RSW validation

If you are going to use a simulated nugget diameter to argue about a weld schedule, you need to know two things: what the number was tuned to hit, and which parts of the model behind it are not validated at all. Both are recorded, and the second list is longer than most people expect.

The acceptance criterion#

Nugget diameter is judged against the geometric acceptance rule used in automotive spot-weld qualification, where t is the thickness of the thinner sheet in millimetres.

The calibration targeted the band between the 5·√t minimum and the 5.5·√t nominal.

Be precise about what that means. Agreement here is between a simulated nugget diameter and the diameter a standard asks for. It is not certification to AWS D8.1M, DVS 2902 or ISO 18278, and the calibration record does not include destructive testing of physical coupons welded to these schedules. A simulated nugget in the band is evidence the thermal model is behaving; it is not a qualified weld.

The calibrated templates#

Twenty-two calibrated schedules ship with the product: 17 symmetric single-grade cases and 5 dissimilar joint pairs. Ask for them in chat:

Show me the calibrated RSW welding templates

Each template fixes the grades, thicknesses, coating, welding current, electrode pressure and timing. All 22 use a 3.0 second simulated cycle with the welding current switched off at 0.42 s.

The five dissimilar templates are DC04 0.8 + DC04 1.5 mm at 5.7 kA / 23.33 MPa; DC04 0.8 + DP780 1.5 mm at 5.5 kA / 32.7 MPa; DC04 1.0 + 22MnB5 1.0 mm and DC04 1.0 + 22MnB5 1.5 mm, both at 4.5 kA / 32.7 MPa; and DP600 1.2 + 22MnB5 1.5 mm at 5.0 kA / 37.3 MPa.

Electrode pressure follows strength class rather than being fitted per grade: 23.33 MPa for mild steel, 28.0 MPa for HSLA, 32.7 MPa for the higher HSLA and dual-phase grades and EN13261, 37.3 MPa for DP780, and 42.0 MPa for the press-hardening steels. Zinc is enabled for the DC, HC and DP grades. It is disabled for 20MnB5 and 22MnB5, whose real coating is AlSi and is not modelled, and for EN13261, which is bare.

One caveat on the currents. The nugget diameters quoted in the calibration reports were measured with the currents used in those runs, and the shipped template files carry currents re-issued after a resistivity correction. The table above is what the product loads. Take the current from the template rather than from a report, and read the nugget diameter the run itself returns.

How close the agreement is#

The thermal-only calibration run covered all 17 symmetric grade and thickness combinations with no errors, and reports every nugget within one mesh step of its target band. The five dissimilar joints all landed inside their target ranges.

With the experimental mechanical path enabled, the same 17 cases give 11 inside the target band and 6 within one mesh step of it.

That mesh step matters. Nugget diameter is quantized by the element size of the fixed 2D axisymmetric mesh — the validation report records steps of about 0.71 mm — so the diameter does not resolve continuously. The 1.5 mm mechanical cases sit 0.03 mm below the 6.10 mm minimum, which is a mesh-resolution artifact rather than a physics failure; a finer mesh or roughly +0.5 kA moves them into the band. Read a simulated diameter with that granularity in mind before treating a 0.1 mm difference between two schedules as real.

Two supporting checks are also recorded: the thermal solver produced physical nuggets in 22 of 22 cases, and DC04's electrical conductivity was corrected during validation to 6.85 1/(mΩ·mm) at room temperature — a resistivity of 146 nΩ·m — after the previous value, roughly twice as high, implied a resistivity below that of pure iron and so was physically impossible.

One result that surprises people is the electrode temperature. In the calibrated cases the electrode contact face exceeds the melting point of copper (1083 °C) during the weld. That is recorded as physically normal for a thin contact region heated by contact resistance, with water cooling bringing the electrode back to roughly 60–110 °C within the 3 s simulated cycle. It is not an electrode-failure prediction.

Hardness in the weld#

When microstructure tracking is enabled, hardness comes from the same routine described in hardness validation. For DC04 0.7 mm at 8 kA the predicted profile is:

The gradient from base metal to nugget is recorded as physically correct. The nugget value sits at the high end of the literature range and is defensible as the theoretical maximum for untempered 100% martensite; real nuggets run roughly 10–15% lower, around 240–260 HV for DC04, because of auto-tempering during cooling, which the model does not represent. For comparison the report cites measured DC51D spot-weld data up to 300 HV0.2, and the Pouranvari 2011 "twice base metal" rule at 200–240 HV.

Two practical conditions apply. Hardness only appears when microstructure tracking is on, which requires the external metallurgy worker. And it is only reported when the model has cooled below 200 °C by the end of the simulated time — a run that ends hot returns no hardness. When microstructure data is unavailable, post-processing falls back to a different rule-of-mixtures formula that gives a noticeably higher value for DC04.

What is not validated#

• The mechanical path is experimental and linear. The active solve is linear thermoelastic with contact. The repository contains an elastoplastic integrator and nonlinear return mapping, but the active mechanical entry point never calls them. Displacement, von Mises stress and the deformed contour must not be presented as validated plasticity, electrode indentation, sheet thinning or expulsion.
• The flow_stress_model setting does nothing. Both simple and tables reach the same linear solve. Historical runs comparing the two returned identical displacement for that reason, which is not evidence that the two constitutive descriptions are equivalent at welding temperatures.
• An applied Phases flow-stress card has no mechanical handoff to RSW. A card affects an RSW run only through the optional microstructure path, where it supplies chemistry and transformation kinetics.
• The mechanical-mode current multipliers (0.85–1.08 by grade family) are empirical. They belong to the linear contact implementation and cannot be attributed to plastic sheet thinning or grade-dependent yielding until a nonlinear path is activated and revalidated.
• The Kaars et al. 2018 case is a reference, not a validation. A reference case reproduces the setup from Kaars, Mayr & Koppe, Determining Material Data for Welding Simulation of Presshardened Steel, Metals 2018, 8(10), 740 — 22MnB5+AS150, 1.5 mm symmetric, 5 kA for 100 ms then 8 kA for 300 ms, 6 kN. That paper validates its own model against experiment. The record states plainly that this is not evidence that this solver reproduces the published indentation, and the paper publishes no hardness profiles.
• Coverage is limited. RSW runs only on the 20 grades with validated bulk electrical, thermal and mechanical properties, and only 17 of those have a calibrated template — 20MnB8, 25MnB5 and 29MnB7 do not. Catalog membership never implies RSW support; see the capability matrix.

Where to go next#

To run one of these schedules or vary it, see resistance spot welding and the weld schedule sweep recipe. For the model-wide caveats, see known limits and how we validate. A run and its stored plots are described in sessions, runs and artifacts.

Weld DC04 0.8 mm to DP780 1.5 mm and show me the nugget diameter