Known limits

Every model has a boundary. This page is where the boundaries are written down, so you can find out before a decision rather than after one. Each entry states the limit and what to do instead.

Read this alongside how we validate, which describes the models and the calibration record these limits qualify.

Heat treatment and CCT#

Mild, HSLA and DP grades over-harden at ordinary cooling rates. This is the largest single defect in the product and it is measured, not inferred. The Kirkaldy alloy factor is inactive (see the next entry) and the fourteen legacy DC, HC and DP decks share one rate-constant set, so all of them inherit a critical cooling rate of roughly 10–20 °C/s. Measured on the shipped worker, cooling from 900 °C:

A 0.08 %C mild steel cannot be hardened by air or oil cooling, and the first two rows say it can. The 22MnB5 row is the same problem in the other direction: that deck's equilibrium ferrite curve is capped at 0.750, so every rate at or below about 10 °C/s returns exactly 75 % ferrite (the catalog card caps it at 0.7353, a 73.5 % ceiling; it was not measured at this rate). Do not use a Phases phase fraction or hardness in the 10–100 °C/s band for mild, HSLA or DP grades. That band was never validated. The DC04 cases are pinned as strict expected failures in the backend test suite, so the gap cannot be closed quietly — recalibrating the cards turns them into failures that force the pin to be re-stated.

Composition does not change how fast a diffusional transformation proceeds. The Kirkaldy alloy factor Fc is identically 1.0 in the shipped solver: it is set to a constant before the phase loop and its routine returns 1.0 on every branch. Chemistry therefore has no path at all into the ferrite, pearlite or bainite rate. It reaches a result only through the card's critical-temperature composition offsets, which move where transformation starts, and through the hardness correlation, which reads composition directly. Sweeping DC04's carbon from 0.02 to 0.80 wt% at 20 °C/s moves predicted hardness from 223 to 872 HV while the microstructure stays 98–100 % martensite throughout — the hardness model sees the chemistry, the kinetics do not.

TTT diagrams are not simulated. Only continuous-cooling transformation is simulated. Asking the transformation-diagram tool for a TTT or isothermal diagram returns "TTT transformation diagrams are not implemented yet." The Material Base transformation panel does draw calculated TTT curves, but those are computed from the card's kinetics parameters, not from a simulated isothermal hold — do not read them as simulation output. For simulated isothermal behaviour at a single temperature, run a V-Gleeble isothermal hold instead — see virtual Gleeble.

The critical-temperature lines are card predictions, not measured equilibrium temperatures. Ae3, Ae1, Bs and Ms are read from the exact material card as a phase-specific base plus its composition-coefficient row, so they vary by grade and by an applied chemistry override. The plotted Ae3 (Fer Tcrit) and Ae1 (Pea Tcrit) are the card's ferrite and pearlite kinetic critical temperatures; they are not the separate equilibrium-curve breakpoints and have not been independently calibrated for every grade. Use measured or separately validated transformation temperatures when an austenitizing decision depends on them.

Fourteen decks share one kinetics constant set. DC01, DC03, DC04, DC05, DC06, HC260LA, HC300LA, HC340LA, HC380LA, HC420LA, DP450, DP500, DP600 and DP780 carry identical ferrite, pearlite and bainite activation energies, grain size and fitted exponents; the simulator-only EN13261 deck shares the same set. For those grades, composition-aware critical-temperature lines do not make the transformation kinetics grade-calibrated, and every one of them inherits the same critical cooling rate. The rest of the catalog carries per-grade authored constants — ASTM grain size alone spans 4.89 to 13.00 across roughly 1,880 distinct values, and every boron press-hardening deck has its own — but authored per-grade constants are not the same as per-grade experimental CCT calibration, which is still incomplete there too.

Pearlite cannot be tuned independently of bainite. Bainite's equilibrium flag forces its available fraction to 1.0, so it can consume all remaining austenite, and its critical temperature sits next to pearlite's. Separating them needs a distinct pearlite transformation path.

Dual-phase steels are simulated from full austenitization. The model does not represent the intercritical annealing route by which real DP steel is produced, so a DP starting microstructure here is an imposed state, not a simulated production history.

Boron is not modelled explicitly. 20MnB5 and 22MnB5 behave reasonably because 22MnB5 was the original calibration basis, not because boron's pinning of austenite grain boundaries is represented.

Documented accuracy gaps. These are recorded in the validation runs and are not fixed. Note the cooling rates: every hardness row below was measured at 0.01 or 100 °C/s, and the 10–100 °C/s band in between has no validation record at all.

The kinetics model has never been validated point-by-point against specific published CCT curves. That remains an open recommendation in the V&V record.

Resistance spot welding#

Twenty grades only. RSW needs validated bulk electrical, thermal, density, latent-heat and mechanical property curves, and only these have them: DC01, DC03, DC04, DC05, DC06, HC260LA, HC300LA, HC340LA, HC380LA, HC420LA, DP450, DP500, DP600, DP780, 20MnB5, 22MnB5, 20MnB8, 25MnB5, 29MnB7, EN13261. The material catalog is roughly a hundred times larger; membership in it does not imply an RSW mapping. Anything else is rejected before the solver runs. To weld a custom chemistry, ride it on one of these base grades — see your own chemistry.

Those 20 grades collapse into two property families. All DC and HC grades use the DC04 property set; all DP, all MnB press-hardening grades and EN13261 use the 22MnB5 property set. Two grades in the same family share identical electrical, thermal and linear-mechanical curves. Only the microstructure path uses each grade's own card.

Three grades have no calibrated template. 20MnB8, 25MnB5 and 29MnB7 are supported for RSW but have no shipped calibrated case file; the 22 templates cover 17 symmetric grades plus 5 dissimilar joints.

Mechanics is a linear thermoelastic contact solve, and it is experimental. The active mechanical entry point calls the linear solver directly. An elastoplastic integrator, a nonlinear return-mapping method and both flow-stress providers exist in the code but are never called by the active solve. Consequences:

• Displacement, von Mises and deformed-geometry output must not be presented as validated plasticity, electrode indentation, sheet thinning or expulsion.
• The flow_stress_model field (simple or tables) is inert. Changing it does not change the mechanical result.
• Applying a Phases flow-stress card has no RSW mechanical handoff. An applied card reaches RSW only through the optional microstructure path, where it supplies chemistry and transformation kinetics.
• The empirical current multipliers recorded for mechanical mode (0.85–1.08 by grade family) belong to the linear contact model and cannot be attributed to plastic yielding.

Nugget diameter is mesh-quantized. The validation report records steps of about 0.71 mm, which is why several 1.5 mm cases sit 0.05 mm below the 6.12 mm minimum — a shortfall more than ten times smaller than one mesh step. Treat a difference smaller than one mesh step as noise, not as a schedule improvement.

Nugget hardness is a theoretical maximum. DC04 is predicted at 280 HV in the nugget — defensible for untempered 100 % martensite, but at the high end of the 200–300 HV literature range. Real nuggets run roughly 10–15 % lower (about 240–260 HV for DC04) because of auto-tempering during cooling, which the model does not represent.

Electrode contact-face temperatures exceed copper's melting point in the calibrated cases. That is recorded as physically normal for a thin contact region heated by contact resistance, with water cooling recovering to roughly 60–110 °C within the 3 s simulation. It is not an electrode-failure prediction.

Virtual Gleeble#

Deformation modes need a bulk flow table. Hot tensile, hot compression and Satoh require an exact per-grade table or an explicitly reviewed grade mapping. Exact tables ship for 17 grades: DC01, DC03, DC04, DC05, DC06, HC260LA, HC300LA, HC340LA, HC380LA, HC420LA, DP450, DP500, DP600, DP780, 20MnB5, 22MnB5, EN13261. 20MnB8, 25MnB5 and 29MnB7 have no exact table and run through a reviewed 16MnCr5 mapping with strength scales of 1.05, 1.18 and 1.25 respectively, and the result carries an explicit surrogate warning.

Custom chemistry cannot run a deformation mode. You get a typed refusal: No validated custom V-Gleeble bulk flow surface exists for {material}; phase P1-P4 is not a compatible substitute. A calculated phase flow-stress result is a different capability and is never a substitute. Thermal modes — CCT dilatometry, HAZ, isothermal, hot stamping, custom — do work with custom chemistry.

Flow stress above 1000 °C is systematically 50–100 % too high. Hot compression at 1000 °C and 10 s⁻¹ is roughly twice published values for 22MnB5 (235 MPa against 80–120). The tables are scaled from room-temperature UTS ratios, which does not capture dynamic recovery and recrystallization in austenite. Separately, 22MnB5 hot tensile flow stress at 800 °C sits about 50 % above Merklein & Lechler (2006), because the boron effect on austenite flow stress is not modelled.

The exact product-grade tables cover 20–1200 °C at 13 temperature points, and several source MSC Marc datasets had gaps — one jumped from 200 °C straight to 900 °C. Missing temperatures in those product tables were filled synthetically: Eurocode EN 1993-1-2 reduction factors below Ac1, the Shida (1969) austenite formula above Ac3, and a linear blend between. The 16MnCr5 surrogate used for 20MnB8, 25MnB5 and 29MnB7 ends at 1100 °C. Those are source-grid facts, not per-material product limits: hot-deformation cards and backend admission use the uniform solver envelope. A request inside that envelope but outside the resolved source grid is extrapolated from its nearest temperature and log-rate planes, so it is a model prediction rather than validation evidence.

Materials and custom chemistry#

Catalog membership is a verified card, not a calibration claim. This is the limit most likely to be misread, so it is worth stating flatly. The built-in catalog holds about 2,060 grades, and every one of them ships marked verified on the v2606 card schema. That marking means the deck parses, its curves are complete, and its integrity checks pass. It does not mean the grade was individually validated against experiment.

The published validation record covers a small reference subset:

Everything outside those sets — the great majority of the catalog — has never been checked case-by-case against measured data in this product. Most cards do cite a published source in their reference field; the largest groups are HTDG 2008 (380 cards), the NIMS database (194) and the Climax Molybdenum atlas (120), while 21 cards cite no source at all. A cited source is provenance for the card's authored data, not a Phases validation run. See families and data sources for where the data came from, and the material catalog for how to find a grade.

The practical rule: if your grade is not in one of the four rows above, treat a Phases result on it as a model prediction from an authored card, and say so when you quote it.

Confirming a custom chemistry is not a calibration claim. It stores a stable material definition and one compatible base-card template. It does not calculate transformation kinetics, phase flow stress or a V-Gleeble bulk flow surface. Each simulator resolves what it needs afterwards and reports unavailable or fallback states explicitly. The full workflow is on your own chemistry; which simulator needs which capability is on the capability matrix.

A preserved-source kinetics result is not interpolation. When no published kinetics model applies to a custom chemistry, heat treatment, CCT, V-Gleeble and microstructure-enabled RSW may retain the verified template's authored kinetics and warn about it. Those kinetics belong to the template, not to your chemistry.

The nearest-grade distance badge measures chemistry proximity to a catalog template. It is not a kinetics-accuracy or calibration score. The simulator's own capability result is authoritative.

Only 20 element rows exist on a card: C, Mn, Si, Ni, Cr, Mo, Cu, S, P, Al, As, W, V, Ti, Bo (entered as B), Nb, N, O, Fe, Zr. A chemistry containing a non-zero element that no verified template can represent is refused rather than silently dropped.

The hardness correlations were fitted over narrow carbon ranges. Kasuya-Yurioka martensite is stated valid 0.12–0.50 %C, Kasuya-Yurioka non-martensite 0.02–0.30 %C, and Maynier bainite 0.10–0.50 %C. The 25-case validation spans 0.02–0.40 %C. Outside that span the model still returns a number, and that number has no validation behind it.

Cross-cutting#

Predicted hardness is a bulk mixture value with hard floors — 60 HV for ferrite+pearlite, 150 HV for bainite, 50 HV for the total. It cannot be reinterpreted as the hardness or ductility of an individual phase. The hardness coefficients are phase-only with fixed values; composition-dependent coefficients are recorded as future work.

A cooling rate is not a quench medium. Phases models the rate you supply. It will not label a rate as water, oil or air, and neither should a report drawn from it.

An imposed initial state is not a solver finding. A fully austenitic start or a hold segment is an input. It is not evidence that austenitization completed.

Questions Phases does not answer at all. Weld preheat requirements, toughness, quench-cracking risk, solute retention and precipitation, and age-hardening response are outside the four solvers. They cannot be inferred from composition or phase fractions, and the product will say so rather than guess.

There is no default material. Asking for "a simulation" without naming a grade, a workspace material or a chemistry returns a question, not a run. Start from the material catalog.

If a limit here decides your case, the honest next step is a physical test or an independently validated calculation. Phases is a fast way to narrow the space you test in; it is not the test.