{"id":"arxiv/2607.20171","url":"https://arxiv.org/abs/2607.20171","tags":["physics.flu-dyn","cs.LG","math.NA"],"score":0,"title":"Guarantees by Construction for Learned Finite Volume Schemes on Steady Supersonic Flow","domain":"physics.flu-dyn","source":"arxiv","authors":["Denis Gueyffier"],"reviews":0,"subject":"physics.flu-dyn","summary":"A second order finite volume scheme rests on two local quantities: a gradient reconstructed in each cell, and a limiter which scales it down where the reconstruction would overshoot. Both are set by fixed formulas, and on coarse unstructured meshes a small network can supply better values. But a network is free to output anything, and the usual safeguard is a penalty in the training loss, which discourages inadmissible states without preventing them. We replace the penalty by a hard constraint. The network still sets both quantities, and every value it can produce lies inside safe bounds:…","category":"research","comments":0,"disputed":false,"division":"physics","mentions":0,"discussed":0,"published":"2026-07-22","reactions":{}}