Field case findings: what 39 bearing and machinery cases show
Across 39 field cases, 5 came down to looseness, soft foot or a weak support, and 7 to lubrication or oil condition. Where before and after readings exist, fixing a loose support cut vibration by 78% to 87%.
The dataset
39 anonymised cases: 26 condition monitoring, 6 failure investigation, 4 oil analysis, 2 design review, 1 energy and reliability. Each case was assigned one main cause theme by reading its summary. The data file is below.
What the cases came down to
| Main theme | Cases | Share |
|---|---|---|
| Early bearing defect detection Trend data gave months of warning | 18% | |
| Lubrication and oil condition Grease or oil state was the cause or the clue | 18% | |
| Balance, resonance, rubs and beats Vibration from the machine, not the bearing | 15% | |
| Looseness, soft foot and weak supports Check the base, bolts and bearing fits before replacing parts | 13% | |
| Gear damage Sidebands and waveforms pointed to the gear | 10% | |
| Design and load inputs The bearing was fine; the inputs decided the life | 5% | |
| Electrical damage Current through the bearing | 5% | |
| Assembly and operating errors Overhaul or operating practice created the failure | 5% | |
| Measurement and operating point Reading the data correctly, and running the machine at the right point | 5% | |
| Alignment and thermal growth A procedure fault repeated across many machines | 3% | |
| Contamination Dirt got into the bearing | 3% |
Counts describe this collection, not how often each cause occurs in industry. Cases that get written up skew towards faults that were found and fixed.
How big were the fixes?
Five measurements from four cases have both a before and an after reading on the same point.
| Case | Before | After | Reduction |
|---|---|---|---|
| Motor soft foot (one bolt loosened and shimmed) | 6.2 mm/s RMS | 1.2 mm/s RMS | 81% |
| Forced-draught fan, loose pedestal fixed | 8.3 mm/s RMS (axial) | 1.8 mm/s RMS | 78% |
| Condensate pump, loose bolts tightened | 2.94 mm/s RMS | 0.38 mm/s RMS | 87% |
| Hydro rotor, two-plane balance, plane 1 | 7.34 mm/s | 0.25 mm/s | 97% |
| Hydro rotor, two-plane balance, plane 2 | 4.38 mm/s | 0.48 mm/s | 89% |
The three looseness-type fixes cut vibration by 78% to 87% for the cost of a spanner and some shims. The balance case shows the other kind of fix: right weight, right place, first time.
How much warning did monitoring give?
| Case | Warning | What happened |
|---|---|---|
| Ball mill pinion bearing | 9 months | Impact-based measurement flagged an outer-race fault; standard vibration on the neighbouring bearing showed nothing until replacement day |
| Rotary calciner support wheels, 3.5 rpm | 9 to 10+ months | Three defects found early and confirmed at teardown |
| Cement mill gearbox | 5 months | Gear wear seen in spectra before the teardown, with cement dust in the oil |
| Pulveriser gearbox thrust bearing | 1 month | Ferrous reading rose from 0 to 352, then 443 in routine oil tests |
| Paper machine press roll, 130 rpm | 8 days | A weak outer-race line at 0.98 mm/s doubled in eight days and was confirmed at teardown |
The pattern is that the warning was there in the data in every case. The value was in the plant acting on it: planning the stop instead of reacting to the failure.
Five patterns worth acting on
- Check the support before replacing the rotor or bearing. 5 cases ended at a bolt, a base or a bearing fit.
- Overall velocity can pass a damaged bearing. Several early detections came from acceleration, envelope or oil data while overall vibration looked acceptable.
- Fixed lubrication intervals do not show condition. Oil tests and measure-before-and-after greasing caught what the calendar missed.
- When many machines fail the same way, suspect the procedure. Fifteen boiler feed pumps shared one alignment method that ignored thermal growth.
- Check the inputs, not just the arithmetic. In two design reviews the life changed by more than ten times depending on a load or shock-factor input.
All cases by theme
Looseness, soft foot and weak supports (5)
- Motor soft foot
- Electrostatic precipitator fan at 113 mm/s RMS
- FD fan
- Condensate pump
- Roughing mill gearbox
Early bearing defect detection (7)
- Ball mill pinion bearing
- Ball mill main bearing
- Rotary calciner at 3.5 rpm
- Fan drive-end NU314M
- Closed-cycle pump
- Paper machine press roll
- Stacker-reclaimer slew bearing
Lubrication and oil condition (7)
- Oil+air system fully "OK", yet the bearings kept failing
- Overgreased bearing
- Lubrication by condition
- Locomotive engine
- Medium-speed diesel
- Pulveriser gearbox
- Wood chipper
Balance, resonance, rubs and beats (6)
- A 160 tonne hydro turbine rotor balanced in place with 42 kg, first ti
- Induced-draught fan, 2,630 kW
- Motor cooling fan at a steel mill
- Blower fan 1X rising and falling every 3 seconds
- Cooling water pump
- A cracked bearing ring, but the bearing was only the victim
Gear damage (4)
Design and load inputs (2)
Electrical damage (2)
- Fan bearing
- Insulated bearings, yet they lasted only one or two months
Assembly and operating errors (2)
- A hot-box detector said all clear, then the axle broke
- A five-year bearing failed in two after an overhaul
Alignment and thermal growth (1)
Contamination (1)
- A brick-factory mill bearing worn out in a year and a half
Measurement and operating point (2)
Method and limits
- The cases are summaries of investigations and monitoring programmes from industrial plants. Customer names are removed.
- One main theme is assigned per case. Many cases involve more than one cause.
- The sample is small and was not chosen at random. Do not read the percentages as industry rates.
- Reductions are calculated from the before and after values quoted in each case.
Questions
What is the most common cause in these field cases?
Lubrication and oil condition (7 cases) and early bearing defect detection (7 cases) are the largest themes, followed by looseness, soft foot and weak supports (5 cases).
How much can fixing looseness reduce vibration?
In the three looseness-type cases with before and after readings, vibration fell by 78%, 81% and 87%.
Can I use these figures as industry averages?
No. The collection is small and not random. It shows what recurred in these cases, not how often each fault occurs across industry.