2 plane Fan balancing 1

Doing a two plane balance with the 4 run method and only using 4 runs would be quite a trick. I would be interested in hearing about it as well. The following was posted by Arne Lindholm at the Reliability Magazine web page and describes a two run method. I suppose if you used this method in two planes and were very very lucky you might get a two plane balance in four runs. But it seems to me that with any significant cross effect that it would expand to 8 or 16 runs very quickly:

Since the four run method seems to have survived vectors methods, maybe you should consider the two-point method which can save some runs.
The Two-point Method. Take first run as reading (a). Place a testmass at 0, measure (b), stop and move the mass across (exactly 180 degrees). Run and take (c). Make a simple construction on a paper. Draw double a in a suitable scale. Make a triangle using b and c starting from the lines ends. Take the smaller of b/c to the right. Extend the right line to double length. The missing line to complete to a triangle to the left is (d). d is related to 2a as test mass is to balancing mass. Angle from best test position (either b or c) to insertion of balancing mass is the full angle to the left d to 2a. To take the angle against or with rotation is the only uncertainty. Locate heavy point (only hor. shafts) and/or check runout to decide which position is the correct one. A drawback is that with a broadband vibrometer, other frequencies can make the precision bad. Neighbour fans should be stopped. The method is old, but reliable. You can test it on your table fan. Regards Arne

Nice to see written instructions for that method, but it won't work as written on a strongly cross coupled system.

Minimising the number of runs requires you to extract the maximum amount of information from each run. This means that in practice your choice of trial weight is reasonably significant, and that you have to know the absolute phase (relative to the driveshaft) and magnitude of both measurements and the trial weights. Speed and temperature control should also be considered.

Here's how to 2 plane balance a driveline, as done 60 times per hour in an assembly plant not far from where I sit, electronically.

2 measurement planes A, B, 2 correction planes a b

1) measure initial unbalance A1, B1 (all vectors, obviously)(if it passes let it go)
2) estimate optimal trial unbalance from old data
3) add trial weight to a
4) measure A2 B2 (if it passes let it go)
5) (optionally remove trial weight from a and) add trial weight to b
6) measure A3 B3 (if it passes let it go)

Now you can work out the actual initial unbalance and all the cross coupling effects, as discussed in your mechanics book.

7) (optionally remove trial weight from b and) add final weights

8) measure A4 and B4. repeat ad infinitum

That's your 4 runs. I strongly suggest removing the trial weight each time, but in theory it is not essential.

The book you need is William Thomson's vibration book, although B&K or many other companies will also have the details. Roughly speaking you are solving the vector equations

A=a*ivAa+b*ivAb
B=b*ivBb+a*ivBa

where iv is an influence vector, and by reciprocity ivBa=ivAb

a is the actual unbalance at that plane, ie the vector sum of the initial unbalance and any trial weights.

Typically cross coupling varies from 3% to 35% on vehicle drivelines. Cheers

Greg Locock

Thanks for the response. However, I am looking for a method to do a 2 plane balance without phase measurements. I read a reference to it in Victor Wowk's book "Machinery Vibration, Balancing". It implies a method similar to the 4-run method but using 7 runs. Has anyone seen this.

Regards,

Siva.

Contact < <nbucska@pcperipherals.com>

Clue 1: I (think I) could theoretically do that in 6 runs.
Clue 2: I would not trust the answer at all!

The first problem with this approach is getting the concepts straight in your head, the actual maths is pretty straightforward.

The big problem is that in the real world the absence of phase information will introduce a lot of potential errors into the process.

You need to sit down and draw 4 complex planes. Using the results of each set of 3 experiments you can determine the cross coupling vectors, and the self-influence vectors, for each of the ends, and the initial unbalances. Reciprocity tells you that 2 of these graphs should be identical. Then add the indicated correction masses.

There is a possible ambiguity in this method, which the 7th run (with no trial masses) removes.

I have done the single plane version of this and would say that the additional time and aggravation required to get the phase measured will repay itself 10 fold, in practice.

Cheers

Greg Locock