Data from: A transfer learning-based hybrid surrogate modeling framework for efficient multi-objective seismic design of long-span cable-stayed bridges

This dataset contains the complete set of computational models, source
code, and reference data supporting the research presented in the
associated article titled “A transfer learning-based hybrid surrogate
modeling framework for efficient multi-objective seismic design of
long-span cable-stayed bridges.” It is organized into five main
components. First, it includes finite element models consisting of SAP2000
models (version V10) for two long-span cable-stayed bridges, which serve
as the high-fidelity simulation basis. Second, it provides neural network
surrogate models with MATLAB (R2024b) source code used to construct and
train three types of surrogate models: Backpropagation Neural Network
(BPNN), Radial Basis Function Network (RBFN), and Generalized Regression
Neural Network (GRNN); a pre-trained RBFN model for Bridge M1 is also
included. Third, the dataset contains a transfer learning module
implemented in MATLAB, which enables adaptation of the surrogate model
trained for Bridge M1 to Bridge M2. Fourth, it includes a hybrid
optimization framework in MATLAB for conducting multi-objective seismic
design optimization while integrating the surrogate models. Fifth, the
dataset provides parameter sampling data with MATLAB scripts for
performing Latin Hypercube Sampling on Fluid Viscous Damper (FVD)
parameters. All code is executable with clearly defined dependencies, and
the included PEER ground motion information table supports replication of
the seismic input data. Overall, this dataset enables full reproduction of
the study’s numerical experiments and offers a reusable computational
framework for researchers and engineers working on efficient seismic
performance assessment and design optimization of long-span bridges.