The night shift clocked in at 22:10 on a machine that had been sitting cold for sixty hours. By 06:00 they had machined 340 valve bodies, and 340 of them had a bore that would not accept the customer’s gauge.
Nothing broke. No tool chipped. The program was the same one that had run the previous Thursday with a Cpk of 1.42. What changed was the temperature of the machine itself.
The part, and why 13 microns mattered
The part is a 316L stainless valve body, about 2.1 kg of bar turned down to a 92 mm envelope. The feature that mattered is a Ø25 H7 bore — 25.000 to 25.021 mm, 0.8 µm Ra — bored on a 40-taper VMC with a carbide bar. Our operator gauges every 25th part with a bore gauge that is checked against a setting ring at the start of each shift.
The bore was the only feature out of tolerance. Every diameter, face and thread on those 340 parts was inside print. That is the signature of a machine problem rather than a process problem, and it is why we spent two days chasing it.
What the log actually showed
We pulled the probe records and the shop temperature log for that night. The spindle housing started at 24.8 °C and finished at 34.4 °C. Z-axis repeatability, measured with a test bar and a 1 µm indicator the next morning, had moved 31 µm negative over the same window.
| Time | Spindle housing | Bore average | Against H7 low limit |
|---|---|---|---|
| 22:10 | 24.8 °C | 25.010 mm | +0.010 |
| 00:00 | 28.9 °C | 25.007 mm | +0.007 |
| 01:30 | 31.4 °C | 25.002 mm | +0.002 |
| 03:00 | 33.1 °C | 24.997 mm | −0.003 |
| 04:30 | 34.0 °C | 24.993 mm | −0.007 |
| 06:00 | 34.4 °C | 24.989 mm | −0.011 |
The bore walked 21 µm in eight hours, monotonically, in step with the spindle temperature. The H7 lower limit on a Ø25 bore is 25.000 mm, so the first bad part left the machine at roughly 03:00. Everything after that was out by 3 to 11 µm — a rejection a caliper will not catch and a plug gauge will.
Why the day shift never saw it
This is the part that took us longest to accept. The same machine had run the same part all the previous week and never drifted more than 8 µm. It was not a shift problem. It was a cold-start problem. During a normal week the machine idles at 30 °C overnight and comes up to 34 °C within an hour of the first cut, so the total excursion is four or five degrees. That weekend it sat through a 60-hour shutdown and started from ambient. The excursion was 9.6 °C, and the last three degrees are where the bore crossed the limit.
Steel moves about 12 µm per metre per degree, and there is close to a metre of structure between the spindle face and the column base, with the Z ballscrew adding its own growth on top. A bare column would have moved close to 115 µm over that range. Ours moved 31, because the base casting lags the head, the screw is anchored at the bottom, and the coolant carries some of the heat away. Thirty-one microns is what reaches the tool tip — more than the entire 21 µm band on that bore.
The bill
| Item | Basis | USD |
|---|---|---|
| Machining already sunk | 340 pcs × 18 min at USD 52/h | 5,304 |
| 316L bar consumed | 714 kg, budget estimate at USD 11.80/kg | 8,425 |
| Scrap credit from the recycler | 714 kg returned | −1,120 |
| Remake on overtime | 340 pcs × 18 min at USD 78/h | 7,956 |
| Total | 20,565 |
We looked at boring the 340 parts out to Ø28 and pressing in a sleeve. It would have been a stainless-to-stainless interference fit in a wet service, and the customer’s valve body is a pressure part, so we scrapped them instead. USD 20,565 for a machine that was simply colder than it had been the week before.
The 20 minutes we now spend
A warm-up cycle runs before the first article, not after it. Twenty minutes at 60 % spindle speed with the axes cycling through the work envelope, then a test-bar reading on Z. More than 8 µm off the morning baseline and the machine gets probed before it cuts anything that matters. We had been skipping it on Monday nights because the machine was needed.
The coolant chiller is held at 21 °C, and the night shift gets a written machine temperature at start-up. A cold start below 26 °C means warm-up first, no exceptions, and the operator signs for it. A rule that lives only in somebody’s head is not a rule.
We probe the critical bore every 25 parts, in process. Not because we expect it to move, but because the probe tells us it has not. The measurement costs 12 seconds a part. The scrap it prevents cost us 18 minutes a part.
What this means if you are buying the parts
Ask your supplier how they handle the first hour after a weekend. It separates shops that think about it from shops that do not.
Ask for first article inspection data from the beginning of the run and from the end. A first article taken 40 minutes in tells you what the shop holds when it is least capable. One taken at hour six tells you what it holds when it is hot. If both are inside print with margin, you have a process. If only one is, you have a lottery.
Ask for Cpk on the critical feature across the whole run rather than a pass/fail list. A feature averaging 25.005 mm with a standard deviation of 0.002 is a completely different risk from one averaging 25.002 with a standard deviation of 0.008, and both print as “in tolerance” on a certificate of conformance. If a feature has a band narrower than 25 µm, say so up front. That is a conversation about temperature, gauging and capability, not about the positioning accuracy on a spec sheet.
Our tolerance chart lists what we hold as standard by feature and size. If the band you need is genuinely tighter, it belongs on the first article inspection sheet as a flagged critical, with gauging agreed before the run starts. The boring above now runs on our CNC milling cells with an in-process probe.
The other half of the problem
Machine growth is only one side of it. The workpiece grows too, and on a 400 mm aluminum part a 10 °C swing in shop temperature is worth about 90 µm on its own. Our guide to thermal expansion and tolerance covers that side of the equation. If you are specifying a tight feature on a large part, both effects are in play at once, and they do not always cancel.
We now run the warm-up on every cold start, and the bore has not gone out since. Send the drawing through the instant quote page if you want to know which features on your part we would flag as thermally sensitive. The team confirms within one business day, and any material figure above is a budget estimate rather than a rate, because 316L moves with the market.