Shell crushing / Adaptive mesh refinement
Adaptive tube crushing
A moving planar rigid wall crushes a thin-walled tube along its axis. Watch the original mesh and an adaptive mesh evolve side by side as the shell folds.
Left: original mesh. Right: adaptive mesh. Both panels show the computed geometry with shell mesh edges and a translucent planar wall. Color shows shell-top effective plastic strain on a fixed 0–0.50 scale; higher values use the top color. The camera stays fixed throughout the sequence.
The original mesh remains at 1,868 elements. The adaptive run adds elements as the tube deforms. The two panels show how the computed deformation and mesh evolve in these configurations.
Model and loading
Thin-walled shell structure
A tube with an initial shell thickness of 2 mm uses Quad4 · fully integrated assumed strain shells, with 2 × 2 in-plane and three through-thickness integration points. The model uses a tabulated elasto-plastic material.
Axial impact and folding
An 800 kg planar rigid wall has an initial axial speed of 8.94 m/s. The tube is fixed at its base in translation and rotation. The model includes wall contact and shell self-contact.
What the two panels compare
| Configuration | Original mesh | Adaptive mesh |
|---|---|---|
| Initial shell elements | 1,868 | 1,868 |
| Final shell elements | 1,868 | 19,970 |
| Saved result states | 21 | 21 |
| Simulation interval shown | 0–20 ms | 0–20 ms |
| Display scale | 0–0.50 plastic strain | 0–0.50 plastic strain |
The two sequences are paired by simulated time and use the same camera and color scale. Wall displacement and folding patterns can differ between the runs. The 5.25-second video shows 20 ms of simulated motion; video length is not solver runtime.
Mean wall resistance: analytical context
The force histories provide a quantitative view alongside the animation. The values below use the same 0–160 mm of rigid-wall travel in both runs. Mean wall resistance is wall work divided by travel: P̄ = ∫ F(s) ds / Δs, using the saved wall-force history. This interval includes an initial geometric gap of about 1.52 mm, so wall travel is not identical to tube-crushing stroke.
| Configuration | Mean wall resistance 0–160 mm travel |
|---|---|
| Adaptive mesh | 62.9 kN |
| Original mesh | 74.2 kN |
Idealized scale estimate
The rigid-plastic folding model of Wierzbicki and Abramowicz (1983) gives P̄ ≈ 9.56 σ₀ t5/3 C1/3. An idealized 116 × 96 mm section gives C = 106 mm. With t = 2 mm and an assumed flow stress σ₀ = 366 MPa, the estimate is 52.6 kN.
Here, 366 MPa is the initial yield stress used as a simple proxy; the material hardens, so it is not a measured mean flow stress. The actual section also has rounded corners and local protrusions, and the simulation is an impact problem. These differences limit the comparison.
This estimate provides context for the force scale. It is not a validated acceptance band, an error metric, or proof that either mesh is converged. An effective-crushing-distance correction would require checking the fold geometry and crushing distance for this model; it is not applied here.
Scope of this benchmark
This is an experimental adaptive-refinement demonstration from a development build. It shows a configured combination of shell plasticity, a moving planar rigid wall, self-contact and mesh refinement. It does not establish that the adaptive result is more accurate, mesh-converged or faster than the original-mesh result.
The analytical estimate above is an idealized comparison, without a calibrated reference curve or accuracy tolerance for this geometry and material. Separate numerical comparisons are available in Accuracy; general capability scope is documented in Evidence.
EXPERIMENTAL · Configured adaptive shell demonstration. The original and adaptive meshes have different resolutions; this is not a convergence or performance comparison.