Searching \ for 'Fuzzy Control Tool (was Re: PID Temperature Contro' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/io/sensors.htm?key=temperature
Search entire site for: 'Fuzzy Control Tool (was Re: PID Temperature Contro'.

Truncated match.
PICList Thread
'Fuzzy Control Tool (was Re: PID Temperature Contro'
1997\09\02@162847 by Aydin Yesildirek

flavicon
face
> > I'm investigating using a cheap toaster oven as a SMD reflow oven
> > for prototype boards.
> >
> > The idea is to use a PIC in provide an accurate temperature
> > controller, possibly providing a programmable temperature profile
> > capability.
>
> I just designed a new toaster oven with a PIC at the heart of the
> controller.  The requirement was for cheap, not for accurate, so I
> cut a lot of corners on accuracy.    All the PIC is really doing is
> analog comparing the thermistor voltage to a potentiometer voltage.
> <snip>

I have been following this list for a while and I could not see
anything related Microchip's fuzzy control development systems.

It seems to me for some people use of PID scheme is an exciting news.
However, the truth about it not so great. This classical old control
scheme is design technique based on a linear systems. Most of the
applications are nonlinear. It is true that PID may work for many
industrial systems around an operating point. However, at least in
theory, it may drive the system into unstability out of the region
defined by the PID coefficients.

The thread about temperature control remainded me a presentation
from Texas A&M in which fuzzy control was outperforming PID based
controller in a real application. If you search by the author name
Langari, i think it was 3-4 years ago you may get some papers to look
at it.

I am not sure whether the subject may fit in PIC list or not, but
anyone out there has any experience with microchips fuzzy control
tools?

BTW-they were announced on microchip BBS, on sale, US$99, for a
limited time but I could not find any one to that price. Then
microchip extended the deadline, i think it is still 66% off the
price, but you may not find a dealer that sells it.

__Aydin

1997\09\02@165533 by Miller, Steve

flavicon
face
>The thread about temperature control remainded me a presentation
>from Texas A&M in which fuzzy control was outperforming PID based
>controller in a real application. If you search by the author name
>Langari, i think it was 3-4 years ago you may get some papers to look
>at it.

>__Aydin

As a graduate of Texas A&M (class of 93)  I do not recall any great interest in
Fuzzy Control.  In fact when, I studied control theory at Texas A&M, the
professor was not in favor of Fuzzy Control.  Fuzzy Control was seen as a sales
tool to sell washing machines in Japan.  Fuzzy Control has its place, but it is
not the great solution or replacement for PID control.  Pease Porridge of
Electronic Design ran a contest for a Fuzzy Control system that he could not
duplicate or exceed the performance of.  He won every challenge using op-amps
and classic control theory. Fuzzy Control and Fuzzy Logic has had lots of
boasting and very little real substance.

---- Steve

1997\09\02@180732 by Aydin Yesildirek

flavicon
face
> >The thread about temperature control remainded me a presentation
> >from Texas A&M in which fuzzy control was outperforming PID based
> >controller in a real application. If you search by the author name
> >Langari, i think it was 3-4 years ago you may get some papers to look
> >at it.
>
> >__Aydin
>
> As a graduate of Texas A&M (class of 93)  I do not recall any great interest
in
> Fuzzy Control.  In fact when, I studied control theory at Texas A&M, the
> professor was not in favor of Fuzzy Control.  Fuzzy Control was seen as a
sales
> tool to sell washing machines in Japan.  Fuzzy Control has its place, but it
is
> not the great solution or replacement for PID control.  Pease Porridge of
> Electronic Design ran a contest for a Fuzzy Control system that he could not
> duplicate or exceed the performance of.  He won every challenge using op-amps
> and classic control theory. Fuzzy Control and Fuzzy Logic has had lots of
> boasting and very little real substance.
>
> ---- Steve


If you check ME department you'll find some news about Dr. Langari,
he might have been a student back then.

I am not trying to sell fuzzy control alone as the greatest solution
at all. But I believe it is a very useful tool which can make a
bridge between modern and postmodern, intelligent, etc. control
techniques. In fact as a control engineer it hurts to see people keep
using PID which is not even defined for nonlinear systems. But it has
been used even in flight control with gain scheduling, btw which can
also be viewed as a fuzzy control. So in this group especially for
the hackers there are great adventures with it. You may go places no
one has gone before:)

I have been in control systems society for about 5 years. My
background is in control theory.

So your EE prof's comment on fuzzy control isn't suprising to me. In
1993 IEEE CDC, San Antonio, TX, biggest control conference. Bode award
winner Prof. Athans from MIT is giving his speech. His contributions
in optimal control are greatly valued. In his speech he blasted Japan
and so called intelligent, neuro-fuzzy, control. According to his expert
opinion, especially neurocontrol, was something like a great scam and
must have been kept out of control systems society. Funny but. the
inventer of fuzzy sets, Lotfi Zadeh happens to be his advisor from
Berkeley and for this reason fuzzy portion of it did not get much
critics from him.

He is not alone. Most of the control profs. with math background
don't like such ideas. If you try to prove by a real example it works
they might even get offended.

I had to ask myself why a person like Prof. Athens attacking some new
techniques and trying to stop its development. I didn't know the
answer but what I know as a fact that most of the industrial
applications do not use modern control because of the assumptions made
are too strong and don't fit most of the practical systems. Thus, those
fancy tools don't work. On the other hand, somebody, most of the
times happens to be from japan, comes up with a neurofuzzy control
scheme without any proof, in a tough real life example showing it
working. Open up a robot controller most of the time you will find a
PID derivative type of algorithm. I could not find any nonlinear
based, e.g., adaptive, robot controller in the market. There books
written about it though. IMHO, this may be the reason why some
control people got frustrated.



PS: Last year I saw Prof. Athans name on a neurocontrol paper :)

Regards,


__Aydin

1997\09\02@185737 by John Payson

picon face
>                                Fuzzy Control has its place, but it is
> not the great solution or replacement for PID control.  Pease Porridge of
> Electronic Design ran a contest for a Fuzzy Control system that he could not
> duplicate or exceed the performance of.  He won every challenge using op-amps
> and classic control theory. Fuzzy Control and Fuzzy Logic has had lots of
> boasting and very little real substance.

Fuzzy logic is a bunch of terminology wrapped around some useful concepts.
While PID loops have their uses, they are often not ideal--even with fully
linear systems.

Consider the problem: you are supposed to pilot a hover-tram from point A
to point B as quickly as posslble, with zero velocity at point B.  You will
not be deemed to have arived until you are at point B with zero velocity; if
you overshoot you will have to come back.  The hover-tram is equipped with
a linear isometric control: applying 1N of forward pressure on the lever will
impart 1000N of forward force on the tram; applying 2N of reverse pressure
will impart 2000N of reverse force on the tram, etc.

Using a PID loop, the hovercraft would be able to go moderately quickly from
A to B regardless of the distance between them.  If the loop is properly
tuned to the mass of the tram, the ride will be smooth and the tram will stop
precisely on target.  If the loop is not tuned properly, it will either over-
shoot the target and then oscillate around it (if the tram is heavier than
expected) or will slow down too soon and crawl extremely slowly toward the
target, Zeno's-paradox-style.

By contrast, a faster method is to give the tram maximum safe forward thrust
until some point (about halfway, though friction will shift the proper spot)
and then maximum safe reverse thrust.  If one does the switch from forward to
reverse at just the right moment, the tram will stop precisely at the desired
spot.

Obviously this faster method would be a little dangerous in practice; if the
driver doesn't reverse direction soon enough, the tram may significantly over-
shoot the target while travelling dangerously fast.  On the other hand, it
is faster than the PID method (which can also overshoot, though not as badly).
The goal then is to find a method which is faster than PID but safer than the
alternative above.

One approach is to consider how a person would solve this problem (or, in
real life, how one would stop a car at a stop sign in potentially icy weather).
If the car is going, e.g., 1mph, then it can be stopped easily within 3" of
any desired point just by pressing the brakes 3" before that point.  If the
car is going 3mph, then it can be gotten to 1mph within 5 feet by pressing
the brakes until it reaches 1mph.  Etc.  So the algorithm becomes (the numbers
vary with weather, etc., but the idea should be clear):

[1] If speed > 55mph press brakes (maximum permissible speed)
[2] If speed > 40mph and distance < 700ft press brakes
[3] If speed > 30mph and distance < 400ft press brakes
[4] If speed > 20mph and distance < 150ft press brakes
[5] If speed > 15mph and distance <  75ft press brakes
[6] If speed > 10mph and distance <  25ft press brakes
[7] If speed >  5mph and distance <  15ft press brakes
[8] If speed >  3mph and distance <  10ft press brakes
[9] If speed >  1mph and distance <   5ft press brakes
[10] If speed > 0 and distance < 3" press brakes
[11] If none of the above, release brakes

This algoritm will stop the car at the proper spot every time, regardless of
weather (unless it's really bad).  It won't always be the fastest way to get
there, but the results are much less dependent upon variations in loading,
friction, etc. than with the PID method.  The only major problem with this
method is that it will produce very jerky results: the brakes are always
"on" or "off"; there's no moderation.  This is where fuzzy logic comes in.

If the above system were entered into a fuzzy logic system and the car was
cruising down the road at 54mph while nearing the destination, the fuzzy
logic interpreter would start applying the brakes as the distance approached
700ft.  As the car slowed down, it would start releasing the brakes as the
speed dropped doward 40mph (assuming the distance was still well over 400ft).
Rather than applying and releasing the break suddenly, the system would do
so gradually.  While this "smoothing" would affect the overall system behavior
somewhat, the effects would be fairly predictable and generally not severe
(though rule 10 might need to be adjusted if the fuzzy system stops short or
long, the car would be moving slowly at that point so the stop should be
"safe" even if it's not perfect.

Fuzzy logic is by no means a panacea in motion-control or other applications;
nonetheless, it allows decent and robust solutions to be produced fairly
quickly.

1997\09\02@212759 by Walter Banks

picon face
----------
> > Pease Porridge of
> > Electronic Design ran a contest for a Fuzzy Control system that he
could not
> > duplicate or exceed the performance of.  He won every challenge using
op-amps
> > and classic control theory. Fuzzy Control and Fuzzy Logic has had lots
of
> > boasting and very little real substance.

I was part of the debate with Bob Pease at two or three of the Fuzzy logic
conferences. Bob forced the issue in a nice way by challenging the fuzzy
people to deal in facts. There are two things that fuzzy does well. It
normalizes the parameters of problems very well. It also allows a simple
solution to be implemented for many non linear problems.

When Byte Craft was developing Fuzz-C we looked hard for applications
where fuzzy worked as well or better than conventional PID. When we did
we found that PID was highly optimized for a narrow range of
applications and fuzzy was a more general case.

For example look at the momentum difference in an aircraft control system
after the fuel stored in the wings has been burned in flight. A PID control
loop in such a system is very hard to tune.

Walter Banks
http://www.bytecraft.com

1997\09\03@081150 by Tom Handley

picon face
re: Hover Craft, Cars, PID, and Fuzzy Logic

  John, thanks for the outstanding example. I'm neither a PID or Fuzzy
logic expert but I have been following it for years. I am a pilot and keep
up with various aircraft control and UAV issues. I think you ought to write
a book ;-)

  - Tom

At 05:42 PM 9/2/97 -0500, John Payson wrote:
>>                                Fuzzy Control has its place, but it is
>> not the great solution or replacement for PID control.  Pease Porridge of
>> Electronic Design ran a contest for a Fuzzy Control system that he could not
>> duplicate or exceed the performance of.  He won every challenge using op-amps
>> and classic control theory. Fuzzy Control and Fuzzy Logic has had lots of
>> boasting and very little real substance.

[snip]

1997\09\03@095206 by Aydin Yesildirek

flavicon
face
> > > Pease Porridge of
> > > Electronic Design ran a contest for a Fuzzy Control system that he
> could not
> > > duplicate or exceed the performance of.  He won every challenge using
> op-amps
> > > and classic control theory. Fuzzy Control and Fuzzy Logic has had lots
> of
> > > boasting and very little real substance.

> When ... developing Fuzz-C we looked hard for applications
> where fuzzy worked as well or better than conventional PID. When we did
> we found that PID was highly optimized for a narrow range of
> applications and fuzzy was a more general case.
>
> For example look at the momentum difference in an aircraft control system
> after the fuel stored in the wings has been burned in flight. A PID control
> loop in such a system is very hard to tune.


It may not be fair to compare PID vs. fuzzy. The biggest strength or
advantage you will obtain with a fuzzy logic is the range of
applications, the ability not only to have a competitive performance
where PID work well but also acceptable results in some applications
PID, or any other conventional control methodology, poorly works or
does not work at all.

But it may not be the most precise controller you can come up with
when you have other alternatives. Nevertheles, it is very robust and
simpler to implement.

__Aydin

1997\09\03@124346 by Pioneer Microsystems

flavicon
face
John Payson wrote:

> >                                Fuzzy Control has its place, but it is
> > not the great solution or replacement for PID control.  Pease Porridge of
> > Electronic Design ran a contest for a Fuzzy Control system that he could not
> > duplicate or exceed the performance of.  He won every challenge using
op-amps
> > and classic control theory. Fuzzy Control and Fuzzy Logic has had lots of
> > boasting and very little real substance.
>
> Fuzzy logic is a bunch of terminology wrapped around some useful concepts.
> While PID loops have their uses, they are often not ideal--even with fully
> linear systems.
>

Hmmm... Want to back that up with some facts?  What is 'ideal'?

> And John went on with some interesting examples...

Now, I may leave myself wide open, not knowing fuzzy theory in any detail, but I
see an unfairness in your comapative example.  First off, I don't think that you
will find anyone that will claim that PID is anything more than an algorithm
that
can achieve a setpoint for you.  In no way is it claimed to be a method of
optimal
setpoint profile control.  Were I to rely on PID to simply slam away at the
final
setpoint without any setpoint profile to follow, I am certain that the tram
would
overshoot, or detuned maybe stop a mile short.  And I am sure that subway trains
in
Japan would 'jolt' when they stop.  But be real, any decent system is going to
have
another piece of code that generates a profile of a setpoint that it will
follow.
An elevator, for example, will have a setpoint profile that will optimize
velocity
and acceleration for the user.  If you research PID, you will find a very
extensive
foundation of well proven math (I'll upset the EE crowd.... Root Loci! Complex
plane!  Imaginary numbers!)  I for one have not seen any hard information about
how
fuzzy tackles the actual complex plane portion of the challenge.  The examples
all
seem to ride the issue of setpoint control and skip the stability issue.  I dare
say that with the lack of concepts being shown under the cover, I will asume
that
under the hood is a good old fashioned PID loop to achieve the goal.  As the
used
car dealer says, Nah, boy, you don't havta look unda tha hood.  Just trust me,
ya
hear?

Folks, please don't give me any static about off topic.  I think all embedded
poeple need to know about feedback control.  If you don't, when your boss or
client
asks you about it someday, you'll look pretty dumb saying 'Duhhh.'

Chris Eddy
Pioneer Microsystems, Inc
Pittsburgh, PA
http://www.nb.net/~ceddy

1997\09\03@141539 by William Chops Westfield

face picon face
   Fuzzy Control has its place, but it is not the great solution or
   replacement for PID control.  Pease Porridge of Electronic Design ran
   a contest for a Fuzzy Control system that he could not duplicate or
   exceed the performance of.  He won every challenge using op-amps and
   classic control theory. Fuzzy Control and Fuzzy Logic has had lots of
   boasting and very little real substance.

Of course, this sounds like the old compiler vs assembly coder debate.  If
fuzzy logic allows less skilled people to produce bettter control system,
then it definately has a place.  Not everyone is Bob Pease...

(Not that *I'm* a believer...)

BillW

1997\09\03@170523 by Eric van Es

flavicon
face
Aydin Yesildirek wrote:
{Quote hidden}

Agreed.

The main thing about PID's popularity is that it is well known by the
people working in the field (the good ol' technicians) and easier to
implement, because you know it. Fuzzy is old, but also new. Designed a
long time ago, took a while to get popular with the Japanese etc. , but
us westerners never really bothered with it, did we?

Thats about as much as I know - so I stand corrected!
--
Eric van Es               | Cape Town, South Africa
spam_OUTvanesTakeThisOuTspamilink.nis.za | http://www.nis.za/~vanes
LOOKING FOR TEMPORARY / HOLIDAY ACCOMODATION?
http://www.nis.za/~vanes/accom.htm

1997\09\03@182555 by trogers

flavicon
face
Lots of talk. I'll suggest the same thing Bob Pease did: let's have some
examples.
Is there any small system that we could propose and actually build that will
work
driven by fuzzy logic and not when driven by a pid algorithm, or is so much
better
when done using fuzzy logic (code size, speed, performance, system cost,
whatever)
that no engineer in his/her right mind would use the pid driven version?

I'll commit some of Time Tech's resources to the actual construction of such a
project, provided it's not too large.

Any ideas?

--Tom Rogers  VP-R&D  Time Tech Inc.

BTW, I'm firmly on the 'I like FL' side of this debate, but not many people I've
debated understood what the heck I was saying, and that includes an exchange
with
Bob Pease. I've got Kosko's text and a bunch of other stuff, and I've used FL in
a
couple of places, but I'm no expert. --TR

1997\09\03@194915 by David Thornton

flavicon
face
> > > Pease Porridge of
> > > Electronic Design ran a contest for a Fuzzy Control system that he
> could not
> > > duplicate or exceed the performance of.  He won every challenge using
> op-amps
> > > and classic control theory. Fuzzy Control and Fuzzy Logic has had lots
> of
> > > boasting and very little real substance.

> When ... developing Fuzz-C we looked hard for applications
> where fuzzy worked as well or better than conventional PID. When we did
> we found that PID was highly optimized for a narrow range of
> applications and fuzzy was a more general case.
>
> For example look at the momentum difference in an aircraft control system
> after the fuel stored in the wings has been burned in flight. A PID control
> loop in such a system is very hard to tune.


It may not be fair to compare PID vs. fuzzy. The biggest strength or
advantage you will obtain with a fuzzy logic is the range of
applications, the ability not only to have a competitive performance
where PID work well but also acceptable results in some applications
PID, or any other conventional control methodology, poorly works or
does not work at all.

But it may not be the most precise controller you can come up with
when you have other alternatives. Nevertheles, it is very robust and
simpler to implement.

__Aydin
I suggest you examine Mitsubishi's newest automatic gearboxes, these
use fuzzy logic, they are not perfect but are a huge improvement over
previous electronic & hydraulic systems found on other cars that I
have driven. Apparently early versions of software could learn bad
habits!!
David.

1997\09\04@095208 by Martin R. Green

picon face
Please, correct me if I'm wrong (OK, so you would have anyway), but I keep
seeing references to FL 'learning' better responses to input.  I thought FL
was just a way to deal with ambiguous (not yes or no) inputs, and that
learning was the domain of neural software.  I can see how the neural
weighting is a form of FL, but not the other way around.  Am I completely
out in left field?


Martin R. Green
.....elimarKILLspamspam@spam@bigfoot.com

----------
From:   David Thornton[SMTP:david-tspamKILLspamSA.SCITEC.COM.AU]
Sent:   Wednesday, September 03, 1997 7:39 PM
To:     .....PICLISTKILLspamspam.....mitvma.mit.edu
Subject:        Re: Fuzzy Control Tool (was Re: PID Temperature Controller

> > > Pease Porridge of
> > > Electronic Design ran a contest for a Fuzzy Control system that he
> could not
> > > duplicate or exceed the performance of.  He won every challenge using
> op-amps
> > > and classic control theory. Fuzzy Control and Fuzzy Logic has had
lots
> of
> > > boasting and very little real substance.

> When ... developing Fuzz-C we looked hard for applications
> where fuzzy worked as well or better than conventional PID. When we did
> we found that PID was highly optimized for a narrow range of
> applications and fuzzy was a more general case.
>
> For example look at the momentum difference in an aircraft control system
> after the fuel stored in the wings has been burned in flight. A PID
control
> loop in such a system is very hard to tune.


It may not be fair to compare PID vs. fuzzy. The biggest strength or
advantage you will obtain with a fuzzy logic is the range of
applications, the ability not only to have a competitive performance
where PID work well but also acceptable results in some applications
PID, or any other conventional control methodology, poorly works or
does not work at all.

But it may not be the most precise controller you can come up with
when you have other alternatives. Nevertheles, it is very robust and
simpler to implement.

__Aydin
I suggest you examine Mitsubishi's newest automatic gearboxes, these
use fuzzy logic, they are not perfect but are a huge improvement over
previous electronic & hydraulic systems found on other cars that I
have driven. Apparently early versions of software could learn bad
habits!!
David.

1997\09\04@171402 by trogers

flavicon
face
Martin R. Green wrote:

> Please, correct me if I'm wrong (OK, so you would have anyway), but I keep
> seeing references to FL 'learning' better responses to input.  I thought FL
> was just a way to deal with ambiguous (not yes or no) inputs, and that
> learning was the domain of neural software.  I can see how the neural
> weighting is a form of FL, but not the other way around.  Am I completely
> out in left field?

No, basic FL is a logic system and has no memory. There are several methods of
rendering a system based on FL adaptive, including some that use standard neural
net techniques.

--Tom Rogers

More... (looser matching)
- Last day of these posts
- In 1997 , 1998 only
- Today
- New search...