Quote from: TinselKoala on October 14, 2012, 08:51:33 AM
@Mondrasek: thanks for doing that. I'll run the numbers shortly, but meanwhile can you give some description that correlates the parameters in your column titles with your verbal description in the earlier post? I don't get these "max" and "min" terms.....

The test started by turning on the shaker.  It was running and therefore vibrating the test ZED slightly the entire time the test was run.  A little water was added to the system through the fill tube to make sure everything was moving freely.  The Pod/Riser system was floating at a point between the bottom and top stops.  It was fully loaded with the addition of a magnet on top of the lift mass.  The magnet was used to pinch one end of a monofiliment fishing line between itself and the lift mass, thus anchoring one end of that line.  The other end of the line was run fairly straight up and through a small pulley that had been suspended from the ceiling above the test ZED.  After looping over the pulley the line was tied to a small ring magnet to anchor that end and give a place for the test weight to be attached.

A steel weight was set on top of the ZED system and caused the Pod/Riser to sink a bit.  The average value of the indicator was recorded at this point in the first column labeled "With Extra (+) Weight."  I say average because the indicator was fluctuating rapidly through a range of ~0.02 mm due to the shaker vibrations.  The weight was then removed and the system given about 20 seconds to fully stabilize as the Pod/Riser system rose back up a small amount.  I then recorded the maximum and minimum readings of the indicator at the new Pod/Riser height.  These are the two columns labeled "After Removing Extra Weight MAX" and "After Removing Extra Weight MIN."

Next, the same steel weight was suspended from the fishing line by attaching it to the hanging ring magnet.  This caused this weight to counterbalance the ZED system a bit by pulling up on the lift mass.  The Pod/Riser system would rise slightly and the average new height was recorded in the column labeled "With Extra (-) Weight."  The weight was then removed and the system given about 20 seconds to fully stabilize as the Pod/Riser system sank back down a small amount.  I then recorded the maximum and minimum readings of the indicator at the new Pod/Riser height.  These are the final two columns labeled "After Removing Extra Weight MAX" and "After Removing Extra Weight MIN."

M.

PS.  Sorry for not posting a description earlier.  I had thought to shoot a short video instead but did not get around to it.

Well, I'll have to think about that description before I can provide a good interpretation of the data. But without knowing anything but the numbers, I can say a few things anyway.

We assume that nothing changes between trials and that each trial is a "duplicate" in all respects.... so the "true" value of any parameter should be the same, across trials. If the measurements really were the same, then you'd see a straight horizontal line across the trial numbers, at the value of the measurement. Since there is some noise to be expected, the mean value of all the trials can be plotted as this straight line, and then each individual value can be compared to the mean to see if it's an outlier, close to the mean, or just crazy noise. "Close" here means within one Standard Deviation of the mean value, plus or minus. You can also define the magnitude of the noise as being the size of the standard deviation: noisy data will have larger variations from the mean than will "quiet" data.

So ideally we'd like to see straight horizontal lines, with small standard deviations, and all data points lying within one SD of the mean line.

But that's not quite what we get. I see noise and a trendline.... I think this: first, your measurement technique got better over the 20 trials. Second, something happened toward the end, you have a small leak or you are losing water or air in the transfer or something, because there is a definite trend in the data that is different from the assumption of a constant mean value.

What it all means in terms of the actual apparatus interpretation, will have to wait until I've slept and thought a bit. But feel free to think and interpret on your own.

For some reason the graph in the spreadsheet lost some of the data symbols when converted. Maybe somebody else can fix it. The image is before conversion though.

Hmm.

Well I think that my test ZED *must* have a leak somewhere.  The data that TK shook down is one thing.  But I have also been trying to establish a repeatable test stroke for about half a day now.  And while that has gone very well (as far as repeatability) it has shown a repeated anomaly that I cannot explain without considering a leak of some sort.

I have been stroking the test ZED system through *exactly* 10.0 mm.  This is very easy to do now.  But the problem comes in when I start to lift from the bottom (sunk) state.  As pressure builds in the ZED I will see some overflow from the Outer Chamber on to the top of the Outer Riser.  This was a normal part of setting up the system before.  So I would remove the excess water with a pipette and continue.  I would repeat the lift and sink cycle until the overflow condition ceased.  That is how I set up for the test runs reported earlier.

But now I have run the same test cycle over a dozen times and *every* time I have to remove some overflow water.  So it is not stabilizing.  There is not loss of air in the system or "blowing skirts."  But there is excess mass of water in the Outer Chamber each time.  So water must be leaking from the inside annulus to the outside annulus?!?

I can leave the system alone at any lift condition for quite some time without any change to the apparent overflow anomaly.  So my guess is that there is a leak between the retainer walls that is affected only when the system is temporarily under a "vacuum" condition as water is vented.  Very weird.  And I am removing the overflow with an eyedropper type pipette so it is only a few ml each cycle.  And fairly repeatable.

The rest of the lift cycle is also very repeatable.  I have tape marking the fill tube at the exact level that the system lifts and sinks.  Those levels repeat within eyeball error.

Very cool results.  But also disappointing.  If there is a leak it is not something I will find easily.  But I wonder if test results from the system as is are worthwhile?

M.

"But I wonder if test results from the system as is are worthwhile?"

Well, to answer that we need to know what the predictions are. If the system were acting as a normal multiple hydraulic cylinder, what would you expect the values in the data table to look like, and if it were producing any extra work or pressure, what would you expect the values to look like?


But regardless of the answer to that, sure, they are worthwhile, because they were good practice for you.... you can see your technique improving over the twenty trials -- and they indicate that leak, or perhaps your fluid removal. In any case they indicate things to be aware of and a possible problem that needs fixing in the apparatus.

Sure, you can find the leak. Use clear water in each chamber except one, make that one dark with food coloring. Stroke. Look for colored seeps. Don't find any, reload system with clear water in each chamber except a different one, repeat stroke, look for colored seeps. Do one chamber at a time with dark color. Eventually you will see where the colored water is bleeding past one of your flexible seams when the system flexes under pressure changes. If it's an air leak..... you can put soap in your water and get all the inner surfaces wet, even the air chambers, and when the leak occurs you will see bubbles forming in the soap.

Quote from: TinselKoala on October 14, 2012, 04:46:39 PM
Well, to answer that we need to know what the predictions are. If the system were acting as a normal multiple hydraulic cylinder, what would you expect the values in the data table to look like, and if it were producing any extra work or pressure, what would you expect the values to look like?

Exactly!  I've been trying to post impartial test data.  But I have also been asking what kind of test to run to put a nail in this thing either way.

So, what is the definitive test?

Thanks, TK, for your input.

M.