Quality · Shaft final grinding AI-assisted A3 Report

Reduce roundness defects in shaft final grinding.

Rockford machining site · PS-24 pump shaft bearing journals · sixteen weeks to week 16

Report PS-24-R · Revision 1Steps 01–08 composedSources & notes
01

Clarify the background

Site / location

Rockford machining site
Rockford, Illinois

Main products

Pump shafts · Valve spools · Bearing sleeves

Report focus

PS-24 pump shaft
Final grinding of bearing journals, G-1 and G-2 → MS-02

Process flow

Process context · schematic · click a box for detail
TURNblank to sizeHEAT TREATharden journalsWASH · PACKto stock
MS-02 records the departure; the breakdown below establishes where it is created. S1

Final-grind first-pass rejects — weekly trend

Rejected shafts / completions (%)
0% 1% 2% 3% Standard ≤0.5% Final grind Wk 1 Wk 4 Wk 7 Wk 10 Wk 13 Wk 16
Standard: site quality plan QP-24, ≤0.5%. First week above standard: Wk 6. Latest: 2.5%, a gap of +2.0 points. Flat for five weeks, then rising every week. S1

Defect types — week 16

Pareto · 75 rejected shafts
0 20 40 60 80 0% 50% 100% 68 Roundness 3 Diameter out of tolerance 2 Surface finish 2 Taper 90.7%
MS-02 primary reject codes. Bars: rejected shafts. Line: cumulative share. S1

Selected priority: roundness — 90.7% of final-grind rejects in week 16, and the category carrying the rise.

02

Break down the problem

Roundness rejects over time — by machine

Roundness rejects / completions (%) · weekly
0% 1% 2% 3% 4% 5% G-2 G-1 Wk 1 Wk 4 Wk 7 Wk 10 Wk 13 Wk 16
G-2 rises every week from week 6 to 4.4% in week 16 (66/1,500); G-1 stays at about 0.1%. Blanks reach both machines at 21–24 µm from the same lots; G-1 grinds them to 1–2 µm, G-2 does not. Within G-2 no difference by shift, operator or part number. S1 S3 S6 detail

Roundness per journal

G-2 rejects · week 16
0 5 10 µm Limit 3.0 µm traced below J1 J2 J3
Out of round across all three journals. One line per rejected shaft; 63 of 66 over on all three, by about the same amount. S3

The departure — journal roundness against the drawing

The shaft → the three journals → the roundness profile

Where on the shaft the requirement applies, and what the measured departure looks like against it, journal by journal.
Interactive 3D
J1J2 · Ø30 mmJ3 J1J2J3
Gold = selected journal · green rings = 3 µm tolerance band on every journal · red = reject profile, exaggerated
ProfileJournalViewLoading 3D…
Drag to rotate · scroll to zoom. Profile: which measured ring is drawn; the tolerance band is the 3 µm limit. Journal: which journal the camera and the traces at right follow. Illustrative geometry; amplitude scaled for visibility, not to µm.
The product standard · PS-24 rev D, note 4
3 µm maximum

Roundness on each bearing journal, peak-to-valley against a least-squares reference circle. MS-02 in-line roundness gauge expanded uncertainty ±0.3 µm. S2

Accepted · G-1 · J2
1.8 µmpeak-to-valley ±1.5 µm band = 3 µm limit · 0° at right
Shaft G1-W16-2231, J2. No dominant lobe.
Rejected · G-2 · J2
11.4 µmpeak-to-valley ±1.5 µm band = 3 µm limit · 0° at right
Shaft G2-W16-1104, J2. Two-lobe (oval) form, 3.8× the limit.

Polar traces exported from MS-02, week 16, day 3; same day, different machines. On the rejected shaft all three journals are over the limit with the lobes at the same angular position. S4

Specific problem
Bearing journals J1, J2 and J3 ground on G-2 final grind measure up to 14 µm roundness against the 3 µm limit.
03

Set the target

Target
Final-grind rejects from 2.5% to ≤0.5% by Wk 20, every journal within 3 µm.
04

Analyze the root cause

Why 1

Why are the G-2 journals out of round? Because the grinding wheel spindle has excessive runout: 4.8 µm against the 2 µm limit.

wheelhead012345dial indicator4.8 µm TIRlimit ≤2 µm
Wheel-spindle radial runout
G-24.8 µm
G-10.6 µm
limit≤2 µm machine limit

Dial test indicator on the spindle at the standard check point, 50 mm from the housing face, spindle at grinding speed after 30 min warm-up, week 16. Total indicator reading, once per revolution. S8

Why 2

Why does the spindle have excessive runout? Because the front bearing pair is worn: raceways pitted, radial play in the pair. The rear pair is sound.

Front bearing pairraceways pitted · grease milky · radial play
Rear bearing pairclean · no play

Spindlehead inspection, week 16, G-2. S9

Why 3

Why is the front bearing pair worn? Because coolant is in its grease: emulsified grease no longer carries load and the raceways pit. Ingress is small per cycle; water builds over years until the film fails, then wear accelerates.

G-10.3 %G-26.8 %
Water content, front-bearing grease
G-26.8 %
G-10.3 %
limit≤1 % grease maker limit

Spindlehead grease analysis, week 16: water content by Karl Fischer titration, front and rear cavities, both machines. S9

Why 4

Why is coolant in the grease? Because a small amount passes through the labyrinth seal at every stop: as the housing cools, the cavity pressure drops and draws pooled coolant and mist through the clearance.

wheelheadgrinding wheelshaft · journalcoolantnozzlesplash from the contactonto the wheelhead frontlabyrinth seal · penetratedfront bearing pairgrease emulsifiedrearbarrier air line · 0.1 bar at the port

Schematic sections, not to scale. Observation log S10, barrier air check S11

Why 5

Why does coolant pass through the labyrinth? Because the air curtain (barrier air) at the G-2 seal is too weak, 0.1 bar against the 0.3 bar it needs: G-2 is fed in series from the one regulator, set to give G-1 its 0.4 bar, and the pressure falls along the branch.

supplyone regulator · set for G-1Plow-pressure switchreads 0.4 bar · no faultG-10.4 bar at the sealG-20.1 bar at the sealregulator and switchneeded at each machine

Gauge at the seal port, week 16. Limit ≥0.3 bar at the seal. Barrier air line layout S11 S12

Other causes checked and cleared: suspects checked.

Root cause · why it occurred

Each grinder has one air connection, and the builder's manual states the barrier air pressure required at the seal, 0.3 to 0.5 bar. At installation both machines were fed from one regulator, set to give G-1 its pressure. The plant's procurement specification and installation acceptance did not require a regulator at each machine or a pressure check at each seal port.

Root cause · why it was not caught

The line's only low-pressure switch sits on the supply ahead of G-1. It reads G-1's pressure and never sees the G-2 branch, so no fault tripped and every PM passed.

05

Develop countermeasures

#AnswersCountermeasureKindOwnerDue
TCustomerContainment: G-2 stopped from week 16; all PS-24 shafts ground on G-1; MS-02 stays at 100 % coverageTemporaryProduction supervisorWk 16
1Why 1–2Replace G-2 front bearing pair, flush the cavity, regrease; verify runout ≤2 µm at the standard check pointRestoreMaintenanceWk 17
2Why 5Regulator at each machine, set 0.3–0.5 bar per builder manual; gauge at each seal portPrevent recurrenceMaintenanceWk 17
3EscapeLow-pressure switch at each seal port wired to that machine's controller: no spindle start below 0.3 bar; alarm, stop and andon on loss during runDetectControls engineeringWk 18
4Why 4Stop-sequence standard: coolant off before spindle stop; barrier air held through cooldown by timerPrevent recurrenceControls engineeringWk 18
5OccurrencePurchasing standard: equipment specification requires per-machine air conditioning per the builder manualPrevent recurrencePurchasing · plant engineeringWk 19
6OccurrenceInstallation acceptance standard: pressure verified at every seal port, switch trip tested, both recordedPrevent recurrencePlant engineeringWk 19
7Why 3, 5PM items: barrier air pressure at each port on the operator daily check, range marked on the gauge; grease water content at interval; spindle runout at intervalDetectMaintenance plannerWk 18
8EscapeMS-02 roundness rejects trended per machine with an action limit, so a rise is acted on before the site limit is crossedDetectQualityWk 17

Each row answers one rung of the Step 4 ladder or one of the two root-cause boxes. T is the temporary item that protects the customer while the permanent items are built.

Options considered

Barrier air at the G-2 seal (Why 5)
  • Raise the one regulator until G-2 reaches 0.3 bar rejected · G-1 would run above the 0.5 bar maximum
  • Separate air line from the supply to G-2 rejected · new pipe run, same single switch problem
  • Regulator and gauge at each machine chosen · each seal set independently, checked at the port
Catching a loss of barrier air (escape)
  • Gauge at each port on the daily check only rejected · administrative; a loss between checks runs unseen
  • Low-pressure switch at each port wired to the controller chosen · no start below 0.3 bar, stop and andon on loss

Plan agreed week 17 by production, maintenance, quality, purchasing and plant engineering; temporary item first, permanent items behind it.

06

See countermeasures through

Wk 16Wk 17Wk 18Wk 19Wk 20targetT · Containment: all shafts on G-11 · G-2 bearing pair replaced2 · Regulator and gauge at each machine3 · Pressure switch at each seal port4 · Stop sequence in the controller5 · Purchasing standard issued6 · Installation acceptance standard7 · PM items added8 · MS-02 chart with action limit
Planned week as a dashed outline, actual week filled; red where an item ran late. Temporary item first, permanent items behind it.
#ItemPlannedActualConfirmed at the machine or by documentStatus
TContainment: G-2 stopped, all shafts on G-1Wk 16Wk 16G-1 weekly roundness rejects 0.1 %, weeks 16–17; MS-02 coverage 100 % S13in place
1G-2 front bearing pair replaced, cavity flushed, regreasedWk 17Wk 17Runout at the standard check point 0.7 µm (limit ≤2 µm) S13in place
2Regulator and gauge at each machineWk 17Wk 17At the seal port: G-1 0.40 bar, G-2 0.35 bar (spec 0.3–0.5) S13in place
3Low-pressure switch at each seal port, wired to the controllerWk 18Wk 19Trip test by bleeding each port to 0.25 bar: no spindle start; stop and andon during run. Switch delivery one week late S14in place · late
4Stop sequence: coolant off before spindle stop; barrier air held through cooldownWk 18Wk 18Controller sequence: coolant off 10 s before spindle stop; barrier air held 15 min after stop, timer log checked S14in place
5Purchasing standard: per-machine air conditioning per builder manualWk 19Wk 20Equipment specification ES-114 rev C issued S15in place · late
6Installation acceptance: pressure at every seal port, switch trip testedWk 19Wk 19Acceptance checklist IA-07 rev B issued S15in place
7PM items: daily port pressure check, grease water content, spindle runoutWk 18Wk 18Daily check sheet rev 4 with marked range; grease water content six-monthly; runout quarterly S15in place
8MS-02 roundness rejects by machine with an action limitWk 17Wk 17Weekly chart carries a 0.5 % action limit per machine; quality reviews it weekly S15in place

Progress reviewed weekly at the production meeting: the maintenance manager reports each item, the plant engineering manager confirms it at the machine before it is marked done. An item is marked in place only on that confirmation, never on a report that it is done. Results against the target wait for Step 7.

07

Evaluate results and process

Final-grind first-pass rejects by machine — weeks 1 to 24

Rejected shafts / completions (%) · weekly
0% 1% 2% 3% 4% 5% Site limit ≤0.5% G-2G-1 Wk 1 Wk 5 Wk 9 Wk 13 Wk 17 Wk 21 Wk 24 1 & 2 4 & 7 3 & 6 5 Countermeasures implemented ↓
Target: ≤0.5% by Wk 20. At Wk 20: G-1 0.33%, G-2 0.27%. Week 24: G-1 0.33%, G-2 0.27%. Implemented (Step 6 item numbers): Wk 17 bearings replaced (1), regulator and gauge at each machine (2); Wk 18 stop sequence (4), PM items (7); Wk 19 pressure switches (3), acceptance standard (6); Wk 20 purchasing standard (5). S16
Accepted · G-1 · J2 · Wk 24
1.7 µmpeak-to-valley ±1.5 µm band = 3 µm limit · 0° at right
Shaft G1-W24-2210, J2.
Accepted · G-2 · J2 · Wk 24
1.6 µmpeak-to-valley ±1.5 µm band = 3 µm limit · 0° at right
Shaft G2-W24-1132, J2.

Polar traces exported from MS-02, Wk 24, day 3; same day, one shaft from each machine. S16

08

Standardize and share

Standards issued

DocumentWhat it now requiresOwnerIssued
ES-114 rev CEquipment specification: per-machine air conditioning (regulator, gauge, low-pressure switch at the seal port) per builder manualPurchasing · plant engineeringWk 20
IA-07 rev BInstallation acceptance: pressure measured at every seal port, switch trip tested, both recorded before sign-offPlant engineeringWk 19
Grinder daily check, rev 4Barrier air pressure at each seal port, range marked on the gauge, recorded by the operator each shiftProductionWk 18
PM plan G-1 / G-2Grease water content six-monthly; spindle runout at the standard check point quarterly; switch trip test monthlyMaintenance plannerWk 18
Controller standard CS-GStop sequence: coolant off 10 s before spindle stop; barrier air held 15 min after stopControls engineeringWk 18
MS-02 weekly chartRoundness rejects by machine with a 0.5 % action limit; quality reviews weeklyQualityWk 17

S17

Change point management — final-grind area

Change pointCheckWhoWhen
Spindle bearing replaced or spindle exchangedRunout at the standard check point; grease sample at 3 monthsMaintenanceBefore restart · 3 months
Any work on the barrier air line, regulator or sealPressure at the seal port; switch trip testMaintenanceBefore restart
Coolant nozzle, pressure or concentration changedPressure at the seal port; MS-02 roundness by machine watchedProduction · qualityBefore restart · next 500 shafts
Controller program change on a grinderStop sequence timing verifiedControls engineeringBefore restart
New grinder or relocated grinderIA-07 rev B acceptance in fullPlant engineeringBefore production

S18

Horizontal deployment — every machine on site with a spindle seal

MachineOperationSealFoundActionStatus
G-3 · G-4Final grind, valve spool lineLabyrinth · barrier airSeries feed from one regulator; G-4 0.15 bar at the portRegulator, gauge and switch at each machineWk 22 · done
IG-1Internal grind, bearing sleevesLabyrinth · barrier airOwn regulator; 0.35 bar at the port; no switchSwitch at the port, wired to the controllerWk 23 · done
CG-1Centerless, pre-grindContact seal · no barrier airNot applicableNone
L-1 · L-2TurningContact seal · no barrier airNot applicableNone

S19

§

Sources & notes

AI-assisted A3 example. Case facts are synthetic example data built on the structure of Art Smalley's 2006 machine-shop grinding case; they are not production records. The 3D model and polar traces are explanatory illustrations drawn from the stated values, not gauge exports.

S1

MS-02 weekly first-cycle record, weeks 1-16: 3,000 completions per week (1,500 per machine); total rejects, roundness rejects by machine, and other reject codes as tabulated. Overall rate = total rejects / completions. Site quality plan QP-24 sets ≤0.5%.

S2

Drawing PS-24 rev D note 4: roundness ≤3 µm per journal, least-squares circle, peak-to-valley. MS-02 gauge expanded uncertainty ±0.3 µm.

S3

MS-02 record for week 16 stratified by machine, shift, operator, part number and worst journal. Same incoming lots were split between G-1 and G-2 throughout. On rejected shafts, all three journals were over the limit in 97 of 102 cases.

S4

Polar traces exported from MS-02 for shaft G1-W16-2231 (accepted: J1 1.6, J2 1.8, J3 1.7 µm) and G2-W16-1104 (rejected: J1 10.9, J2 11.4, J3 11.1 µm, all two-lobe with the lobes at the same angular position on each journal), week 16, day 3.

S5

Roundness value distribution, all week-16 completions by machine, from MS-02 export; bins as shown.

S6

Incoming blank roundness sample log, weeks 12-16: 40 blanks per machine per week, 18-24 µm, no difference between machines.

S7

Lobe count review of 60 randomly selected G-2 reject traces, weeks 14-16: 47 two-lobe, 9 three-lobe, 4 five-or-more.

S8

Spindle runout check sheet, week 16: dial test indicator on the spindle at the standard check point, 50 mm from the housing face, G-1 and G-2, spindle at 3,000 rpm after 30 min warm-up; total indicator reading, once per revolution. Machine limit ≤2 µm.

S9

Spindlehead grease analysis, week 16: G-2 front bearing pair raceways pitted, grease milky; rear pair clean. Grease samples from front and rear cavities of both machines, water content by Karl Fischer titration; grease maker's limit 1 %.

S10

Wheelhead section drawing and observation log, week 16: coolant seen pooling on the G-2 nose after each cycle stop; labyrinth is non-contact and relies on the air curtain (barrier air) from the housing port to keep the flow outward; coolant runs 20 s after spindle stop on both machines.

S11

Barrier air check, week 16: gauge at the seal port, G-1 0.4 bar, G-2 0.1 bar. Barrier air line layout: one supply and regulator set for G-1, low-pressure switch at the supply ahead of the G-1 branch, G-2 branch in series downstream.

S12

Grinder builder's manual: barrier air at the spindle seal 0.3–0.5 bar, one air connection per machine. Plant procurement specification and installation acceptance record: no requirement for a regulator at each machine, no pressure measured at the seal ports. Grinder PM checklist: supply switch and gauge only, passed every month.

S13

Restart record, week 17: G-2 front bearing pair replaced, cavity flushed, regreased; runout at the standard check point 0.7 µm. Regulator and gauge fitted at each machine; port readings G-1 0.40 bar, G-2 0.35 bar. G-1 weekly roundness rejects during containment 0.1 %.

S14

Controls change record, weeks 18–19: low-pressure switch at each seal port wired to the machine controller, trip tested by bleeding each port to 0.25 bar (no start; stop and andon during run); stop sequence coolant off 10 s before spindle stop, barrier air held 15 min after stop, timer log.

S15

Documents issued weeks 17–20: equipment specification ES-114 rev C (per-machine air conditioning per builder manual); installation acceptance checklist IA-07 rev B; grinder daily check sheet rev 4; PM plan revision (grease water content six-monthly, runout quarterly); MS-02 weekly roundness chart with 0.5 % action limit per machine.

S16

MS-02 weekly record, weeks 17–24: first-pass rejects by machine (roundness and other), completions by machine; G-2 restarted week 17 after the bearing change with 500 completions that week. Polar traces exported from MS-02, week 24, day 3, J2 on one shaft from each machine.

S17

Documents issued weeks 17–20 with revision numbers: ES-114 rev C, IA-07 rev B, grinder daily check rev 4, PM plan G-1/G-2, controller standard CS-G, MS-02 weekly chart with action limit.

S18

Change point management sheet, final-grind area, week 21: change points on a grinder that trigger a check before restart, with the check, the owner and the watch period.

S19

Horizontal deployment audit, weeks 21–23: every machine on site with a spindle seal, seal type, barrier air feed, pressure at the port, action and completion week.

Suspects checked and cleared — Step 4
  • Coolant supply and nozzles — Shared tank, same concentration; nozzle settings unchanged on both machines since installation
  • Wheel balance — Residual 0.3 µm on both machines, week 16
  • Dressing — Fresh dress on G-2 left the roundness error unchanged
  • Workhead centers — Center runout 0.4 µm G-2, 0.5 µm G-1
  • Grease life and crash — Same grease and interval as G-1; no crash record; wear progressive, not sudden
Supporting breakdown detail — Step 2

How far out, and how many — week 16, every shaft measured

0 200 400 600 800 3 µm limit 0-1 1-2 2-3 2 30 3-4 20 4-6 9 6-8 4 8-10 3 10-14 Measured roundness, worst journal (µm) · shafts per bin
G-1 · 1,500 shaftsG-2 · 1,500 shafts
G-1 output sits at 1–2 µm with 2 shafts over the limit. G-2 output spreads to 14 µm with 66 over. Same incoming lots, same drawing, same gauge. S5
G-1
0.1% · 2/1,500
G-2
4.4% · 66/1,500

66 of 68 roundness rejects came off G-2: 97.1% of the roundness loss on half the volume.

Incoming blanks · both machines
21–24 µm

Turned and heat-treated blanks, sampled at receipt to final grind. No difference by machine. S6

Leaving final grind · median
G-1 1.4 · G-2 2.1 µm

G-2 95th percentile 2.9 µm. The roundness departure is created in the G-2 final-grind operation. S5

Within G-2 — does the problem follow the shift, the operator, the part number or the journal?

By shift
Shift A 4.4% · 33/750 Shift B 4.4% · 33/750
No difference.
By operator
Operator 1 4.3% · 22/510 Operator 2 4.4% · 22/495 Operator 3 4.4% · 22/495
No difference.
By part number
PS-24 4.4% · 40/900 PS-31 4.3% · 16/375 VS-12 4.4% · 10/225
All three part numbers ground on G-2 are affected alike.
By worst journal
J1 22 J2 22 J3 22
Worst journal among 66 rejects; in 63 of them all three journals were over. Ground in one setup on one wheel spindle, the journals are affected alike.

MS-02 week-16 record, G-2 only. S3 The problem does not follow people, product or journal position. It follows the machine and its setup.

Roundness rejects by machine — sixteen weeks

Roundness rejects / completions (%)
0% 1% 2% 3% 4% 5% G-2 G-1 Wk 1 Wk 4 Wk 7 Wk 10 Wk 13 Wk 16
G-2 climbs steadily from 0.1% in Wk 1 to 4.4% in Wk 16. G-1 stays at about 0.1%. S1

Shape of the departure

Of 60 G-2 reject traces reviewed: 47 two-lobe (oval), 9 three-lobe, 4 five or more lobes. Recorded as observed on the polar trace; what produces a two-lobe form is a question for the root cause analysis. S7

What is not yet known

Whether the G-2 departure varies within a shift or within a wheel-dressing cycle; MS-02 records per shaft, not per position in the cycle. Whether anything on G-2 changed around week 5; the maintenance and setup logs have not yet been pulled.

MS-02 weekly first-cycle records — both machines
WeekCompletionsAll rejectsReject rateRoundness G-2Roundness G-1Other defects
Wk 13,00090.3%216
Wk 23,000100.3%127
Wk 33,00090.3%216
Wk 43,000100.3%127
Wk 53,00090.3%216
Wk 63,000170.6%827
Wk 73,000180.6%1116
Wk 83,000230.8%1427
Wk 93,000260.9%1916
Wk 103,000341.1%2527
Wk 113,000381.3%3116
Wk 123,000471.6%3827
Wk 133,000531.8%4616
Wk 143,000622.1%5327
Wk 153,000672.2%6016
Wk 163,000752.5%6627

Download the example data (JSON)

More A3 reports · 8-step problem solving · A3 (encyclopedia) · TPS Encyclopedia · artoflean.com