Families and data sources

Three labels on every grade#

Every card in the material catalog carries three organising labels, read verbatim from the card header rather than inferred from its chemistry. They are easy to confuse, and they drive different things.

Material type is the coarsest: steel, stainless steel or cast iron. It is what the alloy-type facet in the catalog browser filters on, and families exist only inside one type — which is why the stainless and cast-iron grades vanish from the list while the facet says Steel.

Three of the 106 stainless cards carry a differently-cased type string in the catalog — stainless steel against Stainless Steel; the browser folds the two spellings together, so you see one stainless bucket rather than two.

Family is the application class — Structural steel, Bearing steel, Hot-work tool steel. There are 43 distinct values. This is the label a metallurgist actually navigates by, and it is what the Family facet filters on.

Group is the alloying system — Ni-Cr-Mo, Cr-Mo-V, C-Mn, Mn-B. There are 68 distinct values. This is the field list_materials returns as group, and the field the agent groups a chat answer by. It is also the label the browser's section headings switch to once you have picked a single family.

One consequence worth stating plainly: the family taxonomy on this page is a display taxonomy for browsing 2,060 grades. It is not the same vocabulary the closest-grade matcher uses when you type in a chemistry — that one is a coarser twelve-key set, described in the material catalog.

The 43 families#

The counts below account for all 2,060 grades.

What the large families are for#

• Structural steel (680). Weldable construction and general plate steels — the S235/S355/S460 territory, including normalised and thermomechanically rolled variants. The largest single family in the catalog.
• General engineering steel (411). The medium-carbon and low-alloy machinery steels that get quenched and tempered, through-hardened or induction-hardened: C45, 34CrNiMo6, 18CrNiMo7-6 and the AISI equivalents 4140, 4340 and 8620. Two designations you would expect here are not: 42CrMo4 is filed under quenched and tempered, and 30CrNiMo8 under structural steel.
• General tool steel (210). Tool steels that are not separately classified as hot-work, cold-work, plastic-mould or high-speed — water-hardening and oil-hardening grades, and the broad middle of the tool-steel range.
• Hot-work tool steel (105). Cr-Mo-V die steels that keep hardness and toughness at temperature, for die casting, hot forging and extrusion tooling. H13 and 1.2344 sit here.
• Cold-work tool steel (96). High-carbon, high-chromium steels for blanking, forming and cutting at ambient temperature — the D2 / 1.2379 family and its relatives.
• Welding consumable steel (71). Filler-metal and weld-deposit chemistries. These matter when you want the transformation behaviour of the deposit itself rather than the parent plate.
• Martensitic stainless (60). 12–18% Cr grades that harden by quenching — cutlery, valve, pump-shaft and turbine-blade steels. Much the largest of the six families with “stainless” in the name.
• Cast irons (41). Grey, ductile and compacted-graphite irons. Most are filed under composition-style names rather than trade designations, so search these by chemistry rather than by grade name.
• HSLA (39). Microalloyed high-strength low-alloy sheet and plate, strengthened by Nb, Ti or V precipitation and grain refinement rather than by martensite.
• Bearing steel (33). Through-hardening high-carbon chromium steels, 100Cr6 / 52100 and their variants.
• High-speed steel (30). W-Mo-V tool steels that retain hardness at cutting temperatures.
• Plastic-mould steel (29). Pre-hardened and through-hardened mould steels chosen for polishability and machinability — P20 and the 2311/2738 mould-steel range, including M200, M238 and Formadur 2738. The bare 1.2311 card is filed under general tool steel. Its stainless counterpart, corrosion-resistant mould steel, is a separate family of 16.

The rest#

The mid-sized families cover elevated-temperature service (heat-resistant 25, creep-resistant 22 — the Cr-Mo and Cr-Mo-V power-plant steels), surface-engineering feedstock (nitriding 22 — 31CrMoV9 and the 1.85xx range; case-hardening 19 — 16MnCrS5, 17Cr3, 20MoCrS4), transformation-designed grades (bainitic 18), hardenability-treated grades (boron steel 17), delivery condition (quenched and tempered 17) and spring steel 16.

The advanced high-strength sheet families a body-in-white engineer works with are present but small: TRIP 10, dual-phase 10, complex-phase 3, press-hardening 4 and a single martensitic steel (MS) grade. That is worth knowing before you go looking — the AHSS slice of this catalog is thin compared with its tool-steel and engineering-steel slices, and it is broadly where the twenty reference grades sit.

The long tail runs down through high-strength structural 9, railway 8, near-pure iron 6, mild/drawing 5, pressure-vessel 4, powder-metallurgy 4, electrical 3, wear-resistant 2, weathering 2, interstitial-free 2, austenitic stainless 2, and six families with a single grade each: duplex stainless, super duplex stainless, precipitation-hardening stainless, pipeline and maraging, alongside that lone martensitic sheet grade.

Family labels are imperfect, as any single-label taxonomy over 2,060 grades will be, and the misfilings are not rare edge cases: 16MnCr5, the archetypal case-hardening grade, sits under bearing steel, and 20MnCr5 sits under general tool steel. When a grade's family is not where you would file it yourself, search by name or by alloy system instead — finding your steel covers how.

The 68 alloy-system groups#

Once you are inside a 680-grade family, the family label has stopped helping. The group label — the alloying system — is the more useful axis, because it is the one that correlates with hardenability and transformation behaviour.

Those twenty groups hold 1,795 of the 2,060 grades. The remaining 48 groups share the other 265 between them.

A grade's group is what the agent uses to organise a chat answer, so a question phrased in alloy-system terms tends to get a tidier reply than one phrased in family terms:

Which Ni-Cr-Mo grades are in the catalog?

Where the cards come from#

The catalog is assembled from the published continuous-cooling and isothermal transformation literature. Card headers carry a $ Reference: line naming the source and a $ Confidence level: line, both parsed and stored alongside the card.

These are the fourteen largest reference strings in the catalog:

Together those account for 1,160 cards. The rest cite a long tail of smaller sources, and 21 cards carry no reference line at all.

The spellings above are the card's own, not the publishers'. The stored strings are plain ASCII, so a header reads Bohler, Dorrenberg, Fabrication Francaise and Warmebehandlung der Stahle without the umlauts and cedilla the printed titles carry. Search for them that way. The stored text is not otherwise normalised either — the same atlas appears under several slightly different strings, and a few carry typos — so these counts are per exact string, not per publication.

Two things follow from that mix. First, the catalog is not one dataset — it is a union of published research atlases (IRSID, NIMS, Rose/Wever/Peter), one classic ASM compilation, one Climax Molybdenum monograph, several steelmakers' own published diagrams for their own products, and sources the card headers name only by acronym, including the single largest contributor, HTDG. Measurement conditions, austenitising practice and reporting conventions differ between them, and nothing in the product reconciles those differences for you.

Second, the supplier-published sources are grade-specific by construction. A Bohler or Dorrenberg card describes that maker's product. If your steel is a different mill's version of nominally the same designation, the card is a starting point, not your steel.

Reading a card's own provenance#

The reference and confidence fields are not returned by list_materials and are not shown in the catalog browser, so you cannot look up a grade's source from the material page today.

What does work: every simulation run stores the exact material card it used as a mat_Steel.k artifact (RSW stores one per plate as mat_Steel_<plate>.k). Open that artifact from the run and read its header — the $ Reference: and $ Confidence level: lines are right there, along with the card's material type, family and group. See sessions, runs and artifacts.

That is also the record to keep if you need to justify a result later. The card bytes are content-addressed and hash-verified, so the artifact stored with the run is provably the card that produced it.

A source citation is not a validation claim#

This is the part to be clear about.

A $ Reference: line says where the transformation data on that card came from. It does not say that Phases' predictions for that grade have been compared against measurements, and it does not say the card was fitted or tuned by anyone here. Those are separate activities with a separate record.

The product's calibration and validation work covers a much smaller set of grades than the catalog does — read how we validate for exactly which grades each result set covers, and known limits for what the models do not attempt. Provenance and validation are different axes: a grade can have an excellent, clearly cited 1974 atlas behind its card and still have never been checked against a hardness measurement in this product.

What is uniform, and what is not#

Uniform, across all 2,060 grades:

• quality_state is verified
• the card schema is v2606
• there is no integrity error on any card
• the deck and chemistry hashes are registered and re-checked on use

Not uniform:

• Transformation kinetics. 2,053 grades have readable authored kinetics. Seven do not — hc420, 13cr-4ni, if, if1, 18cr, 316l and saf2507 — and those refuse heat treatment, CCT and the thermal V-Gleeble modes with Required material physics is unavailable. Read the first id carefully: hc420 is a different card from the reference grade hc420la, which does resolve. The other six are not a random selection — they are the stainless and interstitial-free end of the catalog. hc420, filed under general engineering steel, is the odd one out.
• RSW bulk properties. The spot-welding FEM's electrical, thermal and mechanical property database is calibrated for twenty grades only, regardless of how many cards the catalog holds. See resistance spot welding.
• Bulk flow surfaces. V-Gleeble's deformation modes need a tabulated flow surface, which exists for a small set of grades and not for the catalog at large.

None of that is fixed forever — the catalog and its capability coverage both change — which is why every simulator resolves its requirements at run time and reports what it found rather than relying on a table in a document. The current state of play is in the capability matrix.