A Surrogate Model for Urban Wind Flow Prediction Around a High-Rise Building

Document Type

Conference Paper

Publication Date

2025

Publication Title

PROCEEDINGS OF ASME 2025 INTERNATIONAL MECHANICAL ENGINEERING CONGRESS AND EXPOSITION, IMECE2025

Abstract

Urban wind flow, especially around high-rise buildings, affects ventilation, pollutant dispersion, wind energy harvesting, and pedestrian comfort. Accurate prediction of wind statistics is vital for urban microclimate modeling and resilient city planning. However, high-fidelity CFD simulations like LES are computationally expensive, while low-fidelity RANS models lack precision. This study proposes a machine learning-enhanced multi-fidelity surrogate model (MFSM) that integrates RANS and LES data to achieve accurate and efficient wind field prediction. A deep learning model is trained on extensive RANS outputs with selective LES augmentation, enabling robust generalization across wind conditions and elevations. The MFSM is validated against high-fidelity CFD results and evaluated using 2D slices extracted at multiple heights. Results show that the model achieves LES-level accuracy at a fraction of the computational cost, supporting applications in urban ventilation, comfort, and wind energy planning. Its scalability across complex urban geometries demonstrates strong potential for real-time wind flow prediction.

Comments

Paper presented at:
2025 International Mechanical Engineering Congress and Exposition-IMECE, Memphis, TN, NOV 16-20, 2025

DOI

10.1115/IMECE2025-166816

Volume

9

Share

COinS