IDA Database of Peruvian RC Bridge Piers under Subduction Motions

## Dataset **Base_Datos_IDA_Puente.csv** — 43,776 nonlinear time-history analyses (NLTHA) - 192 pier geometries × 19 Peruvian subduction records × 12 PGA levels (0.1–2.3 g) - 25 columns: D_Col_m, H_Centro_m, H_Extremo_m, Sismo_ID, PGA_Original_g, Factor_Escala, PGA_g, Periodo_Fund_s, Rigidez_Efectiva_kN_m, PSa_Original_g, PSa_g, Def_Max_Pila_mm, Def_Max_Estribo_mm, Drift_Pilas_%, Drift_Estribos_%, Def_Neopreno_Centro_mm, Def_Neopreno_Extremo_mm, Cortante_Base_Pila_kN, Cortante_Base_Estribo_kN, Cortante_Neoprenos_Centro_kN, Cortante_Neoprenos_Extremo_kN, Momento_Base_Pila_kNm, Momento_Base_Estribo_kNm, Deform_Corte_Neo_Centro_%, Deform_Corte_Neo_Extremo_% - Source: OpenSeesPy, force-based fibre beam-column elements (Concrete01/Steel02) ## Input files (for re-running the IDA pipeline) | File | Purpose | |------|---------| | `Sismos_Historicos_Peru.xlsx` | Metadata of the 19 Peruvian subduction records | | `sismo_real_7032.txt` … `sismo_real_7060.txt` | 19 accelerogram files (m/s², dt=0.02 s) | ## Scripts | Script | Purpose | |--------|---------| | **`ProcesoIDA_v9a.py`** | OpenSeesPy pipeline — generates the IDA database from scratch (requires OpenSeesPy + 3+ days of computation) | | `utils_sismo.py` | Shared utilities: Arias intensity, Housner intensity, PSA, limit-state classification | | `limpiado-csv.py` | Post-processing and cleaning of raw CSV output | | **`regen_all_figs_EN.py`** | Regenerates all 18 figures (300 dpi, English labels) from the CSV | | **`regen_figs_300.py`** | Alternative figure generation (300 dpi) | | **`make_canonical.py`** | Computes canonical numbers (drift mean, correlations, fragility, ML metrics) | | `make_docx_final.py` | Generates .docx manuscript | | `consolida_v2.py` | Data consolidation and cross-validation | | `insert_ai_decl.py` | Inserts AI declaration into .tex | ## Canonical data - `canonical_numbers_v2.json` — All reported numbers (drift statistics, Pearson correlations, power-law exponents, fragility parameters, ML R², ablation) - `metricas_classif_groupkfold.json` — Leakage-free classification metrics (accuracy, per-class F1) - `resultados_reales.json` — Data validation summary ## Quick start (reproduce figures and numbers) ```python python regen_all_figs_EN.py # regenerates 18 figures in Figures/ python make_canonical.py # computes canonical numbers from CSV ``` Both read `Base_Datos_IDA_Puente.csv` from the same directory. No other dependencies needed beyond pandas, numpy, scipy, scikit-learn, matplotlib. ## Re-running the full IDA (optional) ```bash python ProcesoIDA_v9a.py ``` Requires: OpenSeesPy, pandas, numpy. Computation time: ~3–5 days on a multi-core workstation. The accelerogram files (sismo_real_*.txt) and Sismos_Historicos_Peru.xlsx must be in the same directory. ## Associated manuscript - **Journal:** Engineering Structures (Elsevier) - **Title:** *How pier geometry governs seismic drift in continuous RC bridges: design relations, geometry-resolved fragility, and a leakage-free ML surrogate from 43,776 nonlinear analyses with Peruvian subduction records* - **Authors:** Yordan A. Rocio Maldonado, Jose C. Masias Guillen ## Reproducibility Every figure, table, and coefficient in the manuscript is reproducible from `Base_Datos_IDA_Puente.csv` using the provided Python scripts. The canonical numbers in `canonical_numbers_v2.json` match the values reported in the paper.