Validation.

Liquid in a tank, against measured and exact answers. A liquid with a free surface in a closed tank: still, collapsing, and sloshing after a nudge. Each case runs through the same Liquid in a tank tool a user runs, at its default settings.

2 of 3 cases within tolerance; the miss is in red, and the tool is in beta until it passes. Run on commit 54f81c9, 2026-11-18. The tolerances were set before the runs and not moved after them.

The three casesError against the reference
casevaluereferenceworst errortoleranceverdict
A still tank water half-filling a 100 × 100 × 200 mm box, after 1 spressure with depth, 5 probesρgh, two layers0.07 %±1 %within
A collapsing water column 13 of 14 points within ±5 %; worst at Z 2.78surge front along the floorMartin & Moyce 1952, n² = 25.42 %±5 %outside
A tank nudged sideways 1.2021 Hz against 1.1966 Hzfirst sloshing frequency√(g k tanh kh) / 2π0.46 %±3 %within
A collapsing water columnMartin & Moyce, 1952

A column of water twice as tall as it is wide is let go at one end of a long channel, and the front of the surge is timed along the floor. The measured points are the mean column of the 1952 paper’s table for this shape, read from the table itself. The paper could not record the instant the water was let go, so it set every record to the same time when the front passed 1.44 column widths; MeshWeave’s front is set the same way.

11.522.533.5400.511.522.533.5T = n t √(g/a)Z = front / aT 0.41: measured Z 1.11, MeshWeave 1.005 (-9.44 %), not gatedT 0.84: measured Z 1.22, MeshWeave 1.186 (-2.75 %)T 1.19: measured Z 1.44, MeshWeave 1.440 (0.00 %)T 1.43: measured Z 1.67, MeshWeave 1.647 (-1.38 %)T 1.63: measured Z 1.89, MeshWeave 1.837 (-2.79 %)T 1.83: measured Z 2.11, MeshWeave 2.050 (-2.82 %)T 1.98: measured Z 2.33, MeshWeave 2.220 (-4.73 %)T 2.2: measured Z 2.56, MeshWeave 2.480 (-3.12 %)T 2.32: measured Z 2.78, MeshWeave 2.629 (-5.42 %)T 2.51: measured Z 3, MeshWeave 2.873 (-4.23 %)T 2.65: measured Z 3.22, MeshWeave 3.059 (-4.99 %)T 2.83: measured Z 3.44, MeshWeave 3.305 (-3.93 %)T 2.97: measured Z 3.67, MeshWeave 3.503 (-4.56 %)T 3.11: measured Z 3.89, MeshWeave 3.705 (-4.75 %)T 3.33: measured Z 4.11, MeshWeave 4.033 (-1.87 %)Martin & Moyce 1952, n² = 2 (red: outside ±5 %)MeshWeave, the surge front

What it says: the front runs 1 to 5 % behind the measured surge after that point, and one point is just outside ±5 %. The gap does not close with finer cells (20, 27 and 40 cells across the column give the same curve within half a percent) or with walls that do not drag, nor by tracking a thinner edge of the water. It is not the measurement’s scatter either: the paper’s six runs reach Z 2.78 between T 2.21 and 2.39, and MeshWeave’s front gets there at about 2.45, after all of them. This case is open: until it passes, Liquid in a tank is in beta.

A tank nudged sidewaysThe first sloshing frequency

A 500 mm tank half full of water is moved 5 mm sideways over 0.4 s and held, the movement sent as an acceleration file the way a user uploads one. The water then sloshes at the tank’s first natural frequency, which has an exact answer for a rectangular tank.

0 s1 s2 s3 s4 s5 s6 s255.4244.8liquid level 10 mm from the wall, mm (read from 1 s on)

What it says: the level by the wall swings by 5.4 mm at 1.2021 Hz, against 1.1966 Hz from the formula: 0.46 % apart.

How to read itSaid plainly

One temperature, no evaporation, and the air above the liquid at constant density. The still tank checks the pressure the liquid puts on its walls; the collapsing column checks how fast a free surface moves; the sloshing tank checks a tank under motion. Where gravity dominates a tank (a Bond number of 100 or more, as in the still and the sloshing tank here), surface tension is left out: it would change the liquid’s waves by under 1 %, and runs take far longer steps without it. The collapsing column’s narrow channel keeps it. Cryogenic liquids take their properties from NIST tables (checked within 1 % between the table’s points); heat leak and boiling are not covered here yet.