Titanium machining
Grade selection, cutting data, springback and thin walls, and the drawing callouts that quietly triple a titanium price.
Above 45 HRC the question stops being how to cut the material and becomes whether to cut it at all — machine it soft and harden afterwards, or buy it hard and let CBN do the work. That sequence decision drives cost more than the grade does.
Ryan, founder
Every hardened part has at least two valid process routes, and they have completely different cost curves.
Machine soft, then harden. Annealed tool steel runs at 80–150 m/min with ordinary coated carbide. Cheap tooling, fast cycles, complex geometry is straightforward. The cost is that heat treatment moves the part — and the risk is that quench cracking or distortion beyond allowance destroys it after most of the machining money has already been spent.
| Grade | Typical dimensional change through hardening |
|---|---|
| A2 | ±0.05 % — low, which is its selling point |
| D2 | −0.06 % to +0.10 %, varies with section |
| O1 | +0.10 % to +0.20 %, and distortion-prone |
| H13 | ±0.05 % to 0.10 % |
| M2 | +0.05 % to +0.15 % |
On a 300 mm die plate, 0.1 % is 0.3 mm. That has to be left as grind stock on every critical feature, with a finishing operation afterwards.
Buy hard, machine hard. No distortion after the fact, one operation, dimensions are what you measured. Slower cycles, CBN tooling, higher hourly rate — and a hard ceiling around 65 HRC where turning and milling give way to grinding and EDM.
Hybrid, which is what most real parts do: rough soft, harden, finish hard or grind. An extra setup, and it removes most of the risk from both routes.
For rotational features above 45 HRC, CBN has taken a great deal of work away from grinders — but not all of it, and the threshold is worth knowing.
| Hard turning | Grinding | |
|---|---|---|
| Practical hardness | 45–65 HRC | to 70 HRC and beyond |
| Achievable roundness | 2–5 µm | under 1 µm |
| Typical finish | Ra 0.2–0.8 µm | Ra 0.05–0.4 µm |
| Material removal rate | 3–5× grinding | baseline |
| Setups on a multi-feature part | often one | often several |
The single-setup argument is usually the real reason to hard turn, and it is not about cycle time. A hardened shaft with two journals and a bore, ground conventionally, needs a cylindrical grinder, an internal grinder and careful re-datuming between them — and each re-chuck introduces concentricity error. Hard turned, every feature comes off one chucking and the runout between them comes free.
Where grinding still wins: tolerance below about IT5, finish under Ra 0.2 µm, hardness above 65 HRC, or geometry a lathe cannot reach.
Precipitation-hardening stainless is the family buyers most often under-specify. A drawing saying “17-4 PH” and nothing else omits the decision that controls strength, toughness, corrosion behaviour and machining cost.
| Condition | Hardness | Yield | Character |
|---|---|---|---|
| A (solution annealed) | ~38 HRC | ~760 MPa | Supply condition, not a service condition |
| H900 | 40–47 HRC | ~1170 MPa | Strongest, most notch-sensitive |
| H1025 | 35–40 HRC | ~1000 MPa | Balanced, common choice |
| H1150 | 28–33 HRC | ~795 MPa | Toughest, best stress-corrosion resistance |
H900 gets specified by default because it is the strongest, and it is also the most susceptible to stress corrosion cracking in chlorides. For anything seeing seawater, sour gas or cyclic loading with stress raisers, H1025 or H1150 is usually correct even though the datasheet number looks worse.
Some materials in this bracket are hard to machine for reasons that have nothing to do with hardness.
Invar 36 and Kovar are nickel-iron alloys bought for dimensional stability and glass-sealing respectively. They are soft, gummy, and the real work is the stress-relief sequence — buying Invar and then machining it carelessly wastes the reason you bought it.
Molybdenum and tungsten have a ductile-to-brittle transition at or above room temperature, so the edge propagates a crack rather than shearing a chip. Chips come off as powder, edges break out at cut exits, and interrupted cuts are genuinely dangerous.
Tantalum and niobium are the opposite — extraordinarily ductile, smearing rather than cutting, welding to the tool.
For all of them, material cost usually dominates machining cost, which inverts the normal optimisation: it is worth spending cycle time to reduce scrap risk.
Material: AISI A2 tool steel, supplied annealed. Harden and temper to 58–60 HRC after rough machining. Finish grind features marked G after heat treatment. Hardness verified on the finished part, three locations.
Grade, supply condition, target hardness, sequence, acceptance test. “D2, hardened” leaves all five open, and every open question widens the quote.
Also state whether white layer — the thin re-hardened, brittle surface a worn CBN tool can leave — is acceptable or must be checked. That turns a vague quality expectation into a measurable one.
When CBN hard turning beats grinding on hardened steel, when it does not, and the tolerance and finish thresholds where the decision actually flips.
Condition A, H900, H1025, H1150 — what each does to strength, toughness and machinability of 17-4 PH, and which one your drawing should actually call out.
What each cutting tool material is good for in titanium, nickel alloys and hardened steel, the speed ranges involved, and why the cheapest insert is rarely cheapest.
How H13, D2, A2, O1 and M2 differ, what each does during heat treatment, and how to decide whether to machine before or after hardening.
Residual stress in plate and bar, distortion during machining and heat treatment, and the process sequence that keeps large exotic alloy parts in tolerance.
Low-expansion alloys for tooling, optics and glass-to-metal seals. Why Invar 36 and Kovar are hard to machine, and how to hold tolerance on parts that must not move.
Machining refractory metals — why molybdenum and tungsten fracture instead of deforming, how tantalum gums up tools, and what a realistic process looks like.
Everything above is the thinking. If you have a drawing, the fastest way to find out what it will cost and how long the material takes is to send it: a priced quote from the engineer who read it, within 24 business hours, including the parts of the drawing that are costing you money without buying anything.
Grade selection, cutting data, springback and thin walls, and the drawing callouts that quietly triple a titanium price.
Inconel, Hastelloy, Monel and duplex — the alloys that harden in front of the tool, and how to pick the cheapest one that still works.
Everything that happens before the first chip: packaging an RFQ, reading a quote spread, documentation levels and material lead times.
Milling against turning against grinding against EDM — what actually decides the process, and the drawing habit that locks in the wrong one.
Written for the shop side. What a buyer checks in ninety seconds, why good shops lose quotes, and the documentation that decides who gets the repeat order.