I plan to write about four refrigerators and an icebox; let’s see what happens.... About a year ago, the radio said that Whirlpool had won the $30 million contest for an Efficient Refrigerator. “They used Fuzzy Logic to improve the compressor, and the efficiency was improved by a factor of 3.” Wow! How did they do that? A few months later, I read in Machine Design that Whirlpool’s machine was 29% more efficient than the 1993 Federal standards. Well, that’s not 3:1, but it’s still quite an improvement.
I later read in Popular Science that the basic old refrigerator design was already 20% better than the Federal specs, so they only had to make an additional 9% improvement to win the contest. Well, an improvement from –20% to –29% is a relative improvement of about 11%. And about 1/3 of this was attributed to the Fuzzy Logic, which was not used to improve the compressor, but only to set the best time for the defrost cycle.
Then Mr. R. Shattuck of Gillette, Pa., suggested that you can connect the wiring for the defrost timer over to the switched side of the thermostat. This is an old refrigeration engineer’s trick, so the defrost timer only runs when the compressor is running. If it’s a cold day and nobody opens the refrigerator’s door, then the time between defrost cycles is extended. If it’s hot and the kids are opening the door a lot, the compressor will run a lot, and the time between defrost cycles will decrease. This may provide as much gain in efficiency as is claimed by Fuzzy Logic. The alleged advantages of FL are kind of Fuzzy when you compare it to this old trick.
Now, I was still looking for a case that had a really serious debate between a good FL system and a good “conventional” system. A friend sent me a data sheet on a Fuzzy-Logic-based oven controller made by Space Industries in Texas. It claimed to provide a 10X advantage—0.3°C accuracy vs. 3°C for the “conventional” oven controller.
But I saw some data showing that this controller didn’t just approach the set point and then slow down and stop. It kept cooling long after the oven temperature fell below the set point, and it kept the heater turned on long after the temperature rose above the set point. Why didn’t this controller recognize that it was approaching the set point, then slow down, turn off the heat, and settle promptly? An engineer at this company said that if the mass of the oven was nominal, the overshoot was negligible. But I observed that the data sheet for this oven said that it was rated at full accuracy with large or small thermal mass. This controller appeared to ring and oscillate for hours at any light amount of mass.
I volunteered to help redesign this oven controller, because I know how to make a good controller for a simple case like this. I volunteered to do this with an analog controller (a quad op amp), or a digital controller, or a Fuzzy controller. The company responded that since the controller met specs, they had no interest in further improvements. I knew I could make some serious improvements, because the Fuzzy controller responded only to the temperature error and that error’s rate of change. Thus, it was like a P-D controller—Proportional and Derivative terms. I knew how to make a PID controller, which added an Integral term. By adding an integrating term to increase the gain for long-term errors, I could cut back on the Proportional Gain and improve the damping a lot. (If you don’t have any Integral Term, it forces you to turn up the proportional gain to get good accuracy, which causes the ringing.) In fact, I bet I could meet 0.03°C....
Well, the engineer at this company decided he didn’t want to be helped. That reminds me of the old lady who had to be helped across the street by five Boy Scouts because she did not want to cross the street.
I studied this some more. Some people—even guys at USC—claim that there’s an inordinate number of Fuzzy Rules you have to write in Fuzzy Logic if you advance from 2 dimensions to 3 or 4 variables. If you want to write 7 rules for each dimension, then you’d have to increase the number of rules you write from 49 to 343 to 2401. Wow, that’s a heavy load. I mean, the Fuzzy Logic enthusiasts always say it’s fun and easy to write the rules for Fuzzy Logic. Oh, yeah, but it gets tedious when you have to write 300 rules, and when you have to write 2400, that really gets to be hard work. So most Fuzzy engineers decide that they don’t have to use 3 variables—they can get perfectly fine results with 2 variables (they may be wrong....).
Well, there’s no real reason that you have to write 7 rules in any one dimension, for any one parameter. So, maybe you can write 3 x 2 x 3 rules, rather than 7 x 7 x 7. Now, you might not be able to do this if the system is heavily nonlinear. You might have to write 7 x 3 x 2 rules, which still isn’t more than 49. SO, a person could write the rules for a fairly linear PID FL controller and still not run out of ink. (Note, you might want to avoid getting into a hissing contest with a guy who buys his ink by the barrel.) So, the “explosion of rules” may turn out to be no big deal. And a Fuzzy PID controller can be easy and effective to accomplish—and can perform lots of tasks much better than a simple PD type.
One of the rules I saw for a heater written in Fuzzy Logic went: “If the temperature in the chamber is very cold, turn on the heat very high.” My comment on this is: How many of you guys ever owned a VW, Beetle or Bus? I have put on over 106 miles in VW Buses and Beetles, and I know darn well: “If the temperature in the chamber is very cold, you turn on the heat very high.” Of course, you can turn up the heat in one of these old VWs all the way, and even at full throttle, the heater doesn’t kick out enough heat to keep you warm when it’s below zero outside. Is that a refrigerator? Or is it just an icebox?
It’s not the fault of the controller, it’s the fault of the heater’s capacity. BUT, even if the heater does have enough heating capacity at +40°F outside, you could only get the heat to turn on to full heat when the inside temperature had already gotten too cold, if you used that Fuzzy rule.
Then you could never get the controller to give good results. The weakness there is, if you trust the conventional way that rules are written for FL, most engineers don’t understand the limitations of PD controllers, and they don’t understand the advantages of PID, but they think they can beat the game. They might be wrong. (I don’t mind the weak heater in my ’68 Beetle, because I make sure to bring along plenty of warm clothes for cold-weather travel.)
Recently, we got a letter from an engineer asking, if he buys a “Greenplug,” can he use this NASA-based technology to save on energy running his refrigerator? I got several letters from readers. One of them pointed out that the “Greenplug” does NOT use the old NASA technology, but is just a switch-mode voltage regulator. If the line voltage at your house is typically 124 V, you might save some energy. But if your line is at 106 V, you may not save anything.
Another guy sent in the results of Consumer Reports’ testing which revealed that on old refrigerators you could save a few percent, but on a newer refrigerator, you might not save any electricity at all—or, you might get less efficiency than without. Then someone else said his Greenplug had failed, and he found out only after several days when the contents of his warm freezer let him know. (Send me a SASE for the schematic of an updated low-power detector/alarm for freezer failure.)
Other people sent in glowing claims for other “economizer” controllers that would save energy “by not letting the refrigerator’s motor get too much voltage or too much current.” BUT then I began to see a scary problem: If a refrigerator is running, and power is lost for a few seconds and then restored, the compressor can usually start, even against the back pressure. But if the line voltage is low, and the line’s impedance is soft, the compressor may stall. It might even blow a fuse, or blow out the “economizer.” And this is all compounded by the fact that the “economizer” is trying to provide a soft startup voltage and a small startup current.
If you want to buy an “economizer” for your refrigerator, make sure that the “economizer” is approved by the manufacturer of your refrigerator. They would be best qualified to certify that restart under low-line conditions won't cause stalls or damage.
Finally, I must say that I smile every time I hear about our Secretary of Defense, William Perry. Imagine that—a 340-lb. “refrigerator” in the Pentagon! (Even those of you who aren’t football fans may know, with some prompting, about William “the Refrigerator” Perry who plays for Philadelphia....) But, maybe that was a different guy? Enough!
All for now. / Comments invited!
RAP / Robert A. Pease / Engineer