The heat treat report said “hardness 60–62 HRC, meets specification”. It was telling the truth. It said nothing at all about the shape.
Two hundred and forty D2 punch blanks for a stamping shop in Ohio: Ø32 × 65 mm, a Ø12 H7 bore straight through, two flats milled 6 mm deep on opposite sides, finished OD ground to h6. We machined them, sent them out for vacuum hardening, and left 0.30 mm on the outside diameter — 0.15 mm a side — which is the number this shop had used on hardened punches for years.
It was not enough. Here is what the micrometers found and what it cost us.
What came back from the furnace
We measure every hardened part. Diameter on a bench micrometer, bore with pin gauges, bow on a V-block under a 0.001 mm indicator over the full 65 mm.
| Measured on 240 parts, after hardening | Range | Mean |
|---|---|---|
| OD change (32 mm nominal) | +0.021 to +0.052 mm | +0.036 mm |
| Bore change (12 mm nominal) | −0.010 to +0.018 mm | +0.004 mm |
| Length change (65 mm nominal) | +0.055 to +0.110 mm | +0.080 mm |
| Bow over 65 mm | 0.005 to 0.085 mm | 0.031 mm |
| Hardness | 60.4 to 61.8 HRC | — |
The growth is the part everyone expects. D2 through-hardens with a linear expansion of roughly 0.1%, and 0.1% of 32 mm is 0.032 mm, which is almost exactly what the micrometer reported. The length grew a little more, near 0.12%. The bore hardly moved. None of that is a surprise.
The bow is the part that cost money. 0.085 mm over 65 mm does not sound like much until you remember there was only 0.15 mm of stock on that side to begin with.
Why these particular parts bent
It was not the steel. It was our own process sheet.
A symmetrical D2 blank bows by about 0.0015 mm per mm of length in a gas quench — 0.10 mm on a 65 mm part. But these blanks went into the furnace with two flats milled on opposite sides, so the section that cooled was not a cylinder. It was a cylinder with 12 mm of material missing from two faces. Thin sections lose heat first, thick sections transform last, and the part curls toward whichever side transformed first. Every interrupted section does this: keyways, milled flats, cross holes, a shoulder that is 20 mm on one end and 8 mm on the other.
Doubling the bow allowance for an asymmetric section — 0.003 mm per mm of length — gives 0.195 mm. Add half the growth and a small clean-up margin and this part needed 0.23 mm a side. We had budgeted 0.15.
Twenty-six of the 240 parts could not be cleaned up. We took the full 0.15 mm off each flank, hit the low spots, and the indicator still read 0.004 to 0.011 mm of residual bow with a bright, unground patch down one side. Those went in the scrap bin. There is no way to un-bend a 61 HRC punch.
The bill for 0.08 mm
| Item | Basis | USD |
|---|---|---|
| 26 punches scrapped | USD 18.60 D2 bar + USD 11.20 of pre-hardening machining each | 774.80 |
| Extra grinding to chase the bow | 6.2 h across two cylindrical grinders | 434.00 |
| Replacement parts on a weekend setup | 26 pcs, one shift, rush | 312.00 |
| Second heat treat lot | 26 pcs, treater’s minimum charge | 185.00 |
| Total | 1,705.80 |
That is USD 7.11 per finished part on a part we sold for USD 41.20, and six days late on a delivery the customer had already scheduled press time around.
What we changed afterwards
The allowance is now calculated, not quoted from memory. For through-hardened D2 and A2 we allow 0.10 to 0.15% of the diameter for growth, plus 0.0015 mm per mm of length for bow on a symmetric part, and double that where there is a keyway, a flat, or any interrupted section. This family now runs with 0.45 mm on the OD instead of 0.30 mm. The extra stock costs about 40 seconds of turning per part.
We rough, stress relieve, semi-finish, then harden. These punches were machined to final size minus the allowance in one operation and sent straight to the furnace, which is the cheapest route on paper and the most expensive one in practice. The sequence now is rough within 0.5 mm, stress relief, semi-finish to the hardening allowance, harden, grind. One extra furnace trip at roughly USD 1.10 a part, and the last lot of the same geometry came back between 0.004 and 0.026 mm of bow.
Two coupons travel with every load. Extra blanks, machined to the same allowance, hardened in the same basket, measured before anything touches the good parts. The pair costs about USD 48 including heat treat. The scrap in this story cost USD 774.80. We have not had a bad lot since we started doing it, and when a treater changes their quench pressure or their rack layout, the coupon tells us on the day instead of three days later.
The numbers worth putting on the drawing
Specify a hardness range, not a minimum. “60 HRC minimum” invites a harder quench, and the extra point or two of hardness is exactly what increases the bow. 60–62 HRC gets you the wear resistance you wanted with less movement to grind out.
Anything tighter than 0.02 mm on a hardened feature is a grinding callout, not a machining callout. Machine it oversize, harden it, grind it. Trying to hold a hard tolerance before hardening means you are betting the distortion will be smaller than it is.
Ask for the treater’s lot report, not a pass/fail line. Hardness, quench medium, temper cycle, load position. When something moves you want to know whether the last lot was quenched at 2 bar or 5, and you cannot reconstruct that from a certificate that only lists HRC.
Where the bow matters, tell us where it does not. A punch that is 0.02 mm out of straight over its full length but concentric at both ends is usually fine. A punch that is perfectly straight and 0.03 mm off-centre is not. Datum choice on a hardened part is worth a sentence on the drawing.
The short version
Hardening is not a finishing operation. It is a process that moves metal, predictably, and the amount it moves is a function of geometry you control before the first cut. Every figure above comes from our own invoices and inspection sheets for this lot — read material and heat treat numbers as a budget estimate rather than a fixed rate, because both move with the market.
Our D2 tool steel page covers the grades we stock and the hardness ranges they realistically hold, and the heat treatment page shows how we sequence hardening with machining. If you are sizing a hardened part and not sure what your tolerance actually requires, send the drawing through our instant quote page or straight to the team — we confirm within one business day, and we will tell you where the allowance needs to go before anybody cuts metal.
The hardness certificate was correct. The drawing was wrong. That is the whole story.