Post-Earthquake Forensic Pathology
The Puebla 2017 Seismic Event
Bayesian updating of the collapse probability of reinforced concrete frames with masonry infill, computed with nonlinear models and corrected with the forensic surveys made after the 2017 Puebla earthquake.
Research Overview
Subject. Mid-rise reinforced concrete buildings with unreinforced masonry infill walls affected by the 2017 Puebla earthquake in Mexico (M 7.1).
Purpose. Collapse probability with quantified uncertainty, as input for urban seismic risk management. Calculation in Python with OpenSeesPy.
Data. Collapse probabilities from the numerical model combined with the forensic surveys of 62 mid-rise buildings. Manuscript in preparation for external review.
Preliminary Results
- Prior collapse probability: 46.9% (high epistemic uncertainty)
- Posterior collapse probability: 32.9% (updated with field evidence)
- Critical displacement threshold: 11 cm roof drift
- Brittle failure mechanism in the masonry infills
- Progressive collapse initiated by shear transfer to the columns
Methodology
The infills were modeled in OpenSeesPy with the Pinching4 material to represent the hysteretic degradation of unreinforced masonry. The numerical model was subjected to 73 ground motion records from the 2017 Puebla event.
The Bayesian model uses a Beta-Binomial conjugate pair: the prior distribution of the collapse probability is updated with the observed collapse data, and conjugacy gives the posterior in closed form, which keeps the computation inexpensive.
Application
- Diagnostics: assessment of existing buildings with a model updated against field evidence from a documented earthquake
- Retrofitting: design of interventions with a cost criterion
- Risk management: collapse probability estimates supported by field evidence
Open Data
Structural models, code and seismic records of the study, published in Elsevier's Mendeley Data repository.
DOI: 10.17632/b8szvwjp5k.1Project Details
- Type
- Ongoing research
- Status
- Manuscript in preparation, not submitted
- Focus Area
- Seismic Vulnerability
- Methods
- Bayesian Inference