Vibration After a VFD Retrofit
After a 22 kW induction motor driving a centrifugal fan was fitted with a variable-frequency drive, bearing-housing vibration rose from 2.1 mm/s RMS to 11 mm/s RMS, dominated by 1× running speed. The fan runs at 1,470 rpm. A coast-down test shows the amplitude peaking sharply at 1,465 rpm and falling to about 3 mm/s within 40 rpm either side. A bump test on the assembled skid gives a first natural frequency of 24.4 Hz. The skid is a welded channel base grouted to a concrete floor, but two corners showed 0.15 mm soft-foot gaps before grouting. Operations wants to raise the trip level from 7 to 12 mm/s and run to the next shutdown; maintenance wants to rebalance the fan immediately.
1. Diagnose the dominant problem using the frequency and coast-down evidence.
Read the worked answer
Running speed is 1,470/60 = 24.5 Hz, while the measured natural frequency is 24.4 Hz. The separation is only (24.5 − 24.4)/24.4 × 100 ≈ 0.4%, far below the 15–20% margin normally sought, so the rotor is running essentially on resonance. The coast-down peak at 1,465 rpm (24.42 Hz) confirms that the system passes through a resonance and that the response collapses once it is 40 rpm away. The amplification is also consistent with resonance: 11/2.1 ≈ 5.2, which corresponds to a lightly damped system with an effective damping ratio of roughly 0.1.
The dominance of 1× indicates that the exciter is residual unbalance, possibly with a soft-foot or alignment contribution, not a bearing defect (which would show non-synchronous, high-frequency energy). The 0.15 mm soft-foot gaps matter because pulling the skid down to the floor distorts the frame and can lower its effective stiffness, shifting the natural frequency towards running speed. The conclusion is resonance excited by unbalance, not a sudden deterioration of the bearings.
2. Assess the two proposals and recommend a technically defensible remedy, quantifying the stiffness change needed.
Read the worked answer
Raising the trip from 7 to 12 mm/s is unacceptable: it treats the protection as the problem and exposes the bearings and foundation to sustained resonant response. Rebalancing alone is helpful but insufficient, because even a small residual unbalance is amplified near resonance and the natural frequency can drift with temperature, foundation condition, or fan blade pitch.
First correct the soft foot properly - shim and regrout so the frame is not distorted when bolted down. Then stiffen the skid with diagonal braces, gussets, or a stiffer plinth so that the first natural frequency is at least 20% above running speed. The target is 24.5 × 1.2 ≈ 29.4 Hz. Since frequency scales as √(k/m), the required stiffness ratio is (29.4/24.4)² ≈ 1.45, i.e., about 45% more stiffness, assuming mass is unchanged. Adding mass would lower the frequency and is the wrong direction. Rebalance after stiffening, and use the VFD to lock out a speed band around 1,440–1,500 rpm only as a temporary measure, not as the permanent fix.
3. How would you verify that the remedy has worked before releasing the fan for continuous duty?
Read the worked answer
Repeat a bump test on the assembled, grouted skid and confirm that the first natural frequency is at least 29 Hz and that no other mode has moved close to running speed. Then perform a run-up and coast-down with vibration and phase measurement at the bearing housings, in both horizontal and vertical directions, checking that the 1× response no longer peaks in the operating range and that phase remains stable. Rebalance to the appropriate balance grade and record residual vibration at the normal operating point.
Acceptance should be based on a combination of broad-band vibration at or below the plant alarm level, bearing temperature stability, and no soft-foot indication when bolts are loosened one at a time. Run for 72 hours with trend logging; if vibration or temperature drifts upward, the stiffening or alignment is not yet adequate. The verification report should include the before-and-after bump-test data, so the change in natural frequency is documented rather than assumed.