Seismic Fragility of RC Bridge Columns: 300-Sample Pushover/Dynamic Datasets and Trained ANN Surrogates (MATLAB)
Description
Overview: This dataset supports a series-distributed artificial neural network (ANN) framework for developing seismic fragility curves of reinforced-concrete (RC) bridge columns with reduced computational cost. It contains four curated datasets (300 samples each) generated from nonlinear pushover and time-history dynamic analyses for rectangular and circular sections, six pre-trained ANN surrogate models, and MATLAB scripts to perform Monte Carlo–based fragility assessment. Contents: Datasets (Excel, 300 samples each) 1_dataset_Col_push_300.xlsx — Pushover analysis results for rectangular columns. 2_dataset_Circular_Col_push_300.xlsx — Pushover analysis results for circular columns. 5_dataset_Col_dyna_300.xlsx — Time-history dynamic analysis results for rectangular columns. 6_dataset_Circular_Col_dyna_300.xlsx — Time-history dynamic analysis results for circular columns. Trained ANN models (MATLAB .mat) 1_trained_2lopan_cfnn_push_1.mat — model for rectangular/pushover. 2_trained_2lopan_cfnn_push_1.mat — model for circular/pushover. 5_trained_2lopan_cfnn_dyna_1.mat — model for rectangular/dynamic. 6_trained_2lopan_cfnn_dyna_1.mat — model for circular/dynamic. 5_trained_2lopan_cfnn_IM_1.mat — intensity-measure relation model (rectangular set). 6_trained_2lopan_cfnn_IM_1.mat — intensity-measure relation model (circular set). Code (MATLAB .m) mc_based_fragility.m (main driver), fragility.m (helper), and utilities for parameter fitting (fn_mle_pc.m, fn_mle_pc_probit.m, fn_sse_pc.m). Methods in brief: Ground-motion–intensity relations, drift demands, and damage-state thresholds are learned via coupled ANNs trained on the provided analysis results. The trained surrogates are then integrated with Monte Carlo simulation to produce fragility functions across multiple limit states without assuming a lognormal form.
Files
Steps to reproduce
How to use: Open mc_based_fragility.m in MATLAB (R2020a+ recommended; Statistics and Machine Learning Toolbox). In the section “Define 7 input structural parameters”, set the parameter values/distributions for your case. Load the appropriate trained ANN(s) according to section type (rectangular/circular) and analysis type (pushover/dynamic). Run for each desired limit state to obtain fragility curves and fitted parameters. File formats: .xlsx (datasets), .mat (trained networks), .m (scripts). Reuse notes: Keep dataset–model pairing consistent (e.g., rectangular/dynamic → 5_trained_…_dyna_1.mat). The scripts can be adapted to new parameter ranges or alternative limit-state definitions.