Reproduce “The norm, not the labels: training and evaluating finite-element surrogates for stress analysis in the energy norm” 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

Ruifeng Cao, Xidan Song

The authors train surrogates for linear elastostatics with the assembled energy norm, which needs no solved labels and targets stress-relevant error. In a three-dimensional transformer comparison, changing supervised loss from displacement to energy norm reduced median energy-norm error 7.5-fold and von Mises error 6.1-fold. The two-dimensional mean energy-gap gain did not replicate on held-out data; no network transferred to a finer mesh.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.