A Flow Field Reconstruction Based on Cylindrical Wake Flow Data Using the PSO-CNN-LSTM Algorithm

This dataset contains time-resolved two-dimensional flow-field data for incompressible flow past a circular cylinder in a rectangular channel, generated by a numerical simulation based on a projection-type scheme (velocity prediction followed by pressure correction). The computational domain is a 2D channel of size 1.0 × 0.5 (streamwise × transverse), discretized using a uniform structured grid with 440 points in the streamwise direction and 82 points in the transverse direction. A circular cylinder is embedded in the channel with center at (0.2, 0.25) and radius 0.05.
The fluid is assumed Newtonian and incompressible. The inflow boundary condition is a uniform free-stream velocity of 1.0. No-slip conditions are applied on the upper and lower channel walls as well as on the cylinder surface. At the outlet, the pressure is set to a zero reference value (Dirichlet condition), while Neumann-type treatments are applied for pressure at the inlet and channel walls. The density is 1.0. The Reynolds number, based on the cylinder diameter and the inflow velocity, is 500, which corresponds to a kinematic viscosity of 0.0002.
Time integration uses a constant time step of 1.0 × 10⁻⁴ for 500,000 steps, resulting in a total physical time of 50.0. The flow field is saved every 1,000 steps (saving interval 0.1 in physical time). Therefore, the dataset includes 500 time snapshots spanning t = 0.1 to 50.0 (the initial time t = 0.0 is not stored). Each snapshot contains values at all grid nodes, i.e., 36,080 points (440 × 82).
The data are provided in MATLAB format as cylinder_flow_result.mat, containing a single variable named result_matrix_all. This variable is a two-dimensional matrix obtained by concatenating all saved snapshots in chronological order. Each row corresponds to one grid point at one time instant. The matrix has 7 columns with the following order and meaning:
t — time
x — x-coordinate
y — y-coordinate
u — x-velocity component
v — y-velocity component
s — speed magnitude, computed as √(u² + v²)
p — pressure (relative pressure with outlet set to zero reference)