Seismic Vulnerability Assessment of RC Buildings via Bayesian Updating: Evidence from the 2017 Puebla Earthquake
Description
OverviewThis dataset contains the high-fidelity structural models, processed seismic records, and statistical frameworks supporting the research article: "Seismic Vulnerability Assessment of RC Buildings via Bayesian Updating: Evidence from the 2017 Puebla Earthquake". The repository enables the full reproduction of a forensic engineering assessment of mid-rise RC buildings with unreinforced masonry (URM) infills, characteristic of Mexico City’s transition zone.Methodology & Refinements (Version 2.0)The research integrates numerical and empirical evidence through a robust Bayesian framework, updated in this version to meet Q1-tier reproducibility standards:Numerical Evidence: Non-linear static (pushover) and stochastic reliability analyses performed in OpenSeesPy. The URM infill behavior is modeled via an equivalent strut approach using the Pinching4 constitutive law, with damage parameters (stiffness/strength degradation) strictly anchored to the mode-II fracture energy ($G_f^{II}$) of Mexican artisanal clay bricks.Stochastic Framework: Epistemic uncertainty in masonry strength ($f'_{me}$) is quantified using a Lognormal probability distribution to ensure physical consistency (positive-definite domain) and eliminate algorithmic bias in the Monte Carlo simulations.Empirical Evidence: Forensic field data from 62 buildings (19 observed collapses) following the 19/09/2017 event.Synthesis: A Beta-Binomial Bayesian Updating framework was implemented to refine the analytical prior collapse probability ($P_f = 46.9\%$) into a posterior estimate ($P_f = 32.9\%$), consistent with observed seismic pathologies.Dataset Structure/scripts: Includes optimized Jupyter Notebooks (.ipynb) for the OpenSeesPy structural engine, Lognormal Monte Carlo sampling, and Bayesian inference algorithms./data: Contains the compressed ground motion suite (Acel_2694_047407.zip) consisting of 73 processed seismic records from the 2017 Puebla Earthquake (PGA max = 1.705 m/s²)./results: Raw data in CSV format, including Pushover capacity curves, fragility function coordinates (Prior vs. Posterior), and high-resolution figures for manuscript validation.Usage NotesThe Python environment requires openseespy, numpy, scipy, matplotlib, and pandas. For reproducibility, the ground motion suite must be unzipped into the working directory. The provided scripts include an autonomous file scanner to process the seismic ensemble.
Files
Steps to reproduce
To replicate the structural and probabilistic findings of this research, follow this sequential workflow:1. Software Environment SetupEnsure a Python 3.10+ environment is available (Google Colab is highly recommended for full compatibility). The following libraries must be installed:openseespy: For non-linear structural modeling.scipy.stats: Specifically for Lognormal fitting and Bayesian conjugate priors.numpy, pandas, & matplotlib: For tensor operations, data management, and high-resolution visualization.2. Seismic Data InitializationUpload and unzip the provided Acel_2694_047407.zip file into the working directory. This archive contains 73 processed acceleration time-histories from the 2017 Puebla Earthquake (Transition Zone, Mexico City).Note: The provided scripts include an autonomous file scanner (os.walk) to identify and process these records regardless of the internal directory structure.3. Non-Linear Structural Calibration (Pushover)Execute Seismic_Reliability_Assessment_RC_Infilled_Frames.ipynb. This script:Builds the 2D RC frame with equivalent URM struts.Calibrates the Pinching4 material damage parameters based on mode-II fracture energy ($G_f^{II}$).Performs a displacement-controlled analysis to determine the 11 cm drift collapse threshold.Output: Results_01_Pushover_Capacity_Curve.csv.4. Stochastic Fragility Analysis (Prior)Run Puebla19S_Reliability_Engine.ipynb. This notebook:Executes a Monte Carlo simulation (10,000 iterations) using a Lognormal PDF for masonry strength ($f'_{me}$).Integrates the seismic demand from the 73 records via the Newmark-beta method.Determines the analytical prior probability of collapse ($P_f = 46.9\%$).Output: Results_02_Prior_Fragility.csv.5. Bayesian Updating FrameworkExecute Inferencia_Bayesiana.ipynb. This script:Implements the Beta-Binomial conjugate model.Updates the analytical prior with empirical evidence from 62 inspected buildings (Arteta et al., 2018).Generates the refined posterior probability of collapse ($P_f = 32.9\%$).Output: Results_03_Bayesian_Updating_Densities.csv.6. ValidationVerify numerical consistency by comparing the generated CSV outputs and plots against the high-resolution figures provided in the /results folder of this repository.