Paper urges four-dimensional treatment of learned fluid turbulence
Original title: Let's Do Turbulence Right: In Space and Time!
Using direct numerical simulations larger than 28 TB per run, the authors find spatial and temporal checks can disagree.
The paper uses large time-resolved direct numerical simulations of three-dimensional Navier-Stokes turbulence to study how numerical choices affect resolved space-time physics. It then examines how neural networks represent the recorded motion. The authors report that spatial and temporal assessments can disagree for both the numerical data and learned representations, and argue turbulence should be treated as a four-dimensional phenomenon.
- Simulations exceed 28 TB per run.
- The subject is the multi-scale energy cascade in Navier-Stokes turbulence.
- Spatial metrics and temporal metrics need not agree.
- The appeal is directed at the machine learning community.
From the original article
1 min read at arxiv.orgThe three-dimensional Navier-Stokes equations govern fluid turbulence, a chaotic and non-linear physical phenomenon characterized by complex eddies interacting across vast spatial and temporal scales. A numerically accurate modeling of this multi-scale energy cascade requires fully resolving both the fine spatial structures and the continuous temporal evolution of the flow field.
By Maximilian Plattner · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.