Whet pixel markWhet

Validation

Four textbook cases with closed-form answers, built and solved in Whet, beside the formula.

Method

Each part is built with Whet's own CAD verbs. It is solved with the same gmsh mesh and CalculiX deck a study in the window uses, at two or three mesh sizes, with second-order tetrahedra throughout. The analytic value comes from the formula named in the row; the error is Whet's finest-mesh figure against it, unrounded. The material card is Aluminum 6061 T6, with the modulus, Poisson ratio, density and conductivity read off the card. Nothing was adjusted after the first run.

Cantilever beam, tip load

QuantityFormulaAnalyticWhet, finest meshErrorMeshAcross the meshes
Tip deflectiond = F L^3 / (3 E I)0.5806 mm0.5800 mm-0.09 %1.25 mm, 39491 nodes, 24893 elements5 mm: 0.5784; 2.5 mm: 0.5795; 1.25 mm: 0.5800
Bending stress at the rootsigma = F L (h/2) / I60.00 MPa71.44 MPa+19.06 %1.25 mm, 39491 nodes, 24893 elements5 mm: 53.79; 2.5 mm: 59.18; 1.25 mm: 71.44
Bending stress half way alongsigma = F (L/2) (h/2) / I30.00 MPa30.03 MPa+0.11 %1.25 mm, 39491 nodes, 24893 elements1.25 mm: 30.03
First natural frequencyf1 = (1.875^2 / 2 pi L^2) sqrt(E I / rho A)816.0 Hz814.3 Hz-0.22 %1.25 mm, 39491 nodes, 24893 elements1.25 mm: 814.3
  • Tip deflection: Euler-Bernoulli; with the Timoshenko shear term (kappa 5/6) 0.5852 mm; mean over the loaded tip face.
  • Bending stress at the root: peak von Mises; the fully held root face is singular; convergence verdict on the peak: growing, last move 17.15 percent.
  • Bending stress half way along: surface stress off the field file at x = 50 mm; 9 vertices, spread 0.33 MPa.
  • First natural frequency: Euler-Bernoulli, fixed-free; the square section pairs the mode; modes 1 and 2: 814.26 and 814.27 Hz.
cantilever/cantilever.cadstress.json 3bd1af30ea69; 2026-10-12

Plate with a central hole, tension

QuantityFormulaAnalyticWhet, finest meshErrorMeshAcross the meshes
Peak stress at the holesigma_max = Kt sigma_net, Kt = 3.000 - 3.140 (d/W) + 3.667 (d/W)^2 - 1.527 (d/W)^331.33 MPa30.98 MPa-1.13 %0.5 mm, 98541 nodes, 55609 elements2 mm: 29.77; 1 mm: 30.68; 0.5 mm: 30.98
Peak stress at the hole, force loadthe same, the 250 N quarter load as a uniform traction on the end face31.33 MPa30.98 MPa-1.13 %0.5 mm, 98541 nodes, 55609 elements2 mm: 29.77; 1 mm: 30.68; 0.5 mm: 30.98
  • Peak stress at the hole: Peterson (Pilkey, chart 4.1), d/W = 0.2, Kt = 2.506 on the net section, sigma_net = 12.5 MPa; the end pulled as a pressure; convergence verdict on the peak: settled, last move 0.95 percent.
  • Peak stress at the hole, force load: the end pulled by a total force, applied as the consistent nodal loads of a uniform traction; convergence verdict on the peak: settled, last move 0.95 percent; the finest peak sits 0.08 mm from the hole edge, 49.92 mm from the loaded face, against 30.98 MPa for the same load as a pressure.
plate-hole/plate-quarter.cadstress.json 04adc432d8dc; plate-hole/plate-quarter-force.cadstress.json b592474969a0; 2026-10-12

Thick-walled cylinder, internal pressure

QuantityFormulaAnalyticWhet, finest meshErrorMeshAcross the meshes
Hoop stress at the boresigma_t = p (ro^2 + ri^2) / (ro^2 - ri^2)26.00 MPa25.86 MPa-0.54 %1 mm, 109008 nodes, 73827 elements1 mm: 25.86
Von Mises at the boresqrt(((sigma_t - sigma_r)^2 + sigma_r^2 + sigma_t^2) / 2), sigma_r = -p, open ends32.19 MPa31.97 MPa-0.69 %1 mm, 109008 nodes, 73827 elements4 mm: 31.56; 2 mm: 31.80; 1 mm: 31.97
  • Hoop stress at the bore: Lame; the largest normal stress in the global frame, the hoop on the symmetry faces; the record's own vessel check states the same 26 MPa.
  • Von Mises at the bore: Lame; the study's peak von Mises; convergence verdict on the peak: settled, last move 0.53 percent.
thick-cylinder/tube-quarter.cadstress.json ffac73af3c09; 2026-10-12

Slab, one face held, one face in convection

QuantityFormulaAnalyticWhet, finest meshErrorMeshAcross the meshes
Heat flow through the slabQ = A (Th - Ta) / (L/k + 1/h)94.35 W94.35 W0.00 %2.5 mm, 23884 nodes, 15410 elements10 mm: 94.35; 5 mm: 94.35; 2.5 mm: 94.35
Temperature of the convective faceTc = Ta + q / h95.48 °C95.48 °C0.00 %2.5 mm, 23884 nodes, 15410 elements10 mm: 95.48; 5 mm: 95.48; 2.5 mm: 95.48
  • Heat flow through the slab: k = 167 W/mK off the card; heat out of the held face; a linear field is exact on any mesh; this checks the boundary bookkeeping.
  • Temperature of the convective face: mean over the film faces.
slab/slab-a.cadthermal.json 87418b4c2de5; slab/slab-b.cadthermal.json 9b534332effc; slab/slab-c.cadthermal.json a85ae165d646; 2026-10-12

How to read it

A figure that settles under mesh refinement is resolved by the mesh. A peak that keeps growing is singular: a fully held face or a point load has no finite stress there, and the usable figure is the one read a few elements away. Both kinds are shown as they came out.

Rerun it

One command in the Whet repository runs every case. It writes the part specs, the study specs, each solver result with its hash, and the summary the tables above are printed from. The first case is shown as the two files it ran from:

node tools/validation/run.mjs
{
  "name": "cantilever",
  "material": "Aluminum-6061-T6",
  "solid": {
    "kind": "box",
    "sizeMm": [
      100,
      10,
      10
    ]
  },
  "expect": {
    "solids": 1,
    "sizeMm": [
      100,
      10,
      10
    ],
    "volumeMm3": 10000
  }
}
{
  "name": "cantilever",
  "of": "cantilever.step",
  "material": {
    "name": "Aluminum-6061-T6"
  },
  "restraint": {
    "atMm": [
      0,
      5,
      5
    ],
    "normal": [
      1,
      0,
      0
    ]
  },
  "load": {
    "atMm": [
      100,
      5,
      5
    ],
    "normal": [
      1,
      0,
      0
    ],
    "forceN": [
      0,
      0,
      -100
    ]
  },
  "meshSizeMm": 5,
  "modal": {
    "modes": 6
  },
  "convergence": {
    "sizesMm": [
      5,
      2.5,
      1.25
    ],
    "budgetSeconds": 900
  }
}

Each study is a plain Whet study. The same two files run in the window's Study panel or on the command line of any licensed copy. Strength and heat explains the results.

Saved results

gmsh 4.15.0, CalculiX 2.23, FreeCAD 1.1.3 with OpenCASCADE 7.8.1, run 2026-10-12. The summary is tools/validation/results-2026-10-12/validation.json; each table names its record files and the first twelve characters of each file's SHA-256.