Measured 3D field error on held-out obstacle wakes.

Published 4 Aug 2026, updated 14 Nov 2026

One-jump 3D velocity predictions on a declared 32³ grid (32,768 cells), scored across 48 held-out cases at 32, 64, and 128 solver steps. This is the operator family shipping in Reyn Studio 0.4.1.

Evaluated 21 Aug 2026

Held-out 3D obstacle-flow map (32³).

Across 48 held-out 3D test cases, the one-jump operator reaches whole-field velocity RelL2 0.020 / 0.028 / 0.052 at 32 / 64 / 128 steps, compared with 0.084 / 0.136 / 0.209 for hold-last-state (4.0× margin at 128 steps).

H = 128 RelL2 0.052 vs 0.209 hold-last
Persistence margin 4.0× At 128 solver steps
Held-out corpus 48 cases 6 families × 8 geometries
Declared lattice 32³ 32,768 Cartesian cells
FIG 1.0: One-jump protocol

Single-jump 3D velocity prediction across unseen solid families.

Starting after the numerical solver’s 40-step warm-up, the operator maps the current 3D velocity field and binary occupancy mask directly to step 32, 64, or 128 in a single pass.

Plate 1.0: Held-out irregular 3D body (H = 32, 64, 128) Measured 32³ mid-plane vorticity (simulation vs map vs hold-last)
Held-out irregular body: mid-plane vorticity from the simulation, the learned map, and hold last state at 32, 64, and 128 steps
Unseen non-convex irregular obstacle (32³): Top-to-bottom or column comparison across 32, 64, and 128 steps shows the learned map tracking downstream shear-layer roll-up while static hold-last-state remains frozen at t₀.
Shipping bundle
reyn-flow3d-obstacle-32-v1.reynmodel (SHA-256 0e0524e9c665b1b46138…)
Grid & boundaries
32³ Cartesian lattice (32,768 cells) in a 3D periodic box
Warm-up & jump
40 solver warm-up steps, then single neural jump to H = 32, 64, or 128
Occupied vs empty
Occupied-family mean 0.062 (40 cases); empty-box mean 0.003 (8 cases)
FIG 2.0: Horizon ledger

Whole-field RelL2 from 32 to 128 steps.

Error grows gracefully with jump horizon while the persistence baseline diverges rapidly as wakes advect downstream.

0.052H128
128-step velocity RelL2 across all 48 held-out cases, versus 0.209 for hold-last-state (4.0× margin).
0.020 / 0.028
32-step and 64-step velocity RelL2 on the same corpus, against hold-last-state scores of 0.084 (4.2×) and 0.136 (4.9×).
0.062 / 0.003
Occupied-obstacle mean vs empty-box mean at 128 steps, reported separately so trivial obstacle-free cases never dilute bluff-wake error.
FIG 3.0: Family stratification

Error stratified across all six geometry families.

Multi-body wake impingement is substantially harder than isolated bluff flow. We publish all six eight-case families unaggregated so failure modes are transparent.

Plate 3.1: Horizon error growth
Line chart of whole-field relative L2 versus horizon for the learned map and hold last state
Horizon growth (32 → 64 → 128 steps): Mean RelL2 across all 48 cases, learned map vs hold-last-state.
Plate 3.2: Per-family H = 128 spread
Bar chart of 128-step relative L2 by geometry family for the learned map and hold last state
Per-family means at H = 128: Eight held-out cases per family; multi-body wakes represent the upper error bound (0.076–0.077).
Geometry family (n = 8 each) One-jump RelL2 Hold last state Margin Visual comparison (H = 128)
Single bluff body 0.041 0.187 4.6×
Irregular non-convex body 0.052 0.207 4.0×
Ground-adjacent body 0.063 0.280 4.4×
Multiple bodies (wake interaction) 0.077 0.301 3.9×
Irregular multiple bodies 0.076 0.271 3.6×
Empty periodic box 0.003 0.006 2.0×
FIG 3.1: CAD applicability gate

How 32³ behaves on real STL meshes, including refusals.

When an engineer imports CAD into Reyn Studio, preflight voxelizes the mesh onto this exact 32³ lattice. Select all six reference geometries to inspect solid-cell counts, velocity divergence RMS, and why thin wings are refused pre-solve.

Voxel applicability gate and velocity divergence (reyn-flow3d-obstacle-32-v1.reynmodel, 32³ lattice, Δx = 0.1963)
Recorded in Reyn Studio Studio 0.4.1 (4172f85), 6 imported STL cases
Orange square = 1 model cell (Δx = 0.196) Thinnest axis (y): 2.26 cells
Cleared preflight, inference recorded

Ahmed reference body

1.044 m · R100 nose · 25° slant (500 STL triangles)

Occupied solid voxels
60 / 32,768
Thinnest span (y-axis)
2.26 cells
Axis-ray disagreement
6.25%
Divergence RMS ‖∇·u‖₂
0.17713
Wake deficit peak
0.994
Streamwise Cx
0.1891
Field SHA-256
3f6fc76a44a30d40…
FIG 4.0: Volumetric slices

Uninterpolated 32³ cuts across XY, XZ, and YZ.

We render raw 32³ grid slices without bicubic smoothing so engineers can see the actual spatial resolution of the operator across isolated and multi-body wakes.

Plate 4.1: Orthogonal XY / XZ / YZ cuts
Orthogonal xy, xz, and yz vorticity slices at 128 steps for an irregular body, simulation versus learned map
Orthogonal XY / XZ / YZ cuts (H = 128): Three-plane vorticity structure around an unseen irregular body, simulation vs one-jump map.
Plate 4.2: Isolated bluff-body wake
Single-body mid-plane vorticity comparing simulation, learned map, and hold last state at 32, 64, and 128 steps
Single bluff body (RelL2 0.041 at H128): Mid-plane vorticity across 32, 64, and 128 steps vs hold-last-state (0.187).
Plate 4.3: Multi-body wake interaction stress test Measured Family mean RelL2 0.077 vs 0.301 hold-last-state
Multiple-bodies mid-plane vorticity comparing simulation, learned map, and hold last state at 32, 64, and 128 steps
Multiple-body wake impingement (H = 32 / 64 / 128): Downstream bodies sitting inside an upstream shear layer represent the hardest family in the corpus (0.077 RelL2). Large-scale wake deflection holds, while fine interstitial shear filaments soften at 32³.
Study boundaries

Why 32³ is never upsampled to pretend 64³.

This study covers a 32³ periodic box, six held-out obstacle and empty-box families, whole-field velocity RelL2, and a single jump to 128 steps after a 40-step warm-up. It does not qualify high-Reynolds turbulent boundary layers, compressible aerodynamics, internal ducting, or rotating machinery.

No silent grid resampling in Reyn Studio

CAD Run in Reyn Studio voxelizes imported STL geometry strictly to the bundle’s declared 32³ lattice (32,768 cells). Studio never interpolates a 32³ operator onto a 64³ or 128³ display grid to create an illusion of finer boundary-layer resolution.

Earlier 3D protocol

Solver validation & prior obstacle studies.

Earlier 3D solver verification and obstacle benchmarks conducted under a different grid and horizon protocol.

Archived protocol 3D solver precision (3.9×10⁻¹⁵) & prior obstacle run
Published 9 Jul 2026, updated 21 Aug 2026

Earlier 3D solver and obstacle evaluations

Before the 48-case 32³ one-jump occupancy benchmark above, our 3D reference solver validation reached machine-precision agreement (3.9×10⁻¹⁵ RelL2), and shorter-horizon single-family obstacle experiments reported 0.006–0.007 RelL2.

Protocol non-comparability note: The 0.052 score in the August study above evaluates a much longer 128-step single jump across 48 diverse multi-body and non-convex holdouts rather than a shorter single-obstacle horizon.

Validated three-dimensional vorticity research render
3D vorticity validation: Reference solver verification case (RelL2 3.9×10⁻¹⁵).
Earlier three-dimensional obstacle-flow research render
Earlier 3D obstacle study: Shorter-horizon protocol reporting RelL2 0.006–0.007.
Operating boundaries

What these numbers do and do not claim.

Every held-out score on this site is paired with the grid, regime, and failure gate that bound it in production.

Qualification

Verification, not certification

Relative L2, force RelL2, and Strouhal checks measure discrepancy against reference solvers on held-out test sets. They do not certify Reyn Studio for flight, road, marine, or safety-critical sign-off.

Grid resolution

Declared 32³ voxel grid

The shipping 3D research operator resolves 32,768 cells. Imported CAD thinner than one cell along any axis is refused before inference rather than extrapolated across empty solid voxels.

Flow regime

Tested Reynolds & horizon scope

Published scores cover laminar/transitional bluff-body wake regimes (Re = 100–355 in 2D; Re = 150 to 128 solver steps in 3D, RelL2 0.052). High-Reynolds RANS/LES boundary layers are outside current bundle scope.

Derived loads

Derived pressure & fluid tractions

Surface Cp is recovered from the predicted velocity field, and diffuse-interface tractions represent fluid loads for relative ranking or downstream FEA mapping, not internal structural stress.

Evaluate a geometry.

Studio packages are public to download; credentials are issued after review for external-flow candidate screening.