Non-equilibrium film transfer after pigging

TwoFluidPipe has an opt-in NonEquilibriumFilmTransfer closure for a gas and an aqueous MEG/water liquid with independently transported compositions. The gas and liquid retain the existing two-fluid momentum equations, finite gas holdup, pressure response, terrain forces, and conservative phase and component transport. A liquid inventory can therefore move through cells and be initialized as a larger pool at a low point; it is not pinned to the pipe wall.

This is an experimental isothermal screening capability. It does not qualify field drying duration, spontaneous pool formation, severe slugging, or long-time disappearance of the film. The complete acceptance criteria in issue #4281 remain open until long-horizon numerical and experimental qualification is available.

Distinguish the two questions

PipelineEvaporationStudy marches a flowing two-phase feed over distance. Its phase amounts represent molar flow rates. Selecting its wall-film geometry does not create a stationary post-pigging inventory or calculate elapsed drying time. See the axial evaporation guide.

For the transient question, setInitialAqueousFilm(liquid, holdup) supplies a separate liquid composition and initial aqueous volume fraction for each physical cell. The liquid starts at rest; the initial gas velocity comes from the gas-only steady initialization. This seeded state is a transient initial condition, not a converged two-phase steady solution. Inlet gas composition does not replace the initial liquid composition.

A 99/1 wt% MEG/water film can initially absorb water from methane specified at a liquid-water dew point of -18 C at 70 bara, while MEG evaporates. Water content must not be forced to decrease. Gas dew-point compliance and removal of residual MEG are different completion criteria. The regression uses the issue’s approximately 35.54 ppm inlet water at 5 C and 70 bara; this is a CPA model reference, not an independent experimental measurement or an ice frost-point value.

Transfer closure and units

For each configured species, the gas-directed mass flux is

\[j_i=k_i c_g M_i\left(x_{i,l}\frac{\phi_{i,l}}{\phi_{i,g}}-y_{i,g}\right).\]

Here $j_i$ is in kg/(m2 s), $k_i$ is a user-supplied overall gas-side coefficient in m/s, $c_g=\rho_g/M_g$ is the actual gas molar density in mol/m3, and $M_i$ is component molar mass in kg/mol. The mole fractions and fugacity coefficients are evaluated from separate homogeneous EOS states at the transported gas and aqueous compositions and the local common pressure and temperature. A bulk TP flash is not used to repartition the cell inventory. Equal fugacities give zero flux. This is a dilute, lumped-resistance closure; it does not implement the coupled Maxwell-Stefan, finite-flux, or interfacial-temperature calculation in the axial evaporation API.

Only explicitly named species transfer. Configuring water and MEG holds methane in its existing phase inventory: methane dissolution is not implied. Coefficients are inputs to be measured, correlated, or varied in a sensitivity study; there is no calibrated default.

The source per metre is $j_i S_i$. A positive constructor wetted-perimeter fraction prescribes that fraction of the full pipe circumference for $S_i$; zero uses the stratified interfacial width calculated from the current holdup instead. The former can screen a residual wall film, the latter a stratified pool. Neither option predicts wetting, dry patches, or the coexistence of a separate wall film and pool. Transfer stops if either receiving gas contact or aqueous inventory is absent: this closure does not nucleate new liquid in a dry cell.

Each source is capped at half of its donor component inventory per substep. The remaining half is reserved for advection under the required Euler/CFL limit. Phase sources are the sums of component sources. Each component carries donor-phase momentum, including when opposing component fluxes cancel in the net phase mass source. Component and phase reports must both close; unsupported or nonfinite states reject instead of reporting successful drying.

Configure and inspect a calculation

The executable Java setup is TwoFluidPipeFilmDryingTest.pipe, under src/test/java/neqsim/process/equipment/pipeline. All public methods are also callable through the Java bridge from Python. In sequence:

1. Define a CPA gas and a separate 99/1 wt% MEG/water liquid, with mixing rule 10. Both must have the identical named-component slate; the gas may contain zero MEG. Specify pressure in bara and temperature in K. Prepare each homogeneous phase explicitly.
2. Set pipe length, diameter, cell count and, if needed, N+1 cell-face elevations. Provide one initial holdup per physical cell. A full-circumference film of thickness $\delta$ has $\alpha_l=1-(1-2\delta/D)^2$; a local pool needs its own geometry-based inventory.
3. Enable component transport and mass transfer, select Euler, and use CFL no greater than 0.5. Install NonEquilibriumFilmTransfer and setInitialAqueousFilm before run.
4. Disable energy equations, wall heating, and tracked slugs. The opt-in route holds each cell’s temperature fixed. Pressure remains a hydrodynamic state; it is not prescribed uniformly.
5. Enable setTransactionalTransientEnabled(true) to isolate rejected full intervals. Advance with runTransient, recording the actual accepted elapsed time and each immutable component and phase balance report. setStoreComponentConservationHistory(true) retains the time-aligned reports. Sum their boundary ledgers for cumulative gas use and component removal.
6. Read component mass-fraction, phase-velocity, holdup, pressure and temperature profiles. getEquivalentAqueousFilmThicknessProfile reports the full-circumference thickness equivalent of the local aqueous holdup. For a pool this is not its physical liquid depth.
7. Declare a positive operational residual mass threshold and a maximum simulated horizon before running. getFilmDryingStatus(residualMassKg, horizonSeconds) returns RUNNING, RESIDUAL_THRESHOLD_REACHED, or HORIZON_REACHED without altering mass. Report threshold reached, horizon reached with residual liquid, or solver rejected. Never relabel the threshold as exact mathematical zero. Do not remove sub-threshold mass from the conservation ledger.

The source model currently requires single-stage Euler because the component inventory is updated once per accepted substep. Energy coupling, multi-stage integration and tracked-slug combinations reject explicitly. The legacy bulk-equilibrium relaxation path is unchanged when the closure is not selected. Its supported combinations must not be inferred for this experimental route.

Evidence and remaining qualification

The tests cover opposing water/MEG flux signs, a zero-coefficient control, independent liquid initialization, closed and open component conservation, liquid momentum on a seeded low point, constant temperature, serialization/repeatability, short-time grid/time refinement, a 10 km short transient, equilibrium zero flux, donor limits, absent-film contact, counter-transfer momentum and unsupported-configuration rejection. Numerical tests are not experimental drying validation. The low-point test checks motion and retention of a seeded pool; it does not establish an accurate terrain-trapping model.

Relevant independent VLE data exist in Folas et al., Fluid Phase Equilibria 251 (2007), 52-58, DOI 10.1016/j.fluid.2006.11.001, with a public NIST ThermoML record. That source reports MEG/methane and MEG/water/methane phase equilibria. It has not been used here to fit transfer coefficients or to establish an error bound for this implementation.

Before reporting a 10 km, 5 C drying duration, qualify the actual gas composition, flow, diameter, film inventory, terrain, grid, internal time step, residual threshold and transfer coefficients. Required further evidence includes long-horizon positivity and component closure, spatial/time refinement of pool redistribution and disappearance, carrying-capacity limits, experimental VLE comparisons, and independent drying/transport data. Use the broader TwoFluidPipe evidence matrix as well.

Extended-horizon probes

A 100 m, 0.20 m diameter synthetic low-point case at 0.5 kg/s gas flow and coefficients of 0.001 m/s accepted 30 intervals of 0.1 s with converged component ledgers. Initial aqueous holdup was 0.002, except 0.02 in the seeded pool. Over 3 s, that pool’s holdup decreased to approximately 0.015834 as liquid remobilized. Net transfer into liquid was approximately 0.00034609 kg water and -0.000030632 kg MEG. This reproducible numerical case is retained in seededLowPointRemobilizesOverThreeSeconds; it is not a drying-time measurement.

The long-horizon 10 km gate is not passed. An exploratory 8-cell run with 1 s outer steps and the uncoupled pressure reconstruction accepted its first interval but then generated an unphysical pressure excursion (about 1548 bara in a trial state). The forced-phase property check rejected the trial when the gas EOS root was classified as dense liquid. Enabling the existing coupled pressure/momentum and interfacial-pressure options instead encountered unsupported aqueous inflow through the outlet during the first interval. No outlet liquid composition was provided, so that path correctly rejected rather than inventing incoming fluid. Both probes used full-interval transactions; rejected trials were not published as accepted states.

These failures concern sustained hydrodynamic/pressure/boundary coupling and are recorded against #4281, not hidden by weakening donor, phase-identity or boundary guards. The 0.1 s 10 km regression is an initialization/short-time conservation check only. Do not extrapolate it into a full-line evaporation duration. Further pressure/boundary work, representative mesh refinement, and experimental qualification are required.