Exciting news! We're transitioning to the Statewide California Earthquake Center. Our new website is under construction, but we'll continue using this website for SCEC business in the meantime. We're also archiving the Southern Center site to preserve its rich history. A new and improved platform is coming soon!

Machine Learning Based Regional Seismic Retrofit Design Optimization for Soft Weak Open Front Wall Line Buildings

Zhengxiang Yi, & Henry Burton

Published August 14, 2019, SCEC Contribution #9612, 2019 SCEC Annual Meeting Poster #293

Policies are often enacted to mandate the retrofit of seismically vulnerable buildings. A major challenge in specifying the design requirements for policy-based retrofits is ensuring that the desired performance enhancements are achieved across a portfolio of buildings that differ significantly in terms of structural characteristics and seismic hazard. To address this challenge, a machine-learning-based optimization framework is developed. The proposed methodology seeks to inform the development of prescriptive retrofit measures that maximize seismic performance enhancements at the portfolio scale. The initial framework is developed specifically for woodframe buildings with soft, weak and open front (SWOF) wall lines (or soft-story buildings). There are three key elements to the overall methodology. First, machine-learning-based surrogate models are developed as compact statistical links between building structural characteristics (e.g. number of stories, story strengths, building configuration) and nonlinear response history analysis (including collapse simulation) and performance assessment (e.g. collapse, demolition, repair costs) outcomes. Second, objective functions are defined at the regional or portfolio scale (e.g. maximizing the total increase in collapse margin ratio for all buildings, minimizing total losses for a scenario earthquake) with appropriate constraints (e.g. minimum increase in collapse margin ratio for any single building, maximum cost of retrofit). Lastly, gradient-based stochastic optimization is implemented to determine the retrofit enhancements (e.g. increase in strength and ductility) that would achieve the most desirable combined outcome for all buildings in the portfolio. The framework is demonstrated using the inventory of approximately 12,000 SWOF buildings that are under the purview of the Los Angeles Soft-Story Ordinance.

Citation
Yi, Z., & Burton, H. (2019, 08). Machine Learning Based Regional Seismic Retrofit Design Optimization for Soft Weak Open Front Wall Line Buildings. Poster Presentation at 2019 SCEC Annual Meeting.


Related Projects & Working Groups
Earthquake Engineering Implementation Interface (EEII)