**Subject: **Re: Fuzzy Logic vs. System Dynamics

**From: **Bradd Libby (*bradd@WPI.EDU*)

**Date: **Sun Sep 03 2000 - 06:35:43 MET DST

**sorted by:**[ date ] [ thread ] [ subject ] [ author ]**Next message:**IOANNIS DOKAS: "membership functions"**Previous message:**nhnfix@research-sites.com: "Publish your research papers here FREE 3601"**Maybe in reply to:**Bradd Libby: "Fuzzy Logic vs. System Dynamics"**Maybe reply:**Bradd Libby: "Re: Fuzzy Logic vs. System Dynamics"

================================================

Bradd Libby, bradd@wpi.edu, writes:

*>
*

*> For Fuzzy Logic, the applications seem to be more 'engineering'
*

*> oriented (control of a cement kiln, a subway system, an inverted
*

*> pendulum)...
*

William Siler, WSiler@aol.com, replies:

*>
*

*> Alas, altogether too many people, including many of the fuzzy systems
*

*> crew, have this idea, although it is usually not put forward so
*

*> explicitly. Earl Cox's work in the business field (see one of his
*

*> several books) is certainly non-engineering, and our own recent work has
*

*> been in decision making in a hospital intensive care unit. Unfortunately
*

*> the success of fuzzy systems theory in process control has led to a
*

*> preoccupation with this field to the virtual exclusion of work in
*

*> non-control applications.
*

I hope that I didn't imply that 'engineering problems' were the exclusive

domain of Fuzzy Logic and 'business problems' the domain of System

Dynamics. My main point was that Fuzzy Logic's historical sucesses (to

date) have been 'hard' control problems, while System Dynamic's successes

have been in the area of 'soft' systems analysis (An important sidenote,

though, is that Forrester orignally dubbed System Dynamics 'Industrial

Dynamics,' as he saw its value primarily in areas like supply chain

management, inventory control, logistics/operations research, and the

like, but changed to 'System Dynamics' when he realized, with the help of

a former mayor of Boston, that it could be applied far beyond the

factory.). Indeed, based on the close parallels I've observed between the

two disciplines, it seems to me that any system that can be modelled with

System Dynamics can be handled by Fuzzy Logic as well, and vice-versa

(which of the two is 'more appropriate' for a given problem, easier to

apply, or more intuitive is most likely at the discretion of the

practitioner), which leads me to wonder why we consider them to be

separate disciplines.

I think perhaps a distinction between the two disciplines more interesting

than the 'engineering' vs. 'business' applications is my observation that,

and perhaps this is wrong, System Dynamics is typically employed to

_analyze_ or _understand_ 'problematic' behavior in complex systems (in

fact, some system dynamicists have seriously considered re-renaming their

field 'Problem Dynamics'), while Fuzzy Logic is typically employed to

_control_ or _correct_ problematic behavior. Perhaps the distinction is

academic (Why build understanding if not to correct the problem...and

conversely, how can you correct the problem if you don't understand it?),

but nevertheless, the chief practitioners in these respective fields seem

to have "sold" their disciplines to the Unacquainted in these terms.

It's interesting to hear that you are applying Fuzzy Logic to medical

care, as one of the early applications of System Dynamics to business

simulation was a multiplayer board game called 'Friday Night at the E.R.'

(www.blearning.com) - and this is certainly not the only application of

System Dyanmics to this field.

========================================

Bradd Libby, bradd@wpi.edu, writes:

*> > For Fuzzy Logic the operative maxims are a
*

*> > rejection of IF-THEN/EITHER-OR/BLACK-WHITE in favor of probability
*

*> > distributions and continuous functions;
*

Bob Briggs, zunt@aol.com, replies:

*>
*

*> A couple of quick points off the top of my head...your observation above
*

*> isn't entirely accurate. Fuzzy logicians don't reject if-then (if one does,
*

*> I'd like to hear the argument). FL applications need an inference engine,
*

*> and I use if-then rules all the time as I set up relationships between fuzzy
*

*> variables without feeling polluted. Better to say that fuzzy logic rejects
*

*> bivalence.
*

I agree and disagree (How's that for bivalence for ya'?):

Yes, I agree that my selection of terms was sloppy - I think that just

saying 'BLACK-WHITE' would have conveyed what I intended. (Perhaps

tossing the words 'Boolean' or 'binary' in there somewhere would have

helped as well.) The IF-THEN I had in mind was more along the lines of

absolute causality (A drinking glass falling of the table is a problem

(i.e, IF (position-of-glass) is off the table, THEN problem exists), and

one standing in the center is not a problem (i.e, IF (position-of-glass)

is not off the table, THEN problem does not exist). Boolean logic states

that a glass 90% of the way to the edge is also not a problem (i.e., IF

(position-of-glass) is near edge of table, THEN problem does not exist),

while Fuzzy Logic rejects this sharp distinction.

And No, I disagree that 'Fuzzy logicians don't reject if-then' - I think

that they do (and furthermore, that for most applications, it's good that

they do, and they should continue to do so). I realize that fuzzy

relationships are typically phrased in IF-THEN terms (e.g., 'IF

(position-of-glass) is (far-from-center-of-table) THEN

(push-glass-hard-to-center-of-table)'), but when one rejects Boolean

values (0 or 1) in favor of a continuum of values between 0 and 1,

inclusive, then any IF-THEN statements that are generated containing fuzzy

variables become mathematically equivalent to normalized (and, often,

non-linear) functions.

In other words, there is no difference between my previous example:

IF (position-of-glass) is (far-from-the-center-of-the-table) THEN

(push-glass-hard-to-center-of-table)')

and,

F = q * Dx

where F is the amount of force to be applied in the direction of the

center of the table, Dx is the distance that separates the glass from the

center, and 'q' is a proportionality constant. (I realize my engineering

background is showing in my terminology, but basically this relation says:

'As the position of the glass gets further from the center of the table,

increase the force that's pushing in the direction of the center.') The

two statements are conceptually identical.

If 'q' is a variable (perhaps one that depends on Dx), then the force

curve could be made to assume any desired (linear or nonlinear) shape -

even a triangle. (Preferably, this expression would be such that the force

appraches infinity as the glass approaches the edge, thus ensuring that

the glass would not fall off the table (assuming that the glass doesn't

shatter!).)

So, while fuzzy logicians may retain the IF-THEN *terminology* in applying

Fuzzy Logic, they are actually applying continuous functions, and thereby

rejecting the 'IF A exists/happens, THEN B exists/happens' mindset.

*> I'd be interested to know more about the sets that are used under the system
*

*> dynamics umbrella. Forrester invented magnetic core RAM. Pretty bivalent
*

*> stuff, eh? Does that same bivalence carry through to the math? (This not
*

*> asked very well, but I hope the folks on this list understand what I'm trying
*

*> to convey.)
*

The System Dynamics language (the 'stock and flow' language) has no formal

means of representing IF-THEN statements, although all of the simulation

software packages available, to the best of my knowledge, have added on

the ability to use IF, THEN, OR, NOT, ELSE operators. System dynamicists

tend to disfavor the use of these operators, reserving them only for when

it is much more convenient to use them than to not use them. (In fact,

'Continuous Thinking,' a preference for continuous functions over absolute

causality IF-THEN statements is another of the 7 Systems Thinking skills

identified by Dr. Barry Richmond (who received his Ph.D. in System

Dynamics under Dr. Forrester at MIT).

The reasons for rejecting Boolean logic in System Dynamics models are the

same as for rejecting them in Fuzzy Logic models - they can cause

erroneous results and undesired behavior. Case in point: H. Van Dyke

Parunak, whose work on agent-based systems was mentioned in the cover

article, called 'Ant-based Optimatization' (or something similar), of the

March 2000 issue of Scientific American, and Rick Riolo, an ithink/STELLA

user and professor at U Michigan who is a member of the BACH group (whose

members include John Holland (the 'H'), the inventor of evolutionary

algorithms, and Robert Axelrod (the 'A'), the author of _The Evolution of

Cooperation_ and _The Complexity of Cooperation_), co-authored a project

called DASCh (Dynamic Analysis of Supply Chains), some of the

papers/presentations for which can be found on the web at

www.erim.org/cec/sce.htm and www.erim.org/cec/projects/dasch.htm.

In their project, Dr.s Parunak and Riolo modelled a production-

distribution system and found interesting (read: chaotic) behavior in a

very simple mathematical model. The model is detailed in their reports

"Dynamic Analysis of Supply Chains" and "Agent-Based Modeling vs.

Equation-Based Modeling", which are available at the second URL listed

above. I found it interesting that chaotic behavior was seen in such a

simple model, but was dismayed to discover that the cause of the behavior

was an unruly IF-THEN statement of the form:

IF (x is less than some critical value called x#) THEN (y = 0) ELSE (y =

some large positive value called N)

The problem was that x depended on y, so that if x evaluated above the

critical value, then y became equal to N, causing x to fall, causing y to

become equal to 0, causing x to rise above the critical value again. In

other words, the chaotic behavior results from the discontinuity in the

equation for y, and subsequently the behavior of the system depends on the

time increment selected for the simulating the system over time (a

hallmark that the behavior is an artifact of the formulation of the

equations, rather than true chaos).

When this IF-THEN statement is replaced by a 'fuzzy function' of

the form:

IF (x is less than some critical value called x#) THEN (y = (x/x#)*N) ELSE

(y = N)

(In other words, y is 90% of N when x is 90% of x#, and y is N when x is

greater than or equal to x#), when this substitution is made, all of the

chaotic behavior of the system disappears, and most of the analysis based

on the supposed chaotic behavior is invalidated.

Thus the dangers of Boolean logic and of rejecting Continuous Thinking.

Best regards,

Bradd Libby.

############################################################################

This message was posted through the fuzzy mailing list.

(1) To subscribe to this mailing list, send a message body of

"SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at

(2) To unsubscribe from this mailing list, send a message body of

"UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"

to listproc@dbai.tuwien.ac.at

(3) To reach the human who maintains the list, send mail to

fuzzy-owner@dbai.tuwien.ac.at

(4) WWW access and other information on Fuzzy Sets and Logic see

http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info

(5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html

**Next message:**IOANNIS DOKAS: "membership functions"**Previous message:**nhnfix@research-sites.com: "Publish your research papers here FREE 3601"**Maybe in reply to:**Bradd Libby: "Fuzzy Logic vs. System Dynamics"**Maybe reply:**Bradd Libby: "Re: Fuzzy Logic vs. System Dynamics"

*
This archive was generated by hypermail 2b25
: Sun Sep 03 2000 - 06:52:15 MET DST
*