Resources Numerical accuracy Calculation policy
The defaults and checks of a nonlinear or contact study, and the time steps and loads of a thermal stress history.
These functions build the inputs and read the results of solvers that live in other files. The arithmetic here is small and 64-bit. The solver tolerance and step counts set the accuracy of the results.
4 calculations in 2 source files. The code is read from the released app when this page loads. It is the calculation people run, not a copy kept for this page.
Tests None found by the review
Sets the slope after yield when none is given and flags one that is not below E.
Et = E ÷ 100 when not given
One 64-bit division. The check Et < E is a plain comparison with no tolerance.
const Et = nl.Et === null ? (mat ? mat.E / 100 : 0) : nl.Et;
if (mat && !(Et < mat.E)) problems.push(t('study.nl.problem.et_too_big', { Et, E: mat.E }));Flags a contact penetration over 1 percent of the element size, with the values it reports.
problem when max penetration > 0.01 × h; pct = 100 × pen ÷ h
The comparison is exact. The reported values are rounded to 3 significant digits and the percent to 2 decimals, in the message only.
const c = n.contact;
if (c && c.h > 0 && c.maxPenetration > PEN_LIMIT * c.h) out.push({
id: src + ':nl.penetration', code: 'sim.nl.penetration', severity: 'warning',
msg: { key: 'study.nl.p.penetration', vars: { pen: +c.maxPenetration.toPrecision(3), h: +c.h.toPrecision(3), pct: +(100 * c.maxPenetration / c.h).toFixed(2) } },
about: [{ kind: 'study', id: 'simulation', label: t('study.nl.p.about') }],
...(c.at ? { where: { space: '3d', at: c.at } } : {}),
source: { kind: 'engine', name: src },
req: { op: '<=', value: +(PEN_LIMIT * c.h).toPrecision(3), unit: 'mm' }, found: { value: +c.maxPenetration.toPrecision(3), unit: 'mm' },
});Tests tools/partcraft-test/test-thermal.mjs
Chooses which time steps of the heat run get a stress solve.
stepᵢ = round(i × steps ÷ (n − 1)), i = 0…n−1
Integer steps from one 64-bit division and rounding, with duplicates removed. n is limited to 2 to 51.
/** The time steps solved for stress: [step] for 'time', else `solves` steps evenly over the run, 0 (the start) and the end included. */
export function stressSteps(tr, tl) {
if ((tl.at || 'worst') === 'time') return [stepAt(tr, tl.time).step];
const n = Math.max(2, Math.min(MAX_SOLVES, Math.round(tl.solves || 21)));
const out = new Set(); for (let i = 0; i < n; i++) out.add(Math.round(i * tr.steps / (n - 1)));
return [...out].sort((a, b) => a - b);
}The load vector of each stress step and the temperature range of the worst step.
ΔT = T − Tref; f = f_mech + ft(ΔT)
ΔT and f are Float64Array. Frames of T come from the heat run, so their type is whatever thermal.js gives (not read here). The reported temperature range is taken from the 64-bit ΔT copy of the worst step.
for (let n = 0; n < s.nn; n++) dT[n] = s.active[n] ? T[n] - Tref : 0;
const ft = TH.thermalLoadVector(s.mesh, s.mat, dT);
for (let k = 0; k < f.length; k++) f[k] = fm[k] + ft[k];The arithmetic is 64-bit, but the method or the data sets the accuracy: a solver tolerance, a mesh, a step size, a datum shift, a table. The limit is stated where the result is shown.