Trained Random Forests Machine Learning Models for Longitudinal Seismic Response Prediction of Highway Bridges

This page serves as accessible data materials for a Technical Paper submitted to Structural Safety.

Wang, X., Yang, D., Ye, A. (202X) "Machine learning-aided deterministic, partially probabilistic, and fully probabilistic seismic resilience assessment methods for highway bridges." (Submitted)

Scope Description:

It contains a dataset for training machine learning models for longitudinal seismic response predictions of highway bridge portfolios in China, together with the trained random forests-based machine learning models applied to predict the responses based on a user-input new dataset. 

The scope of highway bridges focuses on seismically designed multi-span continuous RC highway bridges, acting as the most widespread type of bridge in the transportation network in China. Columns are designed as flexure-failure dominated. This excludes bridges prone to shear or flexural-shear failures, ensuring the study remains aligned with modern seismic design principles. Also, soft soil sites are excluded to focus on relatively firm ground. 

Input Variables [Name, Notation (Unit)]: 

Column diameter, D (m)

Axial load ratio, Ra

Column height, H (m)

Rebar reinforcement ratio, ρl

Number of spans, N

Peak ground acceleration, PGA (g)

Peak ground velocity, PGV (cm/s)

Peak ground displacement, PGD (cm)

Spectral acceleration at 1.0s, Sa,1.0 (g)

Spectral velocity at 1.0s, Sv,1.0 (cm/s)

Housner intensity,  HI (cm)

Average spectral acceleration, AvgSa (g)

Output Variables [Name, Notation (Unit)]: 

Peak column drift ratio, γp

Residual column drift ratio, γr

Peak bearing displacement, δp (m)

Residual bearing displacement, δr (m)

Peak expansion joint displacement, Δp (m)

Suggested Platform Versions:

Python version: 3.8

H2O version: h2o-3.46.0.6