High-Dimensional Dataset and Gaussian Process Models for Forced and Free Vibration Experiments

This repository provides a high-dimensional experimental dataset and trained Gaussian Process Regression (GPR) surrogate models for fluid–structure interaction studies of a rigid cylinder under both free vibration and forced vibration conditions. The dataset includes multi-dimensional hydrodynamic response data obtained from large-scale towing-tank experiments. Two main data categories are provided: (1) A four-dimensional dataset for self-excited vibration experiments, covering variations in reduced velocity, structural mass ratio, damping ratio, and Reynolds number. (2) A five-dimensional dataset for forced vibration experiments, including excitation amplitudes, excitation frequencies, phase difference, and flow parameters. Pre-trained GPR surrogate models (*.mat) are provided for both forced and self-excited cases, enabling fast prediction of hydrodynamic response quantities without repeating physical experiments. These models were trained using experimentally measured data and validated within the investigated parameter ranges. In addition, MATLAB scripts (*.m) are included to reproduce prediction results and visualization figures based on the trained models, allowing users to generate response surfaces, parametric maps, and high-dimensional projections of the experimental database. This dataset and model collection is intended to support reproducible research, surrogate modeling, and data-driven analysis of vortex-induced vibration and fluid–structure interaction phenomena, and can be directly used for benchmarking machine learning models and reduced-order modeling methods.