Dataset and Code for "Intelligent Prediction of Aerodynamic Parameters for Flow Past 2D Bluff Body Sections at a Constant Reynolds Number"

This repository contains the dataset and source code supporting the research article entitled "Intelligent Prediction of Aerodynamic Parameters for Flow Past 2D Bluff Body Sections at a Constant Reynolds Number". Dataset The dataset contains 1,000 samples of 2D irregular convex bluff body sections. Each sample consists of: A 9-channel geometric feature tensor (224 × 224 pixels), representing the aerodynamic shape via a consistent shape representation method (distance field + coordinate fields). A corresponding Strouhal number (St) label, extracted from 2D URANS CFD simulations at Reynolds number Re = 2 × 10^4. File: dataset_1000_samples.npz X: shape (1000, 224, 224, 9), dtype float32 — input geometric feature tensor. y: shape (1000,), dtype float32 — corresponding Strouhal number labels (raw values, not standardized). Code The source code for the ResCA-Net model, data loading, and training pipeline is provided in model.py. File: model.py Defines the ResCA-Net architecture (Residual Network with Coordinate Attention). Includes data loading, training loop, and model saving/export functionality. The code expects the dataset file dataset_1000_samples.npz to be in the same directory. Requirements: Python 3.8+, TensorFlow 2.10+, NumPy, SciPy, Matplotlib.