Dataset for Diffusion Map Kernel Regression Model of Pitching and Plunging Flat Plate

This dataset contains pitching–plunging flat-plate flow simulations generated using the Immersed Boundary Projection Method (IBPM). It supports the construction, training, validation, and benchmarking of kernel ridge regression (KRR) models using Diffusion Map (DM) and Radial Basis Function (RBF) kernels, as well as the Residual Dynamic Mode Decomposition (ResDMD) model. The paper can be found here: https://arxiv.org/abs/2512.17203.

The dynamics of the system are governed by two key parameters: the pitching phase and the plunging phase. In this dataset, the pitching phase is fixed at 0°, while the plunging phase varies across 16 trajectories, spanning phases from 0° to 330°.

All data are stored as Python pickle files, with each file representing a single trajectory. The filename convention follows:

• PhaseA_X → pitching phase (fixed at 0)
• PhaseH_Y → plunging phase (varies across trajectories)

A total of 16 trajectories are included. The trajectories with plunging phases except 135° are used for model training and validation, while the trajectory at 135° plunging phase is reserved exclusively for out-of-sample testing. Each training/validation trajectory consists of 200 x 299 x 599 where 200 is the number of time steps, and 299 and 599 are the vertical and horizontal spatial grid points. The test trajectory has the same spatial grid but 801 time steps. All the trajectories are generated with dt=0.01 second and downsampled at dt=0.1 second.