3.1 / 3.2 material certificates
Per EN 10204, with every delivered batch.
Quality and traceability
Documented quality with every batch: 3.1/3.2 material certificates per EN 10204, dimensional reports, first article inspection, SPC and CpK, batch and heat-number traceability, GD&T per ISO 1101 and ASME Y14.5.
What travels with the parts
Batch traceability
Every part can be traced back to the heat number of its material.
First article inspection
FAI per AS9102, before the series is released.
Dimensional reports
3D measurement across the flow, with a report for every batch.
SPC and CpK
Statistical process control on critical dimensions.
GD&T
Interpretation per ISO 1101 and ASME Y14.5.
PPAP and 8D
Production part approval documentation and structured non-conformity handling.
A buyer does not ask for all of this on every part. They ask when the part is safety-relevant — a rail component, one in pressurised hydraulics, one that something else depends on. That is why the list above is complete rather than selective.
Material certificates: what each type states
EN 10204 does not describe the quality of the steel; it defines who guarantees what, and on the basis of which tests. Four types in two families: two declarations without test results and two with them. The difference decides whether a delivered batch can be traced back to the melt it came from.
| Type | What it confirms | Who issues it |
|---|---|---|
| 2.1 | Declaration of compliance with the order. No test results, no measured values. | The manufacturer, with no independent check. |
| 2.2 | Declaration of compliance with results from non-specific testing — samples from current production, not from the batch delivered. | The manufacturer. |
| 3.1 | Results from specific testing, carried out on the material delivered: chemical composition from the melt, mechanical properties from samples. | The manufacturer’s inspection department, independent of production. |
| 3.2 | The same specific testing, but agreed and confirmed by two parties. | The manufacturer’s inspection together with the purchaser’s representative or an authorised inspector. |
Types 2.1 and 2.2 are declarations without testing on the delivered batch; 3.1 and 3.2 carry specific results. The type is settled with the order: the purchaser asks for it, and the standard does not choose on their behalf.
The certificate follows the melt, not the part
A 3.1 certificate attests to the material the part was made from, identified by its melt number. It says nothing about the dimensions of the part — those are evidenced by the measurement report. The two documents answer different questions: what it is made of, and how it came out.
What makes traceability work
A material certificate is worth only as much as the unbroken link between the delivered batch and the melt number: the bar marked at goods-in, the marking preserved through sawing, the number carried onto the batch record. A certificate without that link is a document about some steel or other.
When 3.1 is enough and when it is not
For most machined parts, 3.1 is the level required: results on the material delivered, signed by a department independent of production. 3.2 is asked for where the acceptance conditions call for a second party to confirm, and that is settled before the order, because it changes how the material is bought.
What to state on the drawing or the order
The certificate type, the material with its standard and, where needed, the marking requirement on the part. An order that says only “with certificate” leaves the type to the supplier, and at incoming inspection it may turn out not to be the one expected.
General tolerances: what the drawing omits still applies
On any drawing, most dimensions carry no tolerance of their own. That does not make them free: they follow the general tolerance note, usually ISO 2768. When the note is missing, each side assumes something different — and assumptions surface at incoming inspection.
| Nominal range mm | Class f fine [mm] | Class m medium [mm] |
|---|---|---|
| 0.5 – 3 | ±0.05 | ±0.1 |
| over 3 – 6 | ±0.05 | ±0.1 |
| over 6 – 30 | ±0.1 | ±0.2 |
| over 30 – 120 | ±0.15 | ±0.3 |
| over 120 – 400 | ±0.2 | ±0.5 |
| over 400 – 1000 | ±0.3 | ±0.8 |
Permissible deviations for linear dimensions, from ISO 2768-1. The standard also defines classes c (coarse) and v (very coarse), while part two, ISO 2768-2, covers general form and position tolerances in classes H, K and L.
Why the class matters
On a 300 mm dimension, class f gives a window of 0.4 mm and class m one of 1 mm — two and a half times wider. The difference is not academic: it decides whether a surface is finished in the first cut or needs an extra finishing pass, on every piece of the batch.
Finer is not better
An “ISO 2768-f” note applied everywhere out of caution also tightens the dimensions that carry no function — distances to a free edge, relief lengths, chamfer depths. Each of them then has to be held and measured. The class follows what the part does, and functional dimensions get their own tolerance, written on the dimension.
A general tolerance is no substitute for a fit
A fit — H7 for a bearing bore, h6 for a shaft end — never comes out of a general tolerance: at 50 mm, H7 means a window of 25 µm against the 600 of class m — twenty-four times tighter. Fits, surface finish and measuring datums belong on the dimension, not in the note in the corner.
When the note is missing altogether
A drawing without a general tolerance note is not a drawing without tolerances; it is one whose tolerances are undecided. That gets settled before the quotation, not after the first delivery — otherwise the price is calculated on one assumption and the part is measured against another.
The tolerance chain
A tight tolerance on one dimension does not carry over to the assembly unchanged. Dimensions add up, and the way they add changes the result several times over. This is the question a single part drawing does not answer.
- Worst case
- Statistical (RSS)
- Times tighter
A linear calculation on a chain of parallel dimensions. A real chain may contain dimensions of differing sign and direction; those are summed algebraically, following the assembly diagram.
Direct summation
Every dimension is assumed to sit at the least favourable end of its tolerance, all in the same direction. The result is the sum of the tolerances and grows linearly with their number: ten dimensions at ±0.01 mm give ±0.1 mm. It is the only result that can never be exceeded.
Root sum of squares
If each dimension is centred, independent of the others and normally distributed, the deviations partly cancel. The total then grows with the square root of the number of dimensions rather than with their number: the same ten give ±0.032 mm instead of ±0.1 mm.
What the statistical method does not say
RSS is an expectation, not a guarantee. It assumes centred, independent processes — an assumption that fails if one feature drifts systematically, if two dimensions come from the same setup, or if the batch is too small for statistics to mean anything. For a critical part or a small series, the figure that holds is the one from direct summation.
What a capability index says — and when it says nothing
If you ask us for Cpk during qualification, it is worth agreeing beforehand what the number means. Below is the definition, not a promise: the value a given process reaches is measured on your parts, not announced on a website.
Cp and Cpk do not measure the same thing
Cp compares the width of the tolerance with the spread of the process — the tolerance band divided by six standard deviations. It says how narrow the distribution is, but not where it sits. Cpk adds the position: how close the mean is to the nearer limit. A process can have a good Cp and a poor Cpk — a tight spread, badly centred. That is why Cpk is the number asked for, and why Cp alone says too little.
What the value means
For a stable, centred, normally distributed process: Cpk 1.00 means the spread exactly fills the tolerance — roughly 2,700 parts per million fall outside it. Cpk 1.33 leaves a margin of one standard deviation on each side, about 66 per million; this is the level routinely required in series supply. Cpk 1.67 drops below one part per million and is asked for where an escape is expensive. The step from 1.33 to 1.67 is not “slightly better”: it is around a hundred and twenty times fewer parts outside tolerance.
When the number says nothing
The index assumes a stable process. If the tool wears and the dimension drifts, the calculation yields a handsome figure over a process that will be out of tolerance tomorrow — which is why the trend is watched, not only the index. It assumes enough parts: a Cpk computed on five pieces is arithmetic, not evidence. And it assumes a measuring system to match — if the uncertainty of the gauge consumes a meaningful share of the tolerance, the figure measures the gauge, not the process.
What you receive from us
The characteristics under statistical control are agreed on the drawing at quotation, not after delivery — usually the functional dimensions, not all of them. For each one, the number of parts, the gauge and the sampling interval are settled. The dimensional report per batch accompanies every delivery anyway; the capability index is added where it was asked for and where the run is large enough to mean anything.
Certifications
We do not list certifications we cannot produce on request. What accompanies every delivered batch is set out above, document by document. For the current certification status, write to us.
Request a quote
Send us your drawing. We respond with a quotation within 48 hours, and your data stays confidential.
Send us your drawing