Dataset for the paper "Spatio-temporal Graph Convolutional Autoencoder for Transonic Wing Pressure Distribution Forecasting"

This repository provides the unsteady Computational Fluid Dynamics (CFD) datasets used in:G. Immordino, A. Vaiuso, A. Da Ronch, M. Righi."Spatio-temporal Graph Convolutional Autoencoder for Transonic Wing Pressure Distribution Forecasting."Aerospace Science and Technology, 147 (2025) 109780.https://doi.org/10.1016/j.ast.2025.109780 The dataset contains unsteady Reynolds-averaged Navier–Stokes (URANS) simulations of the Benchmark Supercritical Wing (BSCW) undergoing pitch and plunge excitations at Mach 0.74. These data were used to train and validate a spatio-temporal Graph Convolutional Autoencoder for forecasting wing surface pressure distributions. Test Case Configuration: Semi-span BSCW (rectangular planform, supercritical airfoil). Grid: Unstructured type, 86,840 surface points, y+≈1. Flow: M=0.74, Re=4.49×106, freestream angle of attack =0 deg. Motions: Pitch about 30% chord; plunge allowed. Solver: SU2 v7.5.1, URANS with Spalart–Allmaras model; JST scheme with artificial dissipation; Green–Gauss gradients; ILU-preconditioned BiCGStab solver. Timestep: 2×10−4 s; duration: 2 s. Dataset Structure The dataset comprises 12 unsteady simulations of the BSCW configuration, including damped Schroeder-phased harmonic excitations with varied reduced frequencies and amplitudes, undamped Schroeder signals, single degree-of-freedom motions, and a single-harmonic excitation. Contents Pressure coefficient (CP) distributions at 86,840 surface nodes. Motion inputs: pitch, plunge, and their first and second derivatives. PurposeThe datasets were designed for developing and benchmarking spatio-temporal graph neural networks for unsteady aerodynamics. They capture nonlinear transonic phenomena such as shock motion, shock–boundary layer interaction, and flow separation. KeywordsUnsteady aerodynamics; transonic flow; spatio-temporal graph neural networks; autoencoder; Benchmark Supercritical Wing; CFD; reduced-order modelling; machine learning.