Fourteen hundred 316L hydraulic manifold bodies left our floor on 26 August. A Ø16 H7 bore, 38 mm deep, reamed and deburred, 1.1 kg of stainless in each one. On 4 September the customer’s incoming inspector rejected the entire lot.
I pulled their report, then re-measured the same parts on our CMM. Their two-point bore gauge read 16.019 to 16.031 mm across the lot. H7 stops at 16.018. Our CMM, running a 60-point least-squares circle, read 15.998 to 16.011. Every part in tolerance, on our side of the argument.
Both numbers were real. That is the part nobody warns you about when you buy a CMM.
We cleared the easy explanations first
Temperature, because it is the standard excuse and I wanted it off the table. 316L expands about 16 µm per metre per degree. A 16 mm bore moving through the 4 °C difference between our inspection room and their shop floor grows roughly 1 µm. We were arguing about 20 µm. Not temperature.
Calibration, second. Their gauge carried a June certificate against a setting ring, ours a May certificate against a ceramic sphere. Both current, neither drifting.
Stylus and access, third. Ours was a 2 mm ruby at 60 points on two levels. Theirs was a two-point gauge inserted 15 mm and rocked. Different tools, same nominal feature. After a week of email we stopped asking which instrument was wrong and started asking what shape the hole actually was.
The bore was two-lobed
We ran a roundness trace. The bore was not round. It carried a clean two-lobe form, 8 to 11 µm of amplitude, consistent around the circumference and consistent part to part. Once you know what you are looking for you can feel it: rock the gauge in the hole and it picks up two high spots 180° apart.
A two-point gauge reads the widest span it can find, so on a lobed bore it reads the peaks. A least-squares circle averages the whole trace, so it reads something closer to the middle. On this feature the two definitions of “diameter” differ by almost exactly the 20 µm we spent a week arguing about. Their number was right. Ours was right. The drawing was the thing that was incomplete, because it described a size and said nothing about shape.
Where the lobing came from: a reamer with worn, uneven margins, run in a rigid holder on a machine carrying about 8 µm of spindle runout. It cut fractionally deeper on two opposite flanks than on the other two. Reaming does that. It is not a defect you can see, and a size-only inspection will never catch it.
What it cost
We ran a go/no-go plug gauge through all 1,400. The 96 parts whose low spots had gone under 16.000 could not be brought back at that size and were scrapped and remade. The other 1,304 we honed until the roundness trace closed up and the no-go side came clean, then re-inspected and shipped with the traces stapled to the report. Add air freight on the replacement set, the sorting, the honing, and eleven days of slipped schedule, and the invoice landed near USD 8,900. The part sells for around USD 41 at that quantity, which is a budget estimate for 316L at this size and not a firm number. Anything that needs a firm number goes to our team, and we confirm within one business day.
The four things we changed
One, the drawing. We wrote to the customer and added a roundness callout of 0.008 mm on that bore, because Ø16 H7 with no form control is an argument waiting to happen. They agreed. Two, the process: that feature now gets a boring pass and a light hone instead of ream-and-go, and the reamer was replaced and moved into a floating holder.
Three, the inspection plan. Every critical feature now carries a method statement: instrument, number of points or contact type, evaluation method, stylus, and the temperature the measurement is valid at. A least-squares circle and a two-point span are both legitimate measurements. They are simply not the same one, and the plan has to say which of them the acceptance number refers to.
Four, the plug gauge lives at the machine, not in the inspection room. A go/no-go gauge catches form error the moment the part comes off the spindle, which is nine days earlier than a CMM report does.
If you are the buyer on the other side of this
Ask your supplier what instrument and what evaluation method their inspection report is built on. If the answer is a CMM least-squares circle and your own incoming inspection is a two-point gauge, you will eventually have this conversation with somebody. Put the roundness or cylindricity callout on the drawing, or put the measurement method in the inspection plan, or do both. It costs one line and it prevents a lot rejection.
We have run stainless manifolds and valve bodies most months since. The material side of this part lives on our 316L stainless page, and the boring and honing process sits under CNC turning. If you are writing an inspection plan, the first-article inspection guide covers what belongs in the package, and IT6 versus IT7 tolerance explains why the grade you choose changes the inspection you end up paying for. Nominal sizes and their conversions are on the tolerance chart. Send us the drawing and the mating part and we will tell you which features need a form control before the first chip comes off. The instant quote tool returns a budget estimate to start from.