Seismic Fragility of RC Bridge Columns: 300-Sample Pushover/Dynamic Datasets and Trained ANN Surrogates (MATLAB)

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.概述： 本数据集支持一种串联式人工神经网络（Artificial Neural Network，ANN）框架，用于构建钢筋混凝土（Reinforced Concrete，RC）桥柱的地震易损性曲线，可有效降低计算成本。数据集包含四个精心整理的数据集（每个含300个样本），这些数据集来自针对矩形和圆形截面的非线性推覆分析与时程动力分析；此外还包含六个预训练的人工神经网络替代模型，以及用于执行基于蒙特卡洛（Monte Carlo）的易损性评估的MATLAB脚本。 内容： 数据集（Excel格式，每个含300个样本） 1_dataset_Col_push_300.xlsx — 矩形柱的推覆分析结果。 2_dataset_Circular_Col_push_300.xlsx — 圆形柱的推覆分析结果。 5_dataset_Col_dyna_300.xlsx — 矩形柱的时程动力分析结果。 6_dataset_Circular_Col_dyna_300.xlsx — 圆形柱的时程动力分析结果。 预训练人工神经网络模型（MATLAB .mat格式） 1_trained_2lopan_cfnn_push_1.mat — 针对矩形柱/推覆分析的模型。 2_trained_2lopan_cfnn_push_1.mat — 针对圆形柱/推覆分析的模型。 5_trained_2lopan_cfnn_dyna_1.mat — 针对矩形柱/动力分析的模型。 6_trained_2lopan_cfnn_dyna_1.mat — 针对圆形柱/动力分析的模型。 5_trained_2lopan_cfnn_IM_1.mat — 强度量度关系模型（矩形柱数据集）。 6_trained_2lopan_cfnn_IM_1.mat — 强度量度关系模型（圆形柱数据集）。 代码（MATLAB .m格式） mc_based_fragility.m（主驱动脚本）、fragility.m（辅助脚本），以及用于参数拟合的工具函数fn_mle_pc.m、fn_mle_pc_probit.m、fn_sse_pc.m。 方法概述： 通过对提供的分析结果进行训练的耦合人工神经网络，可学习地震动强度关系、位移需求与损伤状态阈值。随后将训练好的替代模型与蒙特卡洛模拟相结合，能够生成多极限状态下的易损性函数，且无需假设其服从对数正态分布。