Work log — Glaze pump project, spindle bearing preload week
Engineer: M. Ibera (rotating machinery)
Project: GP-15 ceramic glaze pump, 15 kW direct-drive spindle
Log period: Day 01 to Day 05
Document ref: WL-2026-0911-BP07
Day 01 — morning stand-up, the spindle that heats like a stove
Stand-up 08:20, plant floor. The GP-15’s new direct-drive spindle runs its bearing housing at 62 C after forty minutes, the ceramic glaze pump loses delivery pressure at the hot end, and the maintenance supervisor wants to know whether it is the preload or the cooling line. The spindle carries a pair of angular contact bearings back to back at the wet end, a 45 mm bore, 25 degree contact, and the motorized spindle bellows a measured 16 microns of axial growth from cold to warm. The thermal camera in the morning showed the heat concentrated at the bearing housing, not the motor windings, so the wiring is innocent and the bearing is the suspect, and preload is the favourite explanation before a single nut is turned.
Morning I did the discipline that saves a teardown: I measured, rather than assumed. The bearing axial clearance with the preload nut slack: 42 microns at cold, which is a normal value for a lightly preloaded pair, so the factory did not over-tighten; the 62 C is emerging from somewhere else in the train. I pulled the thermal record from the drive: the spindle draws 28 percent more power at the malfunctioning speed than the acceptance test two months ago recorded, 4.6 kW against 3.6 kW, and that extra kilowatt has not been explained by the glaze load, the pump delivered the same volume on the test.
Evening note: an unexplained half-kilowatt of bearing power is the fingerprint, and it smells less like preload and more like a contact condition, a retainer wear or a seal interference, that surfaced after the recent overhaul. Tomorrow the penalty is a bearing removal, and I will not return the same bearing until I can point at the part that caused the power.
Day 02 — the bearing on the bench tells the truth
Bearing removal at 08:40, clean room corner of the pump shop. The pair came out as a matched set, and the bench inspection was quiet until I rolled the front bearing slowly by hand: the cage clicks, a copper cage with a worn pocket, the ball pockets are 0.9 mm oversized in the radial direction, and there is copper transfer smeared on two pockets, textbook cage pocket wear from a churned ball set. The wear pattern spills the story – the retainer was running at speed against a ball set that was not rolling cleanly, and the smear explains the half kilowatt and the 42 micron slack that was not preload at all but cage chaos. The bearing’s own rolling elements and races are clean, no spalling, no heat tint, so the unit is salvageable with a new cage if the supplier can fabricate one, otherwise a matched replacement.
Decision recorded at 11:00: order a matched pair from the bearing supplier, express, and in parallel have the shop machine a nylon retainer for the old pair as a maintenance spare, because a glaze pump cannot wait on customs for a cage. The replacement pair arrives Thursday; until then the pump runs at reduced speed with the salvage-bearing temperature monitored hourly, and the process freezer has enough dressed glaze for one caution day.
Afternoon I measured the bore seat and the shaft shoulder to remove excuses: the housing bore runs true within 1.5 microns of roundness, the shaft shoulder seats within 2 microns of square, so the mounting is boring-book clean and the only fault in the assembly was the cage itself. Preload re-set would have been pointless on a bearing whose balls are swimming, and I logged that sentence in bold.
Evening thought: the kilowatt was the witness, the cage was the criminal, and the preload measured 42 microns of innocence the whole time. The next time a plant blames preload I will reach for the power trace first, because crime is loud before it is hot.
Day 03 — preload method, actually, since the week is named after it
Thursday morning the matched pair landed, and the day became an honest preload exercise after all. The new pair is a factory preloaded set, and I installed it with the correct axial preload method for a wet-end glaze pump: a spring-loaded axial preload rather than a rigid nut lock, because a glaze pump sees intermittent contact with a viscous medium and its spindle cycles through 16 microns of thermal growth, and a rigid preload at cold would climb to a rigid clamp at warm. The preload stack is three disc springs under a jam nut, and I set the stack to deliver 220 N axial preload at cold, which the disc spring data predicts will soften to 150 N at the 60 C operating point, holding the pair in the light preload zone without ever releasing to clearance.
Afternoon I measured what the book promised: the axial deflection of the spindle nose under the 220 N preload came back at 11 microns, matching the bearing supplier’s curve within a micron, and the bearing temperature at empty running sat at 41 C after an hour, down twenty one degrees from the failing run. The power trace followed: 3.5 kW at the same delivery rate, back on the acceptance line, and the drive no longer has a half-kilowatt secret to hide.
Evening note: the spring-loaded preload is not the hero, the correct preload is, and the spring is only there to keep the preload correct when the machine grows warm. A rigid preload would have shipped the 41 C result cold and delivered a 70 C spindle after lunch, so the selection belongs in the same sentence as the thermal budget, not bolted on as an afterthought.
Day 04 to Day 05 — the glaze run under the new front bearing
Day 04 the pump ran the glaze at full production rate for the full shift. Bearing housing peak 58 C in the afternoon, against the 62 C failure and the 70 C rigid-preload projection, and the delivery pressure held flat at 6.8 bar all day with no mid-shift sag. The thermal camera scans hourly showed the heat gradient moving away from the wet-end bearing, which is the signature of a contact condition resolved, not simply a cooler pump.
Day 05 I ran the diagnostics that close the case. Cold-to-warm axial growth on the new pair: 14 microns, within 2 of the designed 16, and the preload stack, measured by its nut torque lift at cold and warm, shifted from 220 N cold to 158 N warm, essentially the designed 150 N figure corrected for the glaze temperature being a touch higher than the estimate. A scope run on the current draw shows no ripple above 4 percent at rated speed, the retainer chatter is gone, and the salvage-pair spare with its machined nylon retainer is racked and labelled for the next unplanned Tuesday.
Closing the pump shop week with the numbers: bearing housing from 62 C to 41 C empty and 58 C loaded, wet-end power from 4.6 kW to 3.5 kW, axial growth 14 microns on the new stack, preload 220 N cold to 158 N warm, and delivery pressure a flat 6.8 bar across the whole glaze run. The damning evidence at the bench was a copper cage with 0.9 mm of pocket slop and a smear, and the moral for the maintenance file is that the preload nut is not always the criminal, sometimes it is the innocent who lives next door to the crime.
Attachment A — bearing condition and preload record
| Check | Failed pump | Rebuilt pump |
|---|---|---|
| Bearing housing, hot | 62 C | 41 C empty / 58 C loaded |
| Wet-end drive power | 4.6 kW | 3.5 kW |
| Cold-to-warm axial growth | 16 microns question | 14 microns |
| Preload type | rigid nut | disc spring 220 N to 158 N |
| Bore roundness | 1.5 microns | 1.5 microns (unchanged) |
| Delivery pressure | sagging | flat 6.8 bar |
Attachment B — bearing preload investigation checklist
1. Read the power trace before the temperature. 2. Measure cold axial clearance before judging the preload nut. 3. Bench-roll the bearing and listen for the cage. 4. Verify bore roundness and shoulder squareness before blaming the bearing. 5. Select spring-loaded preload when the axis cycles through thermal growth. 6. Record preload at cold and warm, the difference is the whole point.
Glossary
Angular contact bearing: a bearing whose raceways meet the balls at an angle to carry combined axial and radial load. Preload: a deliberate axial load applied to remove clearance and stablise the rolling elements. Cage pocket: the window in the retainer that holds one ball. Thermal growth: the axial expansion of the spindle as the machine warms. Matched set: two bearings ground and selected to share preload as a pair.
Week-end reflection
Friday evening, a short sermon. I nearly re-preloaded a bearing whose retainer was dying, and the only reason I did not is that I asked the drive how much power the spindle was eating before I asked the bearing how hot it was. The plant’s first assumption, preload, was wrong, and mine, contact condition, was right, because the 0.9 millimetre of cage pocket slop was visible on a slow hand roll while a torque wrench could never have found it. The rebuild now runs on a spring that keeps the preload honest across the thermal day, and the spare rack holds a nylon-cage pair that will outlive the customs delay. The glow in the log is not the ceramic glaze, it is a pump that finally tells its true temperature.
One handout note, since the seminars will quote it: a one-degree Celsius rise in a bearing race shortens the grease life by roughly ten percent, so the twenty one degree drop at empty running is not cosmetic, it multiplies the regrease interval that the plant can honestly promise. On a bearing running at the 40 to 60 C band, holding the contact cool by fixing a cage, not by skimping on preload, is the difference between a quarterly squirt and an annual service. And for the record of this week, the pair’s rated 45 mm bore, 25 degree contact and 220 N spring preload are now stamped on the pump’s PM card so the next engineer reads the intended preload before they reach for the rigid nut.
A note for the vibration collector: the rebuilt pump’s envelope spectrum at the wet-end bearing shows the ball-pass frequency family sitting at the noise floor, where the failing pump carried a clear 3.1 kHz line from the cage pocket wear. The collector kept that line, and the rebuild’s clean spectrum is now the baseline in the machine’s trending file, so the next contact-condition failure will announce itself three weeks before it gets hot, if anyone remembers to listen. That is the quiet gift of the week: the pump is not only cool, it is also legible.
