Reproduce “CoPiT: A conditional Position-induced Transformer for aerodynamic field prediction on unstructured meshes” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Junfeng Chen, Yongchen Fan, Zhonghua Qiao, Kailiang Wu
CoPiT predicts aerodynamic flow fields from mesh geometry and operating conditions using position-based attention and feature-wise linear modulation (FiLM) to condition features. On AirfRANS full-data tests, it reports surface and volume losses about 54% and 24% lower than Transolver. However, it is not uniformly superior: Transolver performs better on angle-of-attack extrapolation, and CoPiT’s drag correlation does not ensure accurate absolute drag.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.