Welcome to WordPress. This is your first post. Edit or delete it, then start writing!
Author: Codeller
Why your ballscrew is noisy but the pitch reads fine
Every month we get a call that starts with the same sentence: the pitch is still in spec, but it sounds wrong under load. Nine times out of ten the noise is not the lead. It is the recirculation.
What the numbers hide
A laser interferometer measures positional accuracy along the axis. It tells you nothing about the return path inside the nut, where the balls swap lanes. Once a return channel starts to score, the balls no longer roll cleanly. You get audible tick, a temperature rise after twenty minutes of continuous cutting, and a small but stubborn increase in following error at rapid.
What to check before you condemn the screw
- Take a listen at idle rapid, both directions. A one-sided rumble is a strong tell.
- Log axis load over a two-hour production run. A creeping baseline points at the nut, not the drive.
- Pull the wiper covers and look for grease that has gone chalky rather than glossy. Chalky means it has been running hot.
Ballbar readings looked fine. Vibration analysis flagged the Y axis inside a week. That was the nut, not the screw.
If it is caught early, a nut-only rebuild is a two-day turnaround and a fraction of the cost of a full screw and support-bearing pack. Left running, you are on borrowed time before it seizes mid-cycle.
Rebuilding a Big Plus spindle without losing the taper
A Big Plus interface only earns its keep if the face contact is perfect. The extra rigidity comes from the flange kissing the spindle nose at the same moment the taper seats. Miss that by a few microns and you are running a standard BT-40 with more expensive tooling.
Where rebuilds usually go wrong
Most shops know to protect the taper during teardown. Where we see mistakes is the reassembly:
- Bearings pressed on with a machine-shop arbor press instead of a proper induction heater. Cold-pressing a preloaded angular contact set is how you cook a race before the machine ever cuts a chip.
- Grease packed by feel, not by weight. Every spindle we ship goes out with the manufacturer's grease-charge weight recorded on the paperwork.
- Runout checked at the nose only. On Big Plus you have to check runout with a master gauge seated, otherwise the face condition is invisible.
What a good rebuild looks like on paper
A rebuild we sign off comes back with: bearing lot numbers, preload figures at cold and at 30 minutes running, TIR measured at four axial positions, and a run-in chart from the vibration monitor. If the paperwork is thin, the rebuild is thin.
The customer on this job runs aerospace ally at 18,000 rpm all day. Two years on the rebuild, still holding sub-3 µm runout at the nose.
Field notes: a Fanuc 0i-MF that only faulted at 3am
The call was for a Fanuc 0i-MF throwing SV0410 (excess error at stop) on the Z axis. Only on nights. Days ran clean. Two other engineers had been out and reset the parameters, no change.
What the machine was telling us
The alarm log said the fault fired within 30 minutes of shift change every night, then never recurred until the following night. Both engineers had chased the servo pack. Nothing there was wrong.
What clinched it: we asked the shift lead to walk the floor at 2am and note anything that felt different. His answer was that the compressor started cycling harder around then because a neighbouring unit went offline overnight. Voltage sag on the incoming supply was pulling the servo bus low just long enough to trip the safety threshold.
Fix
- Installed a line reactor on the incoming three-phase.
- Rebalanced loads with the site sparky so the CNC cell is not the last thing on the ring.
- Left the alarm history filtered for SV0410 so we would see it come back. Six weeks in, nothing.
Moral: not every alarm on a Fanuc is a Fanuc problem. Half of intermittent faults are power quality, and you will not find them by staring at the servo tune.
The relocation checklist we actually use on aerospace jobs
Aerospace customers do not want to hear that the machine that was cutting 30 µm TP last week now needs a week of ballbar work because a slinger caught the way covers. Here is what we tick off before the hook takes weight.
Before disconnect
- Baseline ballbar and laser interferometer run, filed and signed.
- Photograph the levelling pad heights and torque marks on every anchor bolt.
- Drain and cap the coolant and hydraulic circuits. Label the caps.
On the lift
- Lift points confirmed against the manufacturer's installation manual, not the operator's memory.
- Slings padded at every contact with a hard edge. Way covers wrapped in shrink film, not just packing tape.
- Spindle locked with the OEM shipping bracket. Not a bit of dowel and a prayer.
After set-down
- Levelling to 0.02 mm/m across the bed before any electrical goes back on.
- Warm-up run to spec, 60 minutes minimum, before geometry checks.
- Repeat baseline ballbar and interferometer. If numbers drift more than 20% from the pre-move report, we stay on site until we know why.
Every step gets a signature and a timestamp. On MOD-adjacent work that paperwork is not overhead, it is the reason the customer sleeps.
Renishaw ballbar: reading squareness errors without over-correcting
Ballbar is the fastest way to get a machine graded end to end. It is also the fastest way to talk yourself into work that does not need doing. Squareness is where most of the over-diagnosis happens.
What the plot really means
A squareness reading of, say, 30 µm/m looks alarming until you rerun the test at half feed rate and it drops to 12 µm/m. The delta is not squareness. The delta is servo lag showing up as a geometry error on the plot.
Before we quote for a scraper and a levelling job, we always run:
- Two feed rates at the same radius. If the number moves with feed, it is not geometry.
- Two radii at the same feed. If reversal spikes get worse at a smaller radius, backlash and stick-slip are in play.
- A quick axis-by-axis linearity check. Cheap to do, saves the customer the price of a full geometric alignment when the ballscrew nut is the culprit.
Ballbar reports are diagnostic. They are not a work order.
The rule we live by: no squareness correction is signed off until the number is stable across at least two feed rates and two radii. Otherwise you are chasing an artefact and the customer is paying for the chase.
Coolant mist extraction: what the HSE audit will look for in 2026
The workplace exposure limit for MWF mist is still 1 mg/m³. That has not moved. What has moved is what an HSE inspector will now expect to see on the wall next to the machine.
What's different in 2026
- LEV thorough examination reports every 14 months, not “annually”. Inspectors are pulling dates now and calculating gaps.
- Face-fit test records for anyone still relying on RPE, refreshed every two years or after any facial change (weight, dental work, beard).
- Written scheme of examination for the fluid itself, not just the extractor. That is new for a lot of shops.
Practical fixes
If you inherit a cell with no paperwork, the fastest way to a compliant position is:
- Book an LEV test through a P601-competent tester. Two-week lead time is typical.
- Move the LEV logbook out of the office drawer and onto the machine. Inspectors want it at the point of use.
- Sign off a fluid-management written scheme. A one-pager is fine as long as it names who is doing what, and how often.
The audits we have supported in the last twelve months have all gone well when the paperwork was on the wall, and badly when it was in a file the operator could not find. That is worth more than any spec sheet.
Case study: bringing a 15-year-old YLM back into cycle time spec
The customer bought this YLM new. It had been on continuous shift work for 15 years cutting mid-run production parts. Nothing had failed. The problem was subtler: cycle time on the reference part had crept from 3:12 to 3:55 over three years and management wanted to know why.
What we found
- Rapid rates on X and Y were down to 32 m/min from the rated 40. Servo tune had drifted with age.
- The tool magazine was firing slower than spec. Air pressure at the ATC was at 4.9 bar, not the required 6.
- Spindle warm-up was taking 40% longer to hit thermal equilibrium. Cooler had a slow leak.
What we did
No parts were failing, so nothing needed replacing outright. The job was tuning:
- Rebalanced the servo loops on X, Y and Z. Restored rapids to spec without going near the mechanical side.
- Replaced the ATC solenoid and repaired the air supply line. ATC time down from 4.1 s to 2.6 s per tool.
- Recharged the spindle cooler and replaced the flow switch. Warm-up back to under 20 minutes.
Result
Cycle time on the reference part came back to 3:16, four seconds off the day-one figure. On a run of 800 parts a shift, that is 27 minutes of recovered production per shift. Two shifts a day, 240 shifts a year: 108 hours of extra spindle time.
Anti-bacterial tank wash: when to schedule, when to skip
Anti-bacterial machine and tank wash is not something you do on the calendar. You do it when the coolant is telling you it is time. The rest is money you did not need to spend.
Signs the sump wants a wash
- pH creeping below 8.6 and refusing to come back up with a top-up.
- Rotten-egg smell first thing on a Monday, gone by mid-shift. That is anaerobic activity in the low-flow zones of the tank.
- Dermatitis reports from the operators. That one is not optional. Book the wash.
Signs the sump is fine
- Refractometer reading holding steady week to week.
- Skimmed tramp oil coming off cleanly and consistently.
- No visible biofilm on the return pipe when you dip a torch in.
What a proper wash includes
If it is scheduled, the job we quote covers: pump-out, mechanical clean of the tank walls, chase every dead-leg with the sanitiser, refill and re-inhibit, and a follow-up refractometer check at 48 hours to confirm the ratio held. Anything less is a rinse, not a wash.
The shortest-lived washes we see are the ones where the tank got cleaned but the machine bed did not, so the biofilm was seeded back in on the first cycle.
MOD job? What defence contracts expect from your service records
Defence work used to be about the finished part. Now it is also about the machine that cut it. If a prime cannot see a clean maintenance trail on the equipment, they will find another supplier who can show them one.
What they will ask to see
- Servicing schedule per machine, with dates, engineer names, and the parts touched.
- Calibration certificates that trace back to a UKAS-accredited standard, not “we did it in-house”.
- A list of any interventions in the last twelve months that could have affected geometry, with the ballbar or laser report attached.
- Fluid management records if the parts are corrosion-sensitive.
What good looks like
The customers who breeze through these audits share three habits:
- One logbook per machine, next to the machine, filled in on the day, not the week after.
- Service reports scanned to a shared drive within 48 hours of the engineer's visit.
- An annual review meeting where the maintenance history is read alongside the production log. If a scrap spike lines up with an unrecorded service call, you find it there, not in an audit.
None of this is expensive. It is just a decision to write things down at the time. On a defence contract, that decision is what keeps you on the list.