measure or perceive a
rate directly, without some intervening integration.

Forrester also notes that it is sometimes convenient to use rate-to-rate
connections: "In practice, _rates_ from the last time interval ... may
occasionally be used as a short cut in place of the _level_ of an _average
rate_ where a very short averaging time would have been used."

In fact rate-to-rate connection isnt the issue in the shrimp case. The
various delay functions that one might use in the outflow equation actually
contain internal levels. Vensim delay_n and ithink smthn have n stocks. An
ithink conveyor or Vensim delay_material is an aging chain of (delay
length)/(time step) stocks.

The remaining problem, which is still fairly fundamental, is that its hard
to create a structure that gets the job done technically and has a diagram
thats clear and sensible to the user, particularly if complications like
predation are included. My hope would be that a little experimentation with
3rd vs. higher order delays would demonstrate that the delay order wasnt
policy relevant, so that you could have your cake (reasonable technical
representation of the real system) and eat it too (clear diagram with
low-order explicit aging chain).

Tom

****************************************************
Thomas Fiddaman, Ph.D.
Ventana Systems http://www.vensim.com
8105 SE Nelson Road Tel (253) 851-0124
Olalla, WA 98359 Fax (253) 851-0125
Tom@Vensim.com http://home.earthlink.net/~tomfid
****************************************************

[Tom Fiddaman]

> "Try linking the inflow
> directly to the outflow
> (bypassing the stock) ...

> Perhaps Im missing something here, but
> isnt it a violation of the fundamental
> concepts of the System Dynamics approach
> to do this?

This idea is so important that Forrester mentions it three times in
Industrial Dynamics, e.g. "Rates are not determined by other rates. This is
always true in principle."
"Rate equations in principle depend only on the values of levels...."
The motivation for this is that its not possible to measure or perceive a
rate directly, without some intervening integration.

Forrester also notes that it is sometimes convenient to use rate-to-rate
connections: "In practice, _rates_ from the last time interval ... may
occasionally be used as a short cut in place of the _level_ of an _average
rate_ where a very short averaging time would have been used."

In fact rate-to-rate connection isnt the issue in the shrimp case. The
various delay functions that one might use in the outflow equation actually
contain internal levels. Vensim delay_n and ithink smthn have n stocks. An
ithink conveyor or Vensim delay_material is an aging chain of (delay
length)/(time step) stocks.

The remaining problem, which is still fairly fundamental, is that its hard
to create a structure that gets the job done technically and has a diagram
thats clear and sensible to the user, particularly if complications like
predation are included. My hope would be that a little experimentation with
3rd vs. higher order delays would demonstrate that the delay order wasnt
policy relevant, so that you could have your cake (reasonable technical
representation of the real system) and eat it too (clear diagram with
low-order explicit aging chain).

Tom

****************************************************
Thomas Fiddaman, Ph.D.
Ventana Systems http://www.vensim.com
8105 SE Nelson Road Tel (253) 851-0124
Olalla, WA 98359 Fax (253) 851-0125
Tom@Vensim.com http://home.earthlink.net/~tomfid
****************************************************

[Bill Braun]

X-UIDL: 10441-938505322
Status: O
X-Status:

> Bill Braun wrote:
>
> "Try linking the inflow
> directly to the outflow
> (bypassing the stock) and use
> the delay as such:
>
> flow: shrimp entering lagoon =
> some input
> flow: shrimp leaving lagoon =
> delay(shrimp entering lagoon,
> time in lagoon)"
>
>Ford Stone wrote:
>
> Perhaps Im missing something here, but
> isnt it a violation of the fundamental
> concepts of the System Dynamics approach
> to do this? Of course, algebra lets one
> do anything, -- but should one? I
> thought that the whole thing about the
> System Dynamics way is to foster clear
> and explicit thinking about the system
> one is attempting to model. If the
> dynamics of the situation and/or the
> needs of the modelmaker require a more
> complicated model, then one needs the
> additional structure.

Many thanks for the question and the challenge. My rationale may be based
on a misunderstanding. Ive always understood a delay function to be an
integrating function of sorts. As I look at a number of DYNAMO models and
the user manual, the documentation (as I understand it) treats delays,
averaging and smooting as integrating functions; all of the equations
containing these functions are labeled as levels.

I have never used DYNAMO but find their manual quite useful for clarifying
certain basics of modeling. I also rely on a number of other books based on
DYNAMO and which treat broader issues of modeling. (Richardson and Pughs
Introduction to SD Modeling with Dynamo is priceless.)

I may simply be dead wrong here. If so, I welcome the correction. Id also
welcome as broad an explanation (or cites to some reading) that would help
me see the mistake.

In the suggestion I gave, the out flow is admittedly not using the level
"shrimp in lagoon" to determine its behavior. As an integrating function in
its own right, the net between inflows and outflows, which would be found
in the level, would be an accurate representation of the "shrimp in the
lagoon".

Just thinking about it now, I can see a downside to this. If there are
other outflows from the lagoon (level) that would affect the "shrimp
leaving the lagoon" equation, my suggestion would produce an erroneous
picture of the outflow behavior.

Thanks again for pointing this out, Ford.

Bill Braun
medprac@hlthsys.com

[Bill Braun]

> Bill Braun wrote:
>
> "Try linking the inflow
> directly to the outflow
> (bypassing the stock) and use
> the delay as such:
>
> flow: shrimp entering lagoon =
> some input
> flow: shrimp leaving lagoon =
> delay(shrimp entering lagoon,
> time in lagoon)"
>
>Ford Stone wrote:
>
> Perhaps Im missing something here, but
> isnt it a violation of the fundamental
> concepts of the System Dynamics approach
> to do this? Of course, algebra lets one
> do anything, -- but should one? I
> thought that the whole thing about the
> System Dynamics way is to foster clear
> and explicit thinking about the system
> one is attempting to model. If the
> dynamics of the situation and/or the
> needs of the modelmaker require a more
> complicated model, then one needs the
> additional structure.

Many thanks for the question and the challenge. My rationale may be based
on a misunderstanding. Ive always understood a delay function to be an
integrating function of sorts. As I look at a number of DYNAMO models and
the user manual, the documentation (as I understand it) treats delays,
averaging and smooting as integrating functions; all of the equations
containing these functions are labeled as levels.

I have never used DYNAMO but find their manual quite useful for clarifying
certain basics of modeling. I also rely on a number of other books based on
DYNAMO and which treat broader issues of modeling. (Richardson and Pughs
Introduction to SD Modeling with Dynamo is priceless.)

I may simply be dead wrong here. If so, I welcome the correction. Id also
welcome as broad an explanation (or cites to some reading) that would help
me see the mistake.

In the suggestion I gave, the out flow is admittedly not using the level
"shrimp in lagoon" to determine its behavior. As an integrating function in
its own right, the net between inflows and outflows, which would be found
in the level, would be an accurate representation of the "shrimp in the
lagoon".

Just thinking about it now, I can see a downside to this. If there are
other outflows from the lagoon (level) that would affect the "shrimp
leaving the lagoon" equation, my suggestion would produce an erroneous
picture of the outflow behavior.

Thanks again for pointing this out, Ford.

Bill Braun
medprac@hlthsys.com

[Ford Stone]

X-UIDL: 10527-938505322
Status: O
X-Status:

Bill Braun wrote:

>
> Many thanks for the question and the challenge. My rationale may be based
> on a misunderstanding.

No, I think its my mistake; I just choked on the language
> "Try linking the inflow
> directly to the outflow
> (bypassing the stock)"
and went into overdrive. You were certainly gracious in your response, for
which I thank you.

Bye the bye, is this an actual (i.e., physical) shrimp farm you are
analysing?

ford

[Bill Braun]

At 10:13 AM 6/24/00 +0100, Geoff Coyle:
>Id really be careful about making the model more detailed as a substitute
>for making it smart.

At my stage of learning modeling this embodies the challenge I face. How to
intuit the variables that best describe the structure of the system without
adding variables that may either be a) unnecessary or b) downright
misleading or wrong.

As Ive mentioned on other occasions Ive learned quite a bit from studying
Forresters Urban Dynamics. It weighs in at 150 equations or so, although
that doesnt lead me to disagree with your premice.

Bill Braun
From: Bill Braun <medprac@hlthsys.com>

[Alexander Leus]

I am in agreement with you Professor Coyle. I am an ex control system
engineer, it is amazing what a process model of a single time delay and
a 2nd order differential equation can do for you.

If I am understanding your comments Professor Coyle, a time delay
process with proportional control only will always give you an offset
from your set point or it leads to stability problems as you decrease
the proportional band or increase the gain of the controller. One needs
to consider integral control action and possibly a little rate control.

Getting away from basic control theory, has anyone considered the
personality dynamics of the human controllers. I am referring to work
done by Sandra Seagal and David Horne, refer to humandynamics.com ?


Sincerely,
Alex Leus
From: Alexander Leus <leusa@tds.net>