Refrigerator control

Tim Wescott wrote:


ITYM 'faster'. I don't mean this as a jibe, just clarification.
While there is a consensus that, for a domestic fridge, PID, modelling, etc are unnecessary, I think that as a practical learning exercise it has a lot to offer. Next stop fuzzy logic?
Further thoughts:
Use the door operated inerior light switch as an input to: maybe switch off while open, switch on for longer once closed (shift the setpoint, or feedforward an integrated term), sound an alarm if open too long. Other alarm conditions can also be incorporated.
With the thing connected to a PC we have the potential for a 'web-fridge'. You dont have a barcode scanner to hand do you, Rado?
This has been a fun thread and goes to show that 1 problem + n control engineers = n + m solutions.
Cheers, Andy
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Andy McC wrote:

I assumed he did too.

PWM is just a way to exercise proportional control. An absorption refrigerator is not amenable to proportional control. If enough heat is supplied, it cools; otherwise not. The rate of cooling is nearly independent of the supplied heat within the operating range.
...
Jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Andy McC wrote:

Sigh. Thanks.
One of the best pieces of advice that I was never given (it was related to me by a software engineer who really didn't do control) deals with figuring out whether your controller should have an inverting amplifier or not. The advise is "Count all the sign inversions in the system, then add in the one you missed." I guess this is the one I missed...

I have a knee jerk negative reaction to that term, because "fuzzy logic" is one of those things that makes the less talented get all starry-eyed. From what I can tell it's a useful technique but it's not a panacea -- it's just a structured, formal way to apply common sense to difficult nonlinear control problems. If you can do the control yourself, but can't figure out how to build a controller for it, then fuzzy logic may help you get there.
What it _can't_ do for you is verify that you've built a controller that will actually work, and if you're using it to control a nonlinear system, or if your controller has introduced significant nonlinearities to an otherwise linear system, then you still have to analyze the resulting system to make sure that it will suit. Since the essence of a nonlinear system is that it's behavior is not global this means that you have to either subject it to difficult and messy stability analysis, or you have to run about a gazillion simulations on it from all sorts of screwy starting points.
To date I've always been able to find a common sense solution without using fuzzy logic just by telling myself "well, if I can't do this directly I can always learn fuzzy logic!".

Using the door switch for additional information can't be a bad thing.

I think that's n * m, with m unbounded.
--

Tim Wescott
Wescott Design Services
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Tim Wescott wrote:

[some snipping, without, I hope, changing the meaning]

This is pretty much my take too.

Hmm, yes, I failed to qualify 'm', my bad.
A manager from my past once said words to the effect that the more difficult a system is to control, the more difficult it is to control the engineer who can control it.
--
Andy McC


Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Andy McC wrote:

The actual behavior of the box is the result of the combined effects you list. Since the behavior, not its causes, needs to be controlled, separating the effects isn't necessary. At it's simplest, measure the under/overshoot and adjust the next cycle to correct it. Sophisticated analysis can shorten the convergence time, but that may not be needed.

An anticipator behaves like a D term.
Oh BTW, given the inherent integrating effect of the system, I don't

You do need hysteresis, but not likely that much. In practice if not on paper, every switch needs to be debounced. Hysteresis does that.
Jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

I have my AREF for the ADC at 5.0V and the sensor outputs 10mv/degree kelvin, so at the temperatures I measure is around 2.8V. ADC resulution in this case is 5mv, so I have 0.5 degree K resolution and around 1 degree accuracy (from the ADC), but the sensor itself is rated +-1 degree. I don't need more for the current project I think.

I thought of measuring the lag once and hardcoding it, but your idea to make dynamic adjustments makes the project more interesting. I'll try to do it. This I suppose will give better results when external temperature and refridgerator contents vary.

If I can keep the temperature inside in a 2 degree C band for periods when the door is not opened I'll be more than happy.
When the door is opened for longer than a couple of second the temperature rises and it needs some time to compensate. With such low powered unit I think this is unavoidable (the heater is less than 100W), I'll just have to make up my mind before opening the door ;).
Regards, Rado
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@gmail.com wrote:

Air temperature rises quickly when the door is opened, but the food, because of its high water content, doesn't. The food and other parts of the box will begin to cool the air significantly as soon as the door is closed, even without the refrigerator turning on.
Jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

You'd need to be careful applying any sort of duty cycle control with this system in that the cooling effect is likely to be pretty nonlinear. Applying partial heat is likely to result in minimal cooling below a certain threshhold.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

what an interesting thread this has been......
Never thought that simple on/off cooling control could make such intersting read.
sQuick..
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Yes, indeed!
I'm really impressed with the replies to my first post in the sci.engr.control group.
Thanks everyone for their time and expertise.
I'll post the results when there's some progress, or ask again if I reach some roadblock :).
Greetings, Rado
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@gmail.com wrote:

By all means, do. You generated a lot of interest. Good luck!
Jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.