A team of CFSRC researchers is leading an effort to collect and curate cold-formed steel physical test data that can be used to construct data-driven models. The initiative’s mission is to make high quality physical test data accessible to anyone. The team is affectionately called ‘The Turtles’ because we are making progress at a slow and steady pace.
The data is organized by type (e.g., connections, beams, columns) using custom-built schema and stored in a machine-readable format in the CFSRC GitHub monorepository. The data can be accessed with a few lines of code in your favorite scientific computing language and then used to train machine learning models. Our team’s current focus is collecting data and coding core tools needed for data-driven modeling.
The first large data set coming from this initiative is fastener connection load-deformation response. The database holds 550 monotonic and cyclic load-deformation response curves from tests on screw-fastened and power-actuated fastened ply-to-ply connections. The data set is being used by Dr. Hyeyoung Koh in a Steel Deck Institute (SDI) research project focused on developing a computational alternative to floor diaphragm physical testing (shear strength characterization) that models fastener-level load-deformation response.
Other data sets are also becoming available, including cold-formed steel member load-deformation response (Damir Akchurin), steel stress-strain curves (Dr. Chu Ding), shear wall strength and stiffness (Dr. Deniz Ayhan Moen, Dr. Zhidong Zhang), shear connectors (Dr. Hyeyoung Koh), and data contributions from international researchers, for example Dr. Yuanqi Li at Tongji University.
We are always looking for data and collaborators. Message us on the CFSRC #turtles [Slack channel].