Aerospace High Precision Machining: Materials, Documentation, and Tolerance Expectations Anchored to Real Material States

Ask ten machine shops what "aerospace precision" means and most will answer with a single number: the tighter the tolerance, the more aerospace it is. That answer is wrong in a way that costs buyers real money. On an aerospace machined part, the tolerance that matters is not the tightest band on the drawing — it is the band that the material, in the condition you ordered it, can actually hold across the whole run, and that your evidence chain can prove when the parts arrive. A ±0.005 mm callout on a mating face is only a good callout if the stock was the right temper, the heat-treat step landed in the right place in the sequence, and the material certificate says the same thing the drawing does. Pull any one of those three legs out — the tolerance band, the material state, or the evidence — and the pair that is left cannot stand. This article is about the pairing: which material states change what a band can hold, how to read the tolerance numbers an aerospace drawing actually uses, why the temperature of the measuring room is part of the specification, and which document answers which question when you are deciding whether to accept the parts.

Why does an aerospace precision order fail when the tolerance band, material state, and evidence fall out of sync?

Aerospace hardware is specified in a way that assumes a perfect, stable, featureless material, and then machined from a metal that is anything but. An aluminum bracket, a titanium fitting, or a precipitation-hardened steel part arrives at the machine with a history: it was melted, worked, and heat-treated into one specific condition, and that condition carries its own internal stress, hardness, and dimensional behavior. The tolerance band on the drawing is written for the part in service. The band is actually produced, and later verified, on a part whose material state is doing one thing while the cutter and the gauge are doing another. When those three — the band, the state, and the measurement evidence — do not line up, the order fails in a characteristic way: the parts measure out of tolerance at your dock, the shop's report says they were in tolerance at the factory, and nobody is technically lying. Each side measured a real thing; the material was simply not the same thing in both places, or the file that would have proven which state was measured never travelled with the box.

The failure mode matters more on aerospace work than on almost any other class of machining, because the consequences of a wrong part are severe, and the tolerances sit close to the physical limits of what a metal holds at a given temperature and condition. Notice the shape of the problem: it is rarely that the shop cannot cut tight enough. Precision machining holds ±0.005 mm on milling and turning and ±0.002 mm on grinding as routine, documented bands — ±5 µm and ±2 µm, widths well under the roughly 70–100 µm thickness of a human hair. Cutting to that level is a solved problem. What is not solved by the machine is the rest of the chain: whether the material will stay put through every setup and every thermal cycle, and whether the record trail lets a receiving inspector reproduce the factory's verdict. Aerospace buyers who have been burned are not usually burned by a shop that cut too loose. They are burned by a material state they did not specify, a heat-treatment step that moved a dimension they assumed was final, or a certificate that answered a question they never asked.

The discipline that prevents all three is pairing. Before you send an aerospace drawing for quote, you should be able to state, for the critical features: the tolerance band that feature needs, the material condition that keeps that band honest, and the document that proves both. This page walks through the pairing one leg at a time, in the order a buyer actually meets them — starting with the material, because the state of the metal is the least visible and the most destructive when it is wrong.

Which material states change what a ±0.003 mm band can hold, and what should your drawing name?

On a precision drawing, "7075" or "Ti-6Al-4V" or "17-4PH" is only half the material specification. The other half is the state: the temper, the heat-treatment condition, and in many aerospace programs the lot or heat identity behind it. The state decides how the metal cuts, whether it distorts when thin sections are machined out of it, and how it measures after heat treatment. Two parts machined from the same alloy in different states can land on opposite sides of the same tolerance band with identical cutting programs, because the metal underneath the cutter was not the same metal.

Alloy	States a precision buyer actually sees	What changes between the states	What the drawing and file should carry
7075 aluminum	T6 (peak aged) vs T7351 (overaged, stress-relieved)	T6 is the highest-strength temper but carries higher internal stress from quenching and, in thicker sections under sustained load, is more sensitive to stress corrosion along the grain direction; T7351 gives up some strength for markedly better stress-corrosion resistance and better dimensional stability in large machined panels because it is stress-relieved by stretching after solution treatment	The temper written on the drawing — T6 and T7351 are not interchangeable despite identical chemistry; XRF verifies the alloy but cannot read the temper, so the certificate and the incoming-state record are the proof
Ti-6Al-4V (Grade 5)	Annealed (mill condition) vs ELI (extra-low interstitial)	Annealed is the standard supply condition and the practical default for machined fittings and housings; the ELI grade holds lower oxygen and iron to improve fracture toughness for fracture-critical applications, and it carries its own certificate and lot traceability, not just a different price	Whether the drawing means standard Grade 5 or the ELI variant; because the difference lives in interstitial chemistry, only the mill certificate can prove it — no shop-floor test can
Inconel 718	Solution annealed vs solution annealed and aged	The aged (precipitation-hardened) condition is what carries the alloy's high-temperature strength; solution-annealed stock is softer and cuts faster, but a part that must ship aged has a heat-treatment step in its future, and 718 work-hardens quickly and cuts slowly no matter which state you start from	The required final condition, so the shop plans the machining sequence around the aging step instead of discovering it after the parts are done
17-4PH stainless	Condition A (solution annealed) machined first, then aged to H900 or H1025	H900 is the higher-hardness, higher-strength aging treatment; H1025 trades a little strength for improved toughness, and the aging step itself moves dimensions slightly, so features machined to final size in Condition A can shift a few micrometers in the furnace	Which aging condition the part must finish in, and which features are final after aging versus before it — the state record, not the alloy name, answers this

Two buying habits follow from the table. First, name the state on the drawing the way you name the alloy — "7075-T7351," not "7075." A blank material line invites the shop to buy whatever temper is cheapest that week, and the tolerance story changes with it. Second, understand that some state differences are invisible to any instrument a machine shop owns. XRF confirms the alloy composition; it cannot see a temper, an interstitial level, or a precipitation-hardening cycle. Those live in the material certificate and the heat-treatment record, which is exactly why the quality and traceability documentation side of the order carries weight it does not carry on a loose-tolerance part. The metal's identity and the metal's state are two different facts, and only one of them can be measured on the shop floor.

There is a second, quieter way material state reaches into the tolerance band: residual stress. A bar of 7075-T6 that was quenched hard carries locked-in stress, and when you machine away half the cross-section, the part relaxes and moves — sometimes more than the ±0.005 mm band you are holding on the remaining wall. That is not a cutting error and no amount of spindle accuracy fixes it. It is a material-state problem that shows up at the machine as a tolerance problem, and the usual remedies live upstream: choose a stress-relieved temper where the geometry removes a lot of stock, rough and settle before finishing critical features, and tell the shop which features are critical so it can sequence the work instead of discovering distortion on the last operation. Buyers who treat "material" as a chemistry question and "tolerance" as a machining question miss this coupling entirely; the pairing view exists precisely because they are one decision.

When is ±0.002 mm the right call on an aerospace drawing, and when is ±0.003 mm the honest one?

Precision machining, as this site defines it, is not one number stamped across a part. It is a collection of features, each carrying the band its function needs and nothing tighter. The facility holds ±0.005 mm on CNC milling and turning and ±0.002 mm on precision grinding, and the gap between the two bands is a process fact, not a marketing claim: a grinder takes metal off in micron-sized bites, which is what lets a seat, a journal, or a seal face reach ±0.002 mm where a milling cutter cannot follow. On an aerospace drawing, the tight band almost never belongs on the whole part. It belongs on the two or three features where fit is load-bearing and where the material, in its final state, will hold what the gauge verifies.

The core page of this site — high precision machining, defined feature by feature, quote by quote, report by report — makes the same point without the aerospace framing: precision is a per-feature decision, and the phrase only earns its meaning when the drawing says which features are tight and which are not. Read a typical aerospace print with that lens and a pattern emerges. The parts that fly — brackets, housings, fittings, actuator components, structural mountings — concentrate their ±0.002 mm callouts on ground, fit-critical spots: the bore that receives a bearing, the journal that spins inside a bushing, the flat face a seal closes against. The rest of the part, including the critical mating and interface faces that make the assembly work, is very often specified at ±0.005 mm, and that is not a shortcut — it is the honest answer to what the feature needs. A bolt hole that locates a bracket does not need ±0.002 mm; it needs to be where the drawing says it is within a band that assembly can absorb. Over-tolerancing an aerospace part is not a harmless extra. Every unnecessarily tight band forces slower feeds, more measurement, and tighter process control around a feature that gains nothing from it, and the cost of that discipline lands on every part in the run.

The sharper question is not "how tight can the shop go?" but "which band does each feature need, and does the material state allow that band to be held repeatably?" Three rules keep the answer honest. First, put ±0.002 mm only on features that are actually ground and that function demands — a bearing seat, a seal surface, a journal — and make sure the drawing says the feature is ground, because the process and the band are tied. Second, hold ±0.005 mm as the working band for machined geometry that locates, mates, or interfaces, and resist the urge to tighten features whose function does not ask for it. Third, where heat treatment sits between roughing and finishing — an aged 17-4PH part, a stress-relieved aluminum panel — give the final band to the features that are machined after the state-changing step, and say so in the notes. A dimension cut into Condition A material and then aged is a dimension that moved; a dimension finished after aging is a dimension that stayed. The site's guide to realistic tolerance and finish callouts carries the same feature-by-feature thinking into the question of what your own receiving gauges can confirm.

Statistical tolerance is the quiet third leg of how an aerospace drawing expresses expectations. A print may carry a general tolerance block (the title-block default that applies to every undimensioned feature), feature-specific geometric callouts under the GD&T rules (position, profile, flatness tied to a datum system), and in some cases a statistical tolerance note that asks for capability — typically expressed as Cpk — on key characteristics rather than a hard per-part limit. The facility targets Cpk ≥ 1.67 on key characteristics, which means the process spread sits well inside the band with room to spare. When you see a statistical note on an aerospace drawing, read it for what it is: the drawing is telling you that the feature will be judged on process capability across the run, not on one heroic part, and that is the measurement philosophy this site's whole tolerance-band discussion is built around.

Why is the temperature of the measuring room part of an aerospace tolerance specification?

Aerospace metals expand and contract with temperature at rates large enough to dwarf the tolerance bands on the drawing. The numbers are easy to underestimate, because the drawing never mentions temperature and the shop floor is rarely at the temperature the specification assumes. Standard dimensional measurement practice uses 20 °C as the reference: that is the temperature at which the nominal dimensions on the drawing are defined, and it is the reason serious metrology rooms are climate-controlled. Representative values for the alloys a precision aerospace shop runs: 6061 and 7075 aluminum sit near 23 µm per meter per degree Celsius, stainless steel near 16, titanium near 9, and precipitation-hardened 17-4PH near 11. Those look small until you multiply them by real part lengths and real temperature differences.

Run the arithmetic on an aluminum part, because aluminum is where the effect is largest. Picture a 100 mm length — a bolt spacing, a bore centerline distance, a machined envelope — measured at a 10 °C temperature gap between the factory's inspection room and your receiving area. A 7075 part grows or shrinks by roughly 0.023 mm over that swing; 6061 is close behind at about 0.024 mm. A ±0.005 mm band is 0.010 mm wide in total. Ten degrees of temperature has moved the part more than twice the width of the entire band — before a single cutting error is involved. This is why the same part can read "in tolerance" on the supplier's CMM report and "out of tolerance" on yours, with both measurements honest. The part was not the same size in both rooms. A titanium or steel part suffers the same effect at a smaller scale, which is why long thin titanium parts and precision aluminum panels are the classic cases: the metal moves, the gauge does not, and the argument starts.

The fix is not exotic. Agree on the measurement condition before the order, not after the dispute. The shop should measure critical features in a temperature-controlled environment near the 20 °C reference, or at minimum record the room temperature next to the CMM results so the numbers can be corrected or repeated. Parts that arrive at your dock in a warm truck should soak to your measuring-room temperature before the gauge touches them — measuring a warm part against a 20 °C drawing is measuring a different object than the one you will install. And when you compare the supplier's report to your own incoming inspection, compare numbers taken at a stated common temperature; the difference between the two readings should be explainable as gauge and temperature uncertainty, not as a manufacturing fault. On an aerospace part, the temperature of the room is not a footnote to the tolerance — it is part of the tolerance, because the band is defined at 20 °C and nowhere else.

Which document answers which question, and what happens when one is missing?

The evidence layer of an aerospace precision order is not a stack of paper to be filed and forgotten. Each document answers one specific question, and the questions line up exactly with the pairing this article has been building: what the material is, what state it was in, whether the geometry matches the drawing, and whether the process was followed. A file that answers a question you never asked is decoration; a missing file that would have answered a question you need is a risk you accepted without knowing it. The table below names the documents a serious aerospace order produces and the question each one settles.

Document	The question it answers	What you lose if it is missing
Material certificate (MTC / mill cert)	What is the alloy, by what specification, from which melt or heat, and in what condition did it leave the mill?	The only non-destructive proof of alloy identity and lot origin; without it, a small machined part cannot be proven to be 7075-T7351 versus 7075-T6, or Grade 5 versus ELI, because shop-floor tests cannot see the difference
Incoming inspection / state record	Was the stock verified against the certificate at receipt, and does the recorded incoming condition match the drawing?	The link between the paper and the metal; without it, a certificate can describe the right material while the wrong lot actually went to the machine
First-article inspection report (FAI)	Does the geometry of the first part match every critical callout on the current drawing revision?	The proof that the process produces the drawing; without FAI, "in tolerance" is a claim about parts that were never verified feature by feature
Heat-treatment / process records	For aged or stress-relieved material, was the state-changing step actually done — to which parts, at what cycle, with what result?	For an aged 17-4PH or 718 part, the difference between a part that has the specified state and a part that merely looks like it does
Dimensional (CMM) report and SPC data	What did the measurement system actually find on the critical dimensions, and is the process stable across the run, not just on the first piece?	The repeatable, traceable numbers that let your receiving check reproduce the factory's verdict; without them every acceptance is an act of faith

Notice how the table reads like a conversation, not a checklist. The material certificate and state record answer "what is this metal, in what condition." The FAI answers "does the geometry match the drawing." The heat-treatment record answers "was the state-changing step real." The CMM report and SPC data answer "what did the measurement find, and can it be repeated." When a buyer says "I need full documentation," they usually mean "send me everything", but a precise buyer asks each question separately and checks that the file answering it exists, is current to the drawing revision, and is tied to the actual lot number on the box. Documentation is the twin record of the geometry: the drawing says what the part must be, and the files say what the part was, in the state it was made, as measured under known conditions.

The honest boundary also belongs here. A precision machine shop reports material identity, machining process, and measurement truthfully, and it documents what it does. It does not certify airworthiness, approve a design, or take responsibility for the flight envelope — those live in the buyer's engineering and quality system. Aerospace programs sometimes flow down expectations such as an AS9100-style quality management system or a particular first-article format; whether a given program requires that of its suppliers is a qualification question the buyer's supplier team answers against the current scope of the candidate's certificates. What a build-to-print shop can sign is the layer beneath that: its own three management-system certificates — ISO 9001:2015 (116024/A/0001/UK/En), IATF 16949:2016 (product design excluded) (131941/A/0001/SM/En), and ISO 14001:2015 (116024/B/0001/UK/En) — with Cpk ≥ 1.67 targeted on key characteristics, plus the documents that describe the work it actually did: material state, process, and measurement. That layer is the evidence your quality system builds on, and it is why the pairing question — band, state, evidence — has an answer you can audit.

Which RFQ terms lock the pairing in before the parts are cut, and what do you check when they arrive?

Everything above converts into a short list of habits that cost nothing at quoting time and save an aerospace order at arrival time. The point is to write the material state, the measurement condition, and the evidence list into the RFQ before the parts exist, so the shop prices the honest version of the job and your receiving team knows exactly what to verify.

• Write the full material line on the drawing and the RFQ. Name the alloy, the state, and the governing specification: "7075-T7351 per AMS 4078" style, not "7075." If the part is aged or stress-relieved, say which condition it must finish in and which features are critical after that step.
• Ask for the material certificate and the incoming state record as deliverables. State that the certificate must carry the heat or lot number, and that stock must be verified against it at receipt — the XRF check is the shop's to run, but your RFQ makes it a requirement rather than a courtesy.
• State the measurement basis for critical features. Write the 20 °C reference into the order and require critical dimensions to be inspected in a climate-controlled room, with the room temperature recorded beside the CMM data. This single line prevents the "their report vs my bench" dispute before it starts.
• Require the first-article report against the current drawing revision, and ask which features carry Cpk ≥ 1.67. If the drawing uses statistical tolerance on key characteristics, your acceptance criterion should say the same thing the drawing says — capability across the run, not one part.
• List the documents beside the shipment terms. Material certificate, state record, FAI, heat-treatment records if the state demands them, and the dimensional report travel with the box. A document that arrives a week later is a document your inspector cannot use.
• Give the shop the feature-by-feature tolerance picture. Flag the two or three features that genuinely need ±0.002 mm and confirm they are ground; leave the rest at ±0.005 mm or the general block. A shop that can see which features are critical will sequence the work correctly, which is the difference between a dimension that survives heat treatment and one that does not.

At the receiving dock, run the checks in the order the pairing demands. First, verify the material story: does the certificate's alloy and state match the drawing's material line, and does the lot number on the box match the certificate? Second, let the parts soak to your measuring-room temperature, and measure critical features at a stated, common temperature — compare your numbers to the supplier's report taken at the same basis. Third, confirm the geometry against the first-article report and the drawing revision, not against a vague memory of what you ordered. Fourth, where the state mattered — an aged part, a stress-relieved panel — confirm the heat-treatment record exists and covers the lot you received. Run those four steps in order and the aerospace order reduces to what it should be: a tolerance band, held by a known material state, proven by a file trail you can audit. That is the pairing, and it is the whole game.

Frequently Asked Questions

Does aerospace precision machining require tighter tolerances than medical or automotive work?

Not automatically. Each industry concentrates its tolerance where the function demands it. Aerospace parts commonly carry ±0.005 mm on mating and locating geometry with ±0.002 mm reserved for a few ground features such as bearing seats, journals, and seal faces. What distinguishes aerospace is not the tightest number on the drawing — it is the demand that the material state, the heat-treatment history, and the documentation be tied to the tolerance, because the consequence of a wrong part is severe. The discipline is in the pairing, not in uniformly tighter numbers.

Why can't the shop just check the temper with a spectrometer or XRF gun?

Because those instruments read chemical composition, and a temper is a metallurgical condition, not a chemistry. 7075-T6 and 7075-T7351 have essentially the same composition; what differs is the heat-treatment and stress-relief history, which no surface reading can see. The same is true of Grade 5 versus ELI titanium (the difference is interstitial content) and of an aged versus unaged 17-4PH part. That is why the material certificate and the state record are not paperwork for their own sake — they are the only practical proof of the state the tolerance depends on.

My supplier's CMM report says the part is in tolerance, but our bench says it is out. Who is wrong?

Possibly neither. If the part is aluminum and the two measurements happened at different temperatures, a 10 °C difference moves a 100 mm feature by roughly 0.023–0.024 mm — more than twice the width of an entire ±0.005 mm band. Before arguing about the manufacturing, compare the measurement conditions: the room temperature at each inspection, the soak time of the part, and the datum and sampling method used. State a common measurement temperature and the two readings usually converge. If they still disagree by more than the gauge uncertainty, then you have a real discrepancy worth investigating.

When should I specify ±0.002 mm instead of ±0.003 mm on an aerospace part?

Only on the features where a few micrometers change the assembly, and then only if the feature is ground. On an aerospace part, that tighter band is reserved for the two or three locations where a rotating element lands or a fluid path closes: a journal finished on the cylindrical grinder, a bore a bearing race seats into, a flange face a seal must meet. Everything else — bolt patterns, locating bosses, general machined geometry, even mating faces that carry no seal — performs without it, so specifying ±0.002 mm there spends money on a band the function never asked for. Decide per feature, confirm the tight ones are ground, and let the general tolerance block carry the rest.

Why does heat treatment come up in a conversation about tolerances?

Because heat treatment changes the material state, and the state changes the dimensions. A 17-4PH part machined to final size in Condition A and then aged can shift a few micrometers in the furnace; an aluminum part machined from a hard-quenched temper can relax and distort as stock is removed. If the drawing calls for an aged or stress-relieved condition, the features that must hold the tight band have to be machined after the state-changing step, and the RFQ has to say so, so the shop sequences the work correctly instead of discovering the movement on the last operation.

What documents should an aerospace RFQ list as deliverables?

The material certificate with the heat or lot number, the incoming inspection and state record, the first-article report against the current drawing revision, heat-treatment or process records where the state demands them, and the dimensional (CMM) report with room temperature noted. Each answers one question: what the material is, what state it was in, whether the geometry matches the drawing, whether the state-changing process was real, and what the measurement actually found. List them in the RFQ, require them to travel with the shipment, and verify them at receiving in that order.

My drawing uses general tolerances. Do I need GD&T to buy aerospace precision parts?

You need the drawing to express which features matter and how they will be judged. Many precision parts carry a general tolerance block for the majority of dimensions plus a small number of feature-specific geometric callouts — position, profile, or flatness tied to datums — for the features that function matters for. GD&T is not an end in itself; it is the language that lets you say precisely which features carry which expectation and how they will be measured. If your drawing only carries a general block, the shop can still hold it, but you have not yet told it where the tight band belongs.