Pumps and Systems, August 2009
One of the major contributors to unbalance in rotating machinery is eccentricity. When we disassemble parts, we must also reassemble them. Even if the reassembly requires heating to shrink-fit the elements back together, we probably balanced them in a balancing machine where tolerances due to fit-up on mandrels or roundness of journals may possibly far exceed the heavy spot tolerances that the target standard of balance allows.
This is especially true when the parts are light weight and high-speed, thus making the calculated allowable heavy spot extremely small.
A high speed air compressor rotor weighs 7 lbs and operates at 55,000 rpm in the field. The user wanted to use the API 610 Balance Standard of 4W/N, where W represents the static journal load on each end in pounds and N represents the maximum continuous operating speed in RPM.
4(3.5)/55,000 = 0.00025 oz-in, or 0.0072 gr-in
A test was proposed in which the rotor would be balanced to the proper level in a balancing machine with axial thrust limits, thus ensuring that the part would spin at the same place on the journal each time. After balancing, the part was stopped and re-spun 10 times to test repeatability of the left plane of the machine. The balancing speed was 2,077 rpm. The part showed a maximum deviation of 6 mg-in to 8 mg-in during the 10 runs. In other words, it was balanced to the target set forth above, as shown in Figure 1.
Great care is taken in the design of this particular rotor with respect to attaching the compressor wheel to the shaft. The fit is a polygon shape, rather than round, so that repeatable fits can be ensured. This rotor was set up in the balancing machine so that it could be removed and replaced exactly at the same place for testing purposes. The rotor was removed to a "V" block vise, clamped down and the nut was removed from the nose of the wheel. The compressor wheel was found to have approximately 3 deg of radial movement. It was removed and preloaded clockwise 3 deg at reassembly.
The wheel, shaft and nut were, of course, match marked for reassembly. A precision torque wrench was employed to ensure the nut was repeatable in reassembly. The reassembled part was then placed back in the balancing machine and re-spun. This cycle was repeated 10 times. The average of the runs is displayed in Figure 2.
The average from the test of simply reassembling the part after balancing was 178 mg-in. The tolerance is 7 mg-in. Was the part better when balanced to 4W/N? The part may actually have become worse from a balancing standpoint due to the eccentricity of reassembly. The target should have been 178 mg-in from the beginning.
As if the eccentricity problem with reassembly were not bad enough, a test was performed where the axial position of the part in the balancing machine could be altered while running. It was decided to operate the part at a different place on the journals. The original setup was as shown in Figure 3; the radii were changed to 1 in for the testing.
The axial position of the part was changed by approximately 0.25 in, so that a different journal surface was now in contact with the balancing machine bearings. The result is shown in Figure 4.
The journal test points out that any precision balancing job should be set up in the balancing machine spinning precisely around the same centerline as will be fixed by the bearings in place. If this is not done, eccentricity is balanced out and a false heavy spot is created, which will cause vibration when the part rotates around the centerline that the bearings dictate.
Tightening Tolerances-Fallacy or Fix?
We will now take a typical example of a pump impeller to be balanced. This impeller weighs 20 lbs and turns 3,600 rpm in the field. It also has a 1 mil (0.001 in) slip fit on its shaft. If we calculate API 610 Standard for this part:
4W/N = 4(20)/3600 = 0.02 oz-in or 0.62 gr-in
The impeller weighs 9,072 grams.
If we assume that we shift the impeller centerline by half of the slip fit, or 0.0005 in X 9,072 grams, the result is 4.5 gr-in, and the tolerance is 0.62 gr-in.
Is balancing to 4W/N a waste of time? Probably.