Liquid-Metal Magnetohydrodynamic Lid-Driven Cavity Flow Dataset

This dataset contains two-dimensional liquid-metal magnetohydrodynamic (MHD) lid-driven cavity flow simulations in NumPy format. The data were used as the MHD fine-tuning dataset in the MATEY study, MATEY: multiscale adaptive foundation models for spatiotemporal physical systems The dataset consists of solution trajectories from initial conditions toward steady states for a benchmark lid-driven cavity MHD problem. The governing system is an incompressible Navier–Stokes equation with Lorentz force coupled to an induction equation with divergence cleaning, following the lid-driven MHD cavity formulation described by Fambri et al. The MHD solver is published at Endeve et al. The dataset includes 30 configurations, formed from three magnetic Reynolds numbers, Re_m=100,200,500, and ten imposed horizontal magnetic-field magnitudes, Bx=0.05,0.10,…,0.50. For each configuration, the fluid velocity field and induced magnetic field are stored on a 128 × 128 uniform spatial mesh for 2,000 time steps. The four stored channels are the two velocity components and the two induced magnetic-field components: velocity_0, velocity_1, induced_magnetic_field_0, and induced_magnetic_field_1. The NumPy arrays are organized as time-dependent gridded fields with channel ordering: (time, y, x, channel) where channel = 0, 1, 2, 3 corresponds to: 0: velocity_01: velocity_12: induced_magnetic_field_03: induced_magnetic_field_1 It's used as the MHDDataset in MATEY, where the data loader converts samples to the model-facing format (T, C, H, W) and, for 3D-compatible batching, expands the 2D field to (T, C, D, H, W) with D = 1 in history-to-future prediction tasks with a configurable history length and lead time. File names follow the convention used by the MATEY loader: numpy_Bx_<Bx>_Re_<Re>.npy, for example: numpy_Bx_0.05_Re_100.npynumpy_Bx_0.50_Re_500.npy A typical loading pattern is: import numpy as nparr = np.load("numpy_Bx_0.05_Re_100.npy")print(arr.shape) # expected: (time, y, x, channel)