forked from 1b-agent/fieldcraft-upstream
Christian Svensson
e9ccf10692
Demonstrate temperature slices and signed face-average wall heat with cool air passing an isothermal bracket. Register the original CPU-rendered gallery thumbnail and document the coarse transient setup. Part of #157.
222 lines
14 KiB
Markdown
222 lines
14 KiB
Markdown
# Examples
|
||
|
||
Every example is a built-in sample: open it from **File ▸ Samples** (or the start screen), with
|
||
`fieldcraft --sample NAME`, or with the `file.sample` command (`{"name": "drop-test"}`), then
|
||
press **Solve** (`solve.run`). Each opens fully defined — geometry, mesh, materials, boundary
|
||
conditions, loads and steps — in mm, t, s, N and MPa, and opens and solves within half a minute. The geometry is our
|
||
own: SolidCraft scripts run when the sample opens (a recipe from [`tests/corpus`](../tests/corpus)
|
||
or one written in `crates/engine/src/sample.rs`). Each sample has a test in
|
||
`crates/engine/src/sample.rs` or its modules that opens it, checks it clean, solves it and holds its result to
|
||
the check given below.
|
||
|
||
## Drop test (`drop-test`)
|
||
|
||
|
||
|
||
An aluminium L-bracket (60 × 40 mm foot, 50 mm upright, 8 mm thick, R6 fillet) falls 2 m onto a
|
||
rigid floor tilted 15°, so the end of its foot strikes first. Explicit dynamics for 1 ms with
|
||
gravity: 1 911 swept hex8 elements (one-point, stiffness hourglass control), bilinear plasticity
|
||
(yield 240 MPa), the floor a rigid wall. About 8 400 increments, a few seconds.
|
||
|
||
**Look at:** the frames from the strike (Results ▸ Frame): the compression wave leaves the
|
||
struck corner and the upright's inertia bends the fillet, where von Mises peaks near 130 MPa —
|
||
below yield, so the drop leaves no permanent set. The energies plot shows the kinetic energy
|
||
turning into vibration with a balance under 0.01 %.
|
||
|
||
**Check:** the mass equals ρ V of the bracket (0.5 %); it meets the floor at √(2 g h) =
|
||
6.26 m/s, with ½ m v² = 1 748 mJ of kinetic energy (1 %); energy is conserved (balance < 1 %);
|
||
by 1 ms the floor has taken about half its downward momentum.
|
||
|
||
## Rubber seal (`rubber-seal`)
|
||
|
||
|
||
|
||
A rubber gasket ring (20 mm inside radius, 30 mm outside, 10 mm high) squeezed 2.5 mm (25 %)
|
||
between two plates bonded to it. Generic rubber, E = 6 MPa, ν = 0.49, Mooney–Rivlin with
|
||
C01 = 0.2 MPa; 1 000 hybrid (u–p) hex8 elements in four mapped quarter blocks. Nonlinear static
|
||
with large deformation: Newton–Raphson in nine increments, under a second.
|
||
|
||
**Look at:** the deformed shape at scale 1 — the free inner and outer walls bulge out, and
|
||
von Mises peaks (about 7 MPa) where the bulge meets the bonded plates. Results ▸ Increment steps
|
||
through the converged increments.
|
||
|
||
**Check:** the same ring between frictionless plates compresses evenly, and its force matches
|
||
the closed form for an incompressible Mooney–Rivlin solid in uniaxial compression,
|
||
P = 2 (λ − λ⁻²)(C10 + C01/λ) per original area: 3 430 N against 3 466 N (within 3 %). Bonded,
|
||
the bulge makes it stiffer (shape factor 0.5): 6 518 N, held between 1.05 and 2.5 times the
|
||
frictionless force. The groove and the plates are prescribed displacements today; with implicit
|
||
contact (#161) the ring will be pressed by contact against a groove.
|
||
|
||
## Flow past a bracket (`flow-past`)
|
||
|
||
Air (ρ = 1.2 kg/m³, μ = 1.8 × 10⁻⁵ Pa s) at 25 mm/s past the same L-bracket standing on the
|
||
floor, the upright's back to the wind. A 240 × 120 × 100 mm box of 2.5 mm cells (184 320),
|
||
uniform inlet, outlet at zero pressure, a no-slip floor and slip walls at the sides and top;
|
||
D3Q19 lattice Boltzmann (TRT) for 16 s, about 6.7 flow-throughs, ten frames. Reynolds number
|
||
83 on the upright's 50 mm height: laminar, a steady wake. About 10 s.
|
||
|
||
**Look at:** the step's report (`solve.fluid`, or Results after Solve): the force on every
|
||
named face of the bracket and on each side of the box, final and time-averaged, and the mass
|
||
history; `solve.probe` with `points` gives the velocity and pressure in any cell, and with
|
||
`forces` each face's force history.
|
||
|
||
**Check:** the drag on the bracket, as a coefficient on the upright's frontal area
|
||
(40 × 50 mm), is 2.3 — a bluff plate standing on a wall at low Reynolds number in a box it
|
||
blocks by 17 %, held between 1 and 3.5 (a flat plate broadside to the flow is about 1.2 at high
|
||
Reynolds number and rises as it falls); the lattice's mass changes by less than 1 %.
|
||
|
||
|
||
|
||
The picture shows the bracket as the sample opens. After Solve the viewport draws the flow:
|
||
velocity magnitude on a slice through the middle of the box (`results.slice`; pressure with
|
||
`field: "pressure"`), the bracket coloured by the flow beside it (`results.surface`) and
|
||
streamlines seeded upstream of it, coloured by speed (`results.streamlines`).
|
||
|
||
## Heated bracket (`heated-bracket`)
|
||
|
||
|
||
|
||
The corpus filleted bracket (`tests/corpus/filleted-bracket`: an 8 mm L-bracket, R6 fillet, a
|
||
hole through each leg) in steel, both hole bores clamped. The end of the upright is held at
|
||
120 °C by a heater; the rest of its outside loses heat to still air at 20 °C
|
||
(h = 10 W/(m² K)); steel conducts with k = 50 W/(m K), expands with α = 12 × 10⁻⁶/K and is
|
||
stress free at 20 °C. One coupled temperature–displacement step (steady) on about 23 000 tet10
|
||
elements (the count differs a little between platforms): about 5 s to open (meshing) and 13 s to
|
||
solve.
|
||
|
||
**Look at:** Contours ▸ Temperature: the upright stays within a few degrees of the heater and
|
||
the foot falls to 95 °C, a steady balance between conduction and the air. Contours ▸ Von Mises
|
||
shows the stress of the growth the clamps hold back: tens of MPa through the legs, peaking at the
|
||
edges of the clamped bores, where a rigid clamp is a stress singularity (about 1.5 GPa there, a
|
||
figure that grows as the mesh is refined: read the stress a little away from the bore).
|
||
|
||
**Check:** the temperatures lie between the air's and the heater's and fall away from the heater.
|
||
For the expansion, the test puts the air at the heater's temperature, so the bracket warms
|
||
evenly by ΔT = 100 K, and clamps the foot's hole alone: the bracket then grows freely about it,
|
||
and a point moves α ΔT times its distance from the hole's axis. The upright's top moves −0.0432
|
||
mm along x (within 2 %); the foot's end, 16 mm from the clamped bore, which cannot grow, moves
|
||
0.0228 mm against 0.024 mm (within 6 %).
|
||
|
||
## Hex-meshed bracket (`bracket-hex`)
|
||
|
||
|
||
|
||
The same filleted bracket split into 30 cells in SolidCraft
|
||
(`tests/corpus/filleted-bracket-blocks`: planar cuts through the legs and diagonal cuts round
|
||
the holes and the fillet, each cell a six-sided block), meshed by `mesh.hex.blocks` as one
|
||
conforming all-hex mesh: 5 778 hex20 elements of 3 mm, O-grids round the holes, graded through
|
||
the fillet, every shared face meshed once. Steel; the upright's end clamped, 1 kN down on the
|
||
foot's end. Linear static: about 2 s to open (the cells are rebuilt from the script) and 3 s to
|
||
solve.
|
||
|
||
**Look at:** the mesh (Prepare ▸ Mesh) — mapped blocks meeting face to face with no pyramids or
|
||
tets — and the von Mises contours (tick Deformed for the deformed shape): the stress runs round the fillet and
|
||
peaks at 295 MPa at the upright's hole.
|
||
|
||
**Check:** the tip of the foot moves 0.509 mm down, within 1 % of a 48 016-element tet10
|
||
solution of the unpartitioned part (0.5093 mm, `crates/testkit/tests/bracket_blocks.rs`), and the
|
||
clamp carries the whole 1 kN.
|
||
|
||
## Mirrored crank arm (`crank-mirror`)
|
||
|
||
|
||
|
||
The model of `crank-arm` under the name of its mirror workflow (the sample kept so scripts and
|
||
links made with it still open): seen from below, where the pocket is. How the half is cut,
|
||
meshed and mirrored is told under `crank-arm`.
|
||
|
||
**Look at:** the pocket from below: straight mapped rows along its sides and floor, and at each
|
||
end its corner fillets as three-sided triangle primitives, cut off by the mesher's own blocks;
|
||
the two halves, mirror images across the arm's middle.
|
||
|
||
**Check:** as `crank-arm`.
|
||
|
||
## Bicycle crank arm (`crank-arm`)
|
||
|
||
|
||
|
||
A 170 mm aluminium crank built in SolidCraft (`crates/engine/samples/crank-arm.json`): a 16 mm
|
||
arm tapering from 38 to 24 mm, root arcs into a Ø46 × 22 spindle boss with a 12.7 mm square bore
|
||
(R1.5 corners) and a Ø30 × 20 pedal boss with a Ø14 hole, and an 82 × 14 × 8 pocket in its back
|
||
(R5 corners). It is a stack of extrusions along z, so it sweeps all-hex. To keep it sweepable the
|
||
hex variant leaves out the pocket's floor fillet and the bosses' 1.5 mm chamfers, and its root
|
||
arcs meet the bosses about 20° off tangent (a tangent meeting leaves a zero-angle cusp on the
|
||
arm's top face, where any hex is degenerate). 6061-T6 (E = 68 900 MPa, ν = 0.33); the square
|
||
bore's faces fixed, 1500 N on the pedal hole in the crank's plane, across the arm (−y). Linear
|
||
static: about 5 s to open and 7 s to solve.
|
||
|
||
The mesh needs one datum. A datum plane on the symmetry plane (`Sym`, y = 0) partitions the
|
||
crank (`Half`); `mesh.hex.generate` extrudes the half `CrankArm/a` (hex20 of 3 mm) and cuts its
|
||
cross-section into mappable blocks itself, inside the mesher (nothing is added to the model):
|
||
the half ring round each bore as four-sided blocks, the pocket's R5 corners as three-sided
|
||
triangle primitives, and four-sided blocks along the arm, split where the pocket's fillets meet
|
||
it. Interval assignment gives every block's sides matching counts, with no seeds.
|
||
`mesh.mirror` reflects the half onto `CrankArm/b`: 8 664 hex20 elements, all valid, scaled
|
||
Jacobian min 0.37 (at the corners where the root arcs meet the bosses 20° off tangent), mean
|
||
0.91. Before the mesher cut blocks itself the same crank took seven partitions and eleven exact
|
||
seeds for 0.37 (and the swept half without them: min 0.009, its middle free-paved).
|
||
|
||
**Look at:** the mesh (Prepare ▸ Mesh): rows running straight along the arm, the O-grids round
|
||
the bores, eight layers through the arm, and the von Mises contours: the bending stress grows
|
||
along the arm toward the root; the peak, 117 MPa, sits at a square-bore corner where the rigid
|
||
clamp meets the free face, a support singularity.
|
||
|
||
**Check:** σxx on the arm's sides at x = 29 mm is −62.44 and +62.44 MPa, equal and opposite as
|
||
the mirror-image halves are, against the beam formula
|
||
M/Z = 1500 · 141 / (16 · 35.6² / 6) = 62.6 MPa (0.25 %); a 41 815-element tet10 solution of the
|
||
full part gives 62.3 MPa. The pedal bore moves 0.75 mm down (tet10: 0.70–0.75 mm; Euler–Bernoulli
|
||
with a rigid clamp: 0.58 mm, without shear or the boss's compliance), and the spindle carries the
|
||
whole 1500 N and 255 N·m.
|
||
|
||
## Cantilever (`cantilever`)
|
||
|
||
|
||
|
||
A steel beam of 160 hex8 elements, clamped at one end with a load on the other: a static step
|
||
and a frequency step. **Look at:** the bending stress, highest at the clamped root, and the
|
||
Modes step's mode shapes.
|
||
|
||
## Bar impact (`impact`)
|
||
|
||
|
||
|
||
The same beam flying onto a rigid wall: an explicit dynamic step of ten frames. **Look at:** the
|
||
stress wave running back along the bar (Results ▸ Frame) and the energy history in the Results
|
||
panel.
|
||
|
||
## Hex and tet parts (`parts`)
|
||
|
||
|
||
|
||
A hex8 block beside a tet4 cube: meshes to look at (element types, nodes, picking), nothing to
|
||
solve.
|
||
|
||
## I-beam cantilever (`ibeam`)
|
||
|
||
|
||
|
||
The 200 mm steel I-beam of the README video (`tests/corpus/cantilever-i-beam`), opened as
|
||
SolidCraft geometry with persistent face names, swept all-hex (hex8 of 4 mm, incompatible modes),
|
||
clamped at `F2:start` and loaded with 1000 N down through a kinematic coupling to `F2:top`.
|
||
**Look at:** the bending stress in the flanges, highest at the clamp; the tip moves 0.18 mm down
|
||
(beam theory with shear: 0.184 mm).
|
||
|
||
## Heated flow past a bracket (`heated-flow`)
|
||
|
||
Air at 20 °C enters at 1 mm/s and passes the original upright bracket, whose faces are
|
||
held at 80 °C. The 130 × 80 × 70 mm domain has 5 mm cells (5,824 cells), a pressure outlet,
|
||
slip sides and top, and a no-slip floor. There are 180 steps and six saved output intervals,
|
||
plus the initial frame. Open `file.sample {"name":"heated-flow"}` and solve on CPU or GPU.
|
||
The slow prescribed flow keeps this coarse lattice away from its small-relaxation-time limit.
|
||
|
||
**Look at:** Results → Temperature, or `results.slice {"field":"temperature"}`, for the
|
||
warming air. `results.surface {"field":"wall_heat_flux"}` shows heat received by each
|
||
bracket face per area: the hot wall heating the air is negative. In this model's mm–tonne–second
|
||
units the wall flux is mW/mm². The initial frame has no recorded wall heat and stays uncoloured.
|
||
|
||
**Check:** the fluid warms above its inlet temperature while staying below the held wall
|
||
temperature (with a small numerical tolerance); recorded hot walls give heat to the fluid.
|
||
This coarse, short transient is a quick display demonstration, not a converged convection
|
||
benchmark. It has neither buoyancy nor solid heat conduction. The sample is isolated in
|
||
`crates/engine/src/sample/heated_flow.rs` and its gallery registration.
|