Gravity-only versus seismic-aware form-finding of reinforced concrete shells: 168 sized designs, their A1-A3 embodied carbon, and an offline explorer.
Explorer | Pre-registration | Numbers and sources | Reproduce
How much embodied carbon does seismic resilience cost in a form-found concrete shell, and does form-finding that includes seismic loads deliver it more cheaply than thickening a gravity-only shape?
ShellForm extends the seismic form-finding of Michiels & Adriaenssens (2017, 2018) by sizing each form to pass code checks and counting its A1-A3 embodied carbon. It is built on the validated, tagged modules of ShellBench and never copies a ShellBench value (data contract).
In 79 of the 84 paired cells the two strategies differ in carbon by less than 5 %, which the pre-registered rule (section 7) counts as no meaningful difference.
The other 5 cells (10 m, corner supports, rise 0.15, Sa 0.1-0.5 g) are one 5 mm thickness step next to the buckling limit, and they disappear under every robustness check. Thickness is set by linear buckling (corner supports) or the 50 mm minimum (edge supports), and steel by the minimum ratio; none of these responds to Sa up to 0.6 g. Every number and its source: paper/numbers.md.
The record is code tag freeze-2026-11-09 with ShellBench v0.3.0. The archive is on Zenodo, all versions: 10.5281/zenodo.24084110.
Each shell is form-found as a pin-jointed network in equilibrium (force density method, Schek 1974):
Steps:
- Plan, supports and rise define the network.
- Force density form-finding gives the gravity-only form (Sa = 0) and the seismic-aware form (midrange of the four thrust surfaces).
- CalculiX sizes each form: λ1 ≥ 5, compression and L/300.
- Steel and A1-A3 carbon are counted for each sized design.
- The 5 % rule is applied to the 84 paired cells.
Design space. 2 spans (10, 20 m) × 2 supports (4 corners, continuous edges) × 3 rise-to-span ratios (0.10, 0.15, 0.20) × 2 strategies (gravity-only, seismic-aware) × 7 plateau spectral accelerations (0-0.6 g) = 168 sized designs.
All figures are drawn from results/ by script, at full page width, as 600 dpi PNG and vector PDF (figures/README.md; visual system in docs/visual_style.md).
| Fig. 2: what sets the thickness | Fig. 4: what moves carbon |
|---|---|
![]() |
![]() |
Open the live explorer, or explorer/index.html straight from disk; it needs no server and no internet (three.js and the Inter typeface are bundled in explorer/vendor/). Pick a span, support case, rise and Sa level to see both strategies side by side: the form in 3D coloured by how far the seismic-aware form moves from the gravity-only one, carbon and thickness for each, carbon against Sa, and a clickable matrix of all 84 comparisons. Light and dark themes, keyboard control, a shareable link for every view, and a CSV of all 168 designs.
To try your own plan and supports, ShellLab runs ShellForm's force-density solver in the browser for any rectangular plan and support layout; its tests check it against ShellForm's own output.
Five checks beside the frozen record; no frozen setting or result changed (notes/day17_verification.md, Fig. S1).
| Check | What it shows | Result |
|---|---|---|
| V1 | Complete sphere under pressure, in the same CalculiX path that sizes every shell | Buckles within 2.0 % (R/t 50) and 1.3 % (R/t 100) of the classical value |
| V2 | Mesh convergence | λ1 changes by at most 0.6 % between 61 and 81 nodes per side |
| V3 | Force density equilibrium | Residual below 2.5e-11 N on all 84 form-finding solves |
| V4 | Shell action | 76-100 % of the strain energy is membrane energy |
| V5 | Formwork | No formwork factor or reuse count can move a verdict across the 5 % rule |
The test suite (tests/) covers form-finding, sizing, carbon, the paper tables and the visual system, and runs on every push.
You need Python 3.12 and, for the finite element tests and runs, CalculiX (ccx) on the path (Ubuntu: sudo apt install calculix-ccx). Without ccx the tests that call it are skipped. ShellBench is installed from a clone at the tag pinned in src/shellform/pin.py.
git clone --branch v0.3.0 https://github.com/syednafizreja/ShellBench.git ShellBench-pin
git clone https://github.com/syednafizreja/ShellForm.git
cd ShellForm
python -m venv .venv && source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt
pip install -e ../ShellBench-pin --no-deps
pip install -e . --no-deps
python -m pytest # 125 tests, about 1 minReproduce the study:
python -m shellform.batch # 8 designs (preset smoke8), results/batch_smoke8.csv
python -m shellform.batch --preset all # all 168 designs
python -m shellform.batch --span 10 --supports corners --sa 0.0 0.6 # any subset
python -m shellform.batch --compare-record # also check against the committed results
python scripts/day12_freeze.py # the whole freeze run (about 2 h on 4 workers)
python scripts/day11_figures.py # paper figures (Figs. 1-5)
python scripts/day17_figures.py # Fig. 6 and Fig. S1
python scripts/day17_verification.py # verification checks V1-V5 (about 6 min on 4 workers)
python scripts/day14_paper_tables.py # paper tables in paper/tables/
python scripts/day18_manuscript_tables.py # the same tables as LaTeX, paper/tables/latex/
python scripts/day13_explorer_data.py # explorer/explorer_data.js from results/batch_all.csvEvery results row names the ShellBench tag it used.
Long-term reproduction. Exact versions are in requirements.txt and environment-lock.yml, checksums of every result and figure in SHA256SUMS, and docs/REPRODUCE.md says how to verify the record without running anything and what to do when a dependency no longer installs.
ShellForm/
├── src/shellform/ form-finding, strategies, sizing, steel, carbon, batch runner, figures
├── scripts/ one script per stage of the study (freeze run, figures, tables, checks)
├── tests/ pytest suite, run in CI on every push
├── results/ the frozen record: every CSV names its ShellBench tag
├── figures/ paper figures, PNG and PDF, with the data behind each
├── explorer/ offline interactive explorer (three.js, no server needed)
├── paper/ paper tables (Markdown, CSV, LaTeX) and the numbers-to-sources map
├── notes/ dated working notes, one per step of the study
├── docs/ pre-registration, data contract, reproduction guide, changelog
├── requirements.txt pinned Python packages (pip, any platform)
├── environment-lock.yml exact conda environment of the record (Windows)
└── SHA256SUMS checksums of every file in results/ and figures/
The study design, the decision rule and the robustness checks were fixed in docs/preregistration.md on 30 Oct 2026, before any design-space run. Every later change is a dated amendment that says why. The history is in docs/CHANGELOG.md and the dated notes in notes/. Rules for changes are in CONTRIBUTING.md.
If you use ShellForm, please cite the archived software (the concept DOI always resolves to the latest version):
@software{hasib_shellform,
author = {Hasib, Syed Nafiz Reja},
title = {{ShellForm}: the carbon cost of seismic resilience in form-found concrete shells},
publisher = {Zenodo},
doi = {10.5281/zenodo.24084110},
url = {https://github.com/syednafizreja/ShellForm}
}Machine-readable metadata: CITATION.cff.
Code: MIT (LICENSE). Bundled third-party files keep their own licences: three.js r128 (MIT, explorer/vendor/three.LICENSE) and the Inter typeface (SIL Open Font License 1.1, assets/fonts/OFL.txt).



