LLM agent for structural steel design
A prototype that generates framing layouts from plain-language descriptions, checks column strength with a physics-informed neural network, and writes and runs OpenSeesPy frame analyses.
View on GitHub →Open data for structural engineering
Engineer
with data.
We collect, curate and openly share structural test data so engineers can build better, more reliable design models.
Fcr/Fy versus slenderness λc. The column curve rests on theory from 1870–1920 and tests run before 1970.
Mission
Support data-driven engineering analysis and design.
Our members call themselves the Turtles. Slow and steady, one curated dataset at a time.
When we design a steel-framed building, we lean on semi-empirical equations like the column curve in AISC 360. Those equations trace back to models by Perry, Ayrton and Robertson and to physical tests run up to 1970.
Since then we have accumulated far more knowledge, measurements and simulation results. Most of it is scattered across technical reports and theses, and much of it is getting lost.
What would happen if we carefully curated historical and recent test data and offered it up with an easy, open interface?
Other fields show the payoff. The Worldwide Protein Data Bank has curated protein structures since 1971, and that archive made AlphaFold possible. Steel test data is sparse by comparison, so we couple data with theory and existing models to get the most from every test.
EngineerWithData is a volunteer working group of practitioners, researchers and students. We build open databases, publish them on GitHub, and develop the data-driven and AI models that use them.
- 01 · Intake
Gather
Find test programs in papers, reports and lab archives, and accept contributions from anyone.
- 02 · Curate
Standardize
Convert each dataset to a shared schema with units, geometry, material properties and citations.
- 03 · Host
Publish
Release the data openly on GitHub in CSV and JSON, with viewers and notebooks.
- 04 · Use
Model
Calibrate design equations, train ML and physics-informed models, and inform specification committees.
Open databases
Test data you can download today.
| Database | Format | Records |
|---|---|---|
| Fastener connection testsScrew and power-actuated fastener connections to cold-formed steel, with full monotonic and cyclic load-deformation curves. | CSV | 612tests |
| Hollow section column bucklingSquare and rectangular hollow section columns, hot-rolled and cold-formed. | CSV | 695tests + 3,974 FE |
| Cold-formed steel column testsColumn tests on cold-formed steel members from the literature. | CSV | 600+tests |
| Shear connector push-out testsHeaded studs in composite steel deck, with an interactive data viewer. | JSON | 551tests |
| Cold-formed steel couponsTensile stress-strain curves, including advanced high-strength steels. Archived on Zenodo. | JSON · MAT | 400+curves |
| Cold-formed steel imperfectionsMeasured bow, camber, twist, local and distortional imperfections along member length. | CSV | 53specimens |
Every dataset keeps its original files, its source citation and a README describing the test program. Have data to add? Tell us about it.
Models and initiatives
Putting the data to work.
Fastener backbone prediction
A neural network trained on the fastener database that predicts load-deformation backbones for cold-formed steel connections.
View on GitHub →Instantaneous center method for diaphragms
Notebooks for inelastic shear strength of steel deck diaphragms with mixed fastener types, built on measured fastener behavior.
View on GitHub →Building databases with agents
AI agents that extract test data from published papers, with confidence scoring and manual checks so the curated values can be trusted.
Code compliance database
Design codes stored as data: standardized variable names, BIM property mapping and equation logic in JSON, so any software can run the same transparent checks for AISC, ACI, ASCE or Eurocode.
A data-first steel industry
We advocate for an industry-led steel data advisory council to plan data intake, curation and open access, starting with the data behind AISC specification updates.
Talks and papers
Where we've presented.
- NASCC 2026Designing with Data · Moen, Zhang
- NASCC 2026Fasteners and data-driven models · Moen
- SSRC 2026LLM agents for structural steel design · Zhang et al.
- NASCC 2025Engineering with Data · Moen, Arber, Zhang, Çevik
- NASCC 2024Database Driven Design Methods · Blum, Moen
- CFSEI 2023Cold-Formed Steel Framing Design with Data-Driven Models
Get involved
You can EngineerWithData too.
- We meet regularly to share data, models and tools.
- We help with data formats and structures, so contributing takes minimal work.
- Engineers, researchers, students and manufacturers are all welcome.