Hydrodynamics · analytical study

Connected-tank sloshing analysis

Illustrative U-tube anti-roll response and a single-tank tuned-absorber comparison.

Summary

An invented, illustrative U-tube system comprises two tank legs connected by a conduit. Analysis is performed to determine the connected-tank response and compare it with a single-tank tuned absorber. The geometry represents no particular vessel or client. The results are summarized below.

Analytical predictions are evaluated using the connected-tank model and the tuned-absorber model, with published aggregate calibration results. Archived computational fluid dynamics (CFD) evidence is presented separately from the illustrative predictions. The predictions support screening; design qualification is not established.

The published moment-fit band limits the use of the aggregate error statistic. Transfer to the invented geometry remains an assumption, including inside that band. The single-tank comparison illustrates peak splitting and merging; its amplitudes do not establish a design ranking.

Related: generic hull roll screening estimate.

Draft for maintainer review.

Revision history

RevisionDateWhat changed and why
Related pages2026-10-27Link to the generic hull roll screening estimate added for related anti-roll context.
Display precision2026-10-23Damping-axis labels distinguished and comparison percentages rounded to one decimal place; source precision retained in the evidence record.
Evidence2026-10-23Archived CFD reductions added with provenance and evidence-class limits to distinguish measurements from illustrative predictions.
Geometry2026-10-23Independent illustrative dimensions and conduit area selected to separate the predictions from the reference geometry.
EditorialNot assignedEngineering descriptions clarified to distinguish illustrative predictions from calibration evidence.
InitialNot assignedPublic-source analysis added to document model scope and remaining work.

Illustrative geometry

Invented round dimensions are selected independently of the published reference geometry. Validation of the illustrative plan areas and centroid offset is not established. Fill depth is an input; a validated fill-percentage range is not established.

InputIllustrative valueSource and assumption
Each leg length × width30.0 × 8.0 mIndependent invented round dimensions
Each leg plan area240.0 m²Product of the illustrative leg length and width
Leg centroid offset from centreline10.0 mIndependent invented round offset
Fill depth4.0 mIndependent invented round depth
Conduit area for period sweeps8.0 m²Independent invented area within the published conduit-area input range
Effective conduit length5.2 mCalibrated reference value from the connected-tank model; transfer is assumed
Published conduit-area input range3.4–13.5 m²Reference period inputs; individual CFD measurements are not reproduced

The illustrative conduit area lies inside the published input range in the table. The calibrated effective length is transferred as an assumption; a measured natural period for the new geometry is required before predictive use, as described in the model calibration basis.

Published validation

The calibration quantities below are reported in the model calibration summary and published calibration checks. RMS denotes root mean square error; each percentage applies to the corresponding reference fit.

Published quantityValue
Quadratic conduit loss: calibrated area exponent2.138
Calibration cases (aggregate results only)35
Calibrated loss-law RMS (%)8.6
Refitted linear-in-area law RMS (%)36.5
Closed-form moment RMS (%)4.3
Moment fit applies from forcing period (s)20
Degradation reported below about (s)16
Effective-length inflation from forced-sweep peak (times)2.76

The loss comparison is based on a separate refit of each area law, as documented in the calibration checks. A fixed-coefficient exponent substitution is excluded. Effective conduit length is calibrated against a measured natural period, as specified in the calibration basis; the forced-sweep peak is loss-controlled and varies with roll amplitude.

Computed curves

The connected-tank curves are evaluated using the analytical model at the illustrative inputs listed above. Frozen-fluid prismatic roll inertia is calculated in the inertia curve; a validation error for rigid-column conduit inertia is not established by that curve.

Rigid-column natural period Period (s) against Conduit area (m²). Rigid-column model. Rigid-column natural period 0.0 3.4 5.5 5.4 11.0 7.4 16.5 9.5 22.0 11.5 27.5 13.5 Conduit area (m²) Period (s) Rigid-column model Frozen-fluid roll inertia Inertia (million kg m²) against Conduit area (m²). Frozen-fluid prismatic model. Frozen-fluid roll inertia 0.0 3.4 46.2 5.4 92.4 7.4 138.6 9.5 184.8 11.5 231.0 13.5 Conduit area (m²) Inertia (million kg m²) Frozen-fluid prismatic model

Damping is evaluated at forcing period 20.0 s with imposed level-difference amplitudes from the published damping inputs. The calculated damping depends on amplitude; it is not a fixed tank property.

Quadratic-loss equivalent damping Damping ratio (%) against Conduit area (m²). Level-difference amplitude 0.1 m; Level-difference amplitude 0.5 m. Quadratic-loss equivalent damping 0.0 3.4 39.1 5.4 78.1 7.4 117.2 9.5 156.3 11.5 195.4 13.5 Conduit area (m²) Damping ratio (%) Level-difference amplitude 0.1 m Level-difference amplitude 0.5 m

The moment is calculated at 5.0° imposed roll and the illustrative conduit area, using the tank-moment model.

Tank-on-vessel moment Moment (MN m) against Forcing period (s). In phase; Quadrature; Magnitude. Shaded periods are outside the published moment fit band. Tank-on-vessel moment -28.8 16.0 -13.2 24.8 2.5 33.6 18.1 42.4 33.7 51.2 49.3 60.0 Forcing period (s) Moment (MN m) In phase Quadrature Magnitude

The coupled response is evaluated with the published response inputs: vessel roll period 20.0 s, hull damping 5.0 % and tank authority 12.0 %. Tank authority denotes a normalized stiffness scale, distinct from a mass ratio. In the coupled-response model, the forcing reference is fixed by the bare-vessel roll amplitude. The calculated response is roll per unit wave moment at that finite reference; quadratic loss makes it amplitude-dependent. Shading marks periods outside the published moment-fit band. Wave-spectrum and dimensional vessel calibration are not established.

Normalized coupled roll response Roll / unit wave moment against Forcing period (s). Bare vessel; Connected-tank model. Shaded periods are outside the published moment fit band. Normalized coupled roll response 0.0 16.0 2.0 24.8 4.0 33.6 6.0 42.4 8.0 51.2 10.0 60.0 Forcing period (s) Roll / unit wave moment Bare vessel Connected-tank model

Single-tank comparison: split and merging

The tuned-absorber model is evaluated at the frequency grid and damping values in the published response record. The mass ratio, denoted by μ, is 0.05; the frequency tuning ratio is 1.0. A single equivalent tank is coupled to an undamped primary vessel. The tank is not isolated, and its geometry is not matched to the U-tube system. Different normalizations and hull damping are used in the two models; their response amplitudes do not establish a design ranking.

Single-tank tuned-absorber benchmark Roll / static deflection against Forcing / vessel natural frequency. Damping 3.0 %; Damping 9.5 %; Damping 12.7 %; Damping 18.0 %; Damping 30.5 %. Single-tank tuned-absorber benchmark 0.0 0.8 4.8 0.9 9.6 1.0 14.5 1.1 19.3 1.2 24.1 1.3 Forcing / vessel natural frequency Roll / static deflection Damping 3.0 % Damping 9.5 % Damping 12.7 % Damping 18.0 % Damping 30.5 %
Tank damping (%)Resolved peaksPeak normalized roll
3.0224.1
9.529.9
12.728.9
18.019.2
30.5113.5

Two peaks are resolved in the light-damping curves; they merge as damping increases, and the peak rises again at the highest damping shown in the table. These curves are evaluated using the single-tank model and the published response record. New connected-tank CFD results are not included. The connected-tank curves are presented separately; a matched single-tank and connected-tank comparison is not established.

Validity band

The moment model degrades below about 16 s and the CAUSE IS NOT ESTABLISHED. An earlier explanation was withdrawn. The residual also depends on roll amplitude.

The reported moment RMS applies at forcing periods of 20 s or longer for the published calibration. Intermediate periods are not qualified by that aggregate statistic. Even inside the band, transfer to the invented geometry is an assumption. No new physical mechanism is proposed.

CFD evidence

The CFD cases belong to an earlier, archived campaign on a twin-tank geometry that is not shown. That geometry is separate from the invented illustrative geometry above. These CFD results were not produced for this page. The evidence record identifies the source reduction, field and digest for each value. Raw fields and geometry are excluded.

Conduit-area sweep

Exchange amplitudes are compared at forcing period 6.25 s and imposed roll 5.0°. Ratios are normalized to the smallest area. The historical analytical amplitude uses geometric conduit length; it is not the later calibrated model.

Exchange amplitude against conduit area Exchange amplitude (m³); conduit area (m²). Discrete archived cases; connecting lines guide the eye. Exchange amplitude against conduit area 5.9 7.9 12.2 16.3 25.6 34.6 3.4 6.8 13.5 Exchange amplitude (m³); conduit area (m²) CFD Historical analytical
Area (m²)CFD amplitude (m³)Analytical amplitude (m³)CFD ratioAnalytical ratio
3.47.95.91.0001.000
6.816.312.22.0622.056
13.534.625.64.3774.322

Exchange-amplitude comparison; ratios retain the archived precision.

Exchange natural period against conduit area Natural period (s); area (m²). Discrete archived cases; connecting lines guide the eye. Exchange natural period against conduit area 10.6 11.0 14.3 14.3 19.3 19.7 3.4 6.8 13.5 Natural period (s); area (m²) CFD-derived natural period Calibrated inertia model
Area (m²)CFD-derived natural period (s)Calibrated prediction (s)Historical geometric-length prediction (s)
3.419.319.727.0
6.814.314.319.4
13.511.010.614.1

Period bases differ: the historical period is analytical, not a CFD measurement. The middle area is the calibration reference.

Inertia and natural period

The archive publishes a ±3 % band for the inertia law using one calibrated effective conduit length. The reductions below include an exception: the low-fill check exceeds that band. The error is prediction minus CFD-derived period, divided by CFD-derived period. This checks exchange-mode inertia through natural period; it does not validate frozen-fluid roll inertia. The calibration period is 14.31 s against 14.31 s from the reduction; it is not an independent check.

Period prediction error: area / fill Period error (%); area (m²) / fill (%). Discrete archived cases; connecting lines guide the eye. Period prediction error: area / fill -3.0 -3.0 -0.8 2.4 3.0 3.7 3.4 / 50 13.5 / 50 6.8 / 25 6.8 / 70 9.5 / 50 Period error (%); area (m²) / fill (%) Prediction error Published upper band Published lower band
Area (m²)Fill (%)Prediction (s)CFD-derived (s)Error (%)Band comparison
3.45019.719.32.4Inside
13.55010.611.0-3.0Inside
6.82514.013.53.7Outside: exception
6.87014.614.41.4Inside
9.55012.312.4-0.8Inside

All archived off-calibration period checks; the independent intermediate-area check is included.

Fill sweep

Discrete low and high fill checks bracket the calibration fill; continuous fill coverage is not established. Equivalent damping depends on forcing period and imposed roll. The table retains each available condition; the damping chart selects the lower-roll case nearest each predicted natural period.

Natural period against fill Natural period (s); fill (%). Discrete archived cases; connecting lines guide the eye. Natural period against fill 13.5 14.0 14.4 14.6 25 70 Natural period (s); fill (%) CFD-derived Predicted Equivalent damping against fill Damping ratio; fill (%). Discrete archived cases; connecting lines guide the eye. Equivalent damping against fill 0.514 0.536 25 70 Damping ratio; fill (%) Near predicted natural period; lower imposed roll
Fill (%)Area (m²)Forcing period (s)Roll (°)Equivalent damping ratio
256.812.05.00.509
256.813.95.00.514
256.816.05.00.469
256.813.910.00.731
706.813.05.00.540
706.814.65.00.536
706.817.05.00.553
706.814.610.00.720

Fill-dependent damping at the archived forcing conditions; period predictions and measurements appear above.

Loss scaling and independent area check

The fitted area exponent is n = 2.138, with 8.6 % RMS over 35 cases. The naive inverse-area form has 36.5 % RMS after its own coefficient refit: it underpredicts damping at the smallest area (group RMS 55.1 %) and overpredicts at the largest (group RMS 108.3 %). These group RMS magnitudes are distinct from signed individual errors.

Equivalent damping against conduit area Damping ratio; area (m²). Discrete archived cases; connecting lines guide the eye. Equivalent damping against conduit area 0.338 0.358 0.441 0.559 0.672 0.775 3.4 9.5 13.5 Damping ratio; area (m²) CFD-derived, near predicted natural period Original inverse-area prediction
Area (m²)Fill (%)Forcing period (s)Roll (°)CFD-derived dampingOriginal inverse-area damping
3.45019.75.00.7750.358
9.55012.35.00.4410.559
13.55010.65.00.3380.672

Conditions near each predicted natural period. The original inverse-area prediction uses the reference calibration, not the separate all-data coefficient refit.

At the independent intermediate area, the existing law evaluated without refitting gives 8.7 % RMS. Including that area in the fit gives n = 2.112 and 8.7 % RMS over 38 cases. Percentages are rounded to one decimal place; the evidence record retains the source precision. The independent prediction and all-data refit are separate checks.

Forced-roll period sweep and numerical checks

The medium-mesh forced-roll cases are accepted for screening after cycle convergence. The exchange plateau is separate from the exchange natural period. Fine-mesh confirmation applies at the selected period only; it does not establish convergence throughout the sweep.

Medium-mesh forced exchange Exchange amplitude (m³); forcing period (s). Discrete archived cases; connecting lines guide the eye. Medium-mesh forced exchange 114.2 123.3 129.6 130.2 132.4 132.4 16 18 20 22 24 26 Exchange amplitude (m³); forcing period (s) Exchange amplitude
Forcing period (s)Exchange amplitude (m³)Roll-moment amplitude (N m)
16114.27473614.6
18123.38433125.4
20129.69103025.1
22132.49497852.5
24132.49673479.6
26130.29683196.3

Completed, cycle-converged medium-mesh sweep.

Numerical checkResult
Fine confirmation period (s)24
Exchange change (%)0.8
Moment change (%)1.5
Fine transient maximum Courant / nominal target0.549 / 0.5
Free-decay frequency error (%)0.3–0.3
Observed Richardson order0.191
Indicative fine-grid GCI (%)0.1

Fine-mesh Courant exceeds the nominal target. Free-decay convergence is non-asymptotic; the GCI is indicative. Free-decay validates frequency in its own case family, not twin-tank anti-roll response.

Evidence classes and withheld conclusions

Analytical screening cases are not CFD solves. Failed wrappers and open transfer are rejected. Short transfer checks are verification only. Trustworthy phase and anti-roll benefit remain withheld.

Analytical cases and failed wrappers: separate evidence classes Case count. Discrete archived cases; connecting lines guide the eye. Analytical cases and failed wrappers: separate evidence classes 9.0 144.0 Analytical screening Rejected wrappers Case count Reviewed case counts; not completed CFD totals
Evidence classResult familyDispositionFindingLimitation
CFD no rollfree decay convergenceaccepted for frequency validation7/7 completed; measured 0.755580-0.755807 Hz vs 0.758054 Hz analytical; 0.3-0.3 % errorObserved Richardson order 0.191; GCI 0.1 % is indicative because convergence is non-asymptotic
CFD forced rollperiod refinementaccepted for screening6/6 medium ten-cycle cases completed and cycle-converged; broad exchange plateau at 22-24 sSource-neutral twin-tank idealization; not a full vessel/tank structural-load model
CFD forced rollfine confirmationaccepted for screening24 s fine case completed; medium-to-fine change 0.8 % exchange and 1.5 % roll momentFine transient max Courant 0.549 exceeded nominal 0.5 target
CFD transferclosed and coupled verificationverification onlyClosed-outlet and coupled-roll checks completedShort-duration checks are not periodic-response evidence
CFD transferopen transferrejectedStopped at 0.682764 s without End markerUnphysical boundary-flux growth; exclude from conclusions
CFD matrixnine selected casesrejected0/9 wrapper cases succeededGenerated inputs are not CFD results
Analytical screeningscreening manifestmethodology only144/144 analytical cases valid with zero reported mass residualNot 144 CFD solutions
Reduced ordernormalization and response algorithmsmethodology onlyUseful trend screening and normalization collapseNot direct validation of the resolved twin-tank CFD
Reduced orderanti roll phase assessmentwithheldKinematic magnitudes may be individually consistentInertia-free model cannot establish trustworthy phase or anti-roll benefit

The source-neutral review controls the evidence classification; methodology and validation are not interchangeable.

Way forward

Further assessment is required before the illustrative response is applied to a design. The geometry assumptions, comparison limits and calibration limits define the outstanding work below.

Left

The archived period checks do not establish a direct frozen-fluid roll-inertia error bound. Raw CFD fields, meshes, unstaged case quantities and the archived geometry are omitted. Continuous fill coverage, matched single-tank versus connected-tank performance, and anti-roll phase or benefit remain unestablished. The earlier campaign does not qualify the invented illustrative geometry.

Analysis basis

The analytical scope is described in the sloshing study. The connected-tank calculations and aggregate calibration statistics are defined in the analytical model and published calibration checks. The single-tank comparison is defined in the tuned-absorber model and response record.