Coupled release and inventory depletion

ReleaseInventory is an opt-in process unit for one rigid, well-mixed equilibrium inventory and one short opening. Unlike a hypothetical source sample, its native runTransient removes mass and energy. The changed temperature, pressure and composition then determine the next release calculation. It works within both ProcessSystem and ProcessModel; neither container nor the existing separator/blowdown APIs are changed.

The original constructor remains a single-gas compatibility API. An overload ending in an explicit PhaseType selects GAS, OIL, LIQUID or AQUEOUS withdrawal. The selected equilibrium phase supplies the upstream state and withdrawn composition; the entire remaining inventory is then reflashed. This is a boundary condition, not a level/geometry model: no entrainment, slip, interfacial transfer rate or automatic fallback phase is inferred.

An additional opt-in ReleaseInventoryPhaseGeometry overload adds a cylindrical-vessel geometry gate to the explicit phase selection. It calculates the live liquid level and gas/liquid interface area from phase volumes, verifies that the selected phase occupies the declared opening elevation, and rejects an opening inside a caller-declared interface exclusion band. The band makes unresolved waves and entrainment machine-readable and fail-closed; it is not a carry-over correlation.

An overload accepting List<PhaseType> and phaseExhaustionMassFraction adds a caller-declared ordered transition plan. The first phase must initially be present above the threshold. A transient step advances only to the next named phase when the active phase inventory mass fraction reaches the explicit threshold. It never selects a phase from density, position or phase count. If the immediately next planned phase is not usable, the complete call fails atomically.

The constructor clones the supplied EOS system and scales its amount to the specified volume at its initial temperature and pressure. The supplied mole count defines composition, not vessel size or kg/s. Returned fluids and accounting snapshots are independent copies.

The dynamic-capability audit classifies the unit as DYNAMIC_LUMPED because it owns component masses and internal energy. This state-ownership category persists when the opening is closed or steady-state mode is requested. It does not certify runtime readiness: activation remains UNVERIFIED and source-frame evidence remains UNQUALIFIED.

Balance equations and numerical method

For component inventory $m_i$ in kg, selected-phase component mass fraction $w_{i,s}$, internal energy $U$ in J, fixed volume $V$ in m3, and release rate $\dot m$ in kg/s:

\[\frac{dm_i}{dt}=-w_{i,s}\dot m,\qquad \frac{dU}{dt}=\dot Q_{ext}-h_{0,s}\dot m,\qquad V=\mathrm{constant}\]

$h_{0,s}$ is the selected phase’s EOS upstream stagnation enthalpy in J/kg. The discharge removes enthalpy, including flow work; subtracting only specific internal energy would predict the wrong cooling. Enthalpy and internal energy can be negative under the selected EOS reference. Their signed balance is retained. The default ReleaseInventoryHeatTransfer.adiabatic() boundary sets $\dot Q_{ext}=0$. Opt-in constantPower(...) and linearConductance(...) boundaries add a caller-declared signed heat rate in W, positive into the inventory. The linear form uses $\dot Q_{ext}=UA(T_{environment}-T_{inventory})$ with explicit W/K and K inputs. Both require provenance and are evaluated from the immutable beginning-of-substep state. They do not predict fire heat flux, wall thermal inertia, radiation, interfacial heat/latent transfer or phase-transfer rates. No boundary work, inflow, kinetic inventory energy or potential-energy change is included.

Each explicit Euler substep removes selected-phase composition, integrates $\dot Q_{ext}-h_{0,s}\dot m$, then calls NeqSim’s volume/internal-energy flash with explicit m3 and J units. It verifies component, volume and energy closure to relative $10^{-7}$, including cumulative closure $U+H_{released}-Q_{ext}=U_0$ against the original inventory. The substep is limited by maxSubstepS, remaining duration, and 1% of current selected-phase mass. This bound prevents overdraw; it is not an accuracy estimate. Refine maxSubstepS and the external output timestep for each study. A call needing more than 10,000 substeps fails without committing this unit’s state. An isolated inventory can continue integrating an active heat boundary while its opening is closed. The cumulative signed heat is exposed by Balance.getHeatTransferEnergyJ() and by deterministic source-frame provenance together with the boundary type and provenance string.

For an ordered plan, the phase-exhaustion threshold is a model input, not a numerical tolerance. It represents the minimum phase inventory for which the selected homogeneous withdrawal boundary is considered applicable. The transition preserves all residual material in the reflashed vessel; it changes only which explicitly named equilibrium phase supplies subsequent outflow composition and enthalpy. Threshold sensitivity therefore belongs in each study’s validation matrix.

The general VU flash’s stopping tolerance is looser than the cumulative accounting target. Each substep therefore permits up to 32 solves/refinements of the same VU equations until volume and energy residuals are at most $10^{-11}$ relative to their respective scales. After the first general VU solve, a nearby single-phase state can use a coupled Newton correction including both volume/energy cross derivatives. A bounded line search reflashes each trial with the configured EOS and mixing rule; it accepts only an unchanged phase and a smaller scaled residual. Multiphase states, invalid Jacobians and rejected trials retain the general VU refinement path. This prevents dense-liquid stagnation caused by repeated diagonal updates without changing the inventory closure target or general VU solver. Failure to meet the target produces INVENTORY_VU_REFINEMENT_FAILED. The total number of VU solves/refinements is available through getLastVolumeEnergySolves() and frame provenance; it does not count the TP trials internal to each refinement.

ReleaseInventoryTest.linearHeatBoundaryRefinesAndWritesExecutableBenchmark executes the lumped-conductance boundary at 1.0, 0.5, 0.25 and 0.125 s maximum substeps. It checks monotone successive refinement, temperature bounds and the signed energy ledger, and writes target/source-term-benchmarks/release-inventory-heat-boundary.csv. This is a numerical boundary and conservation benchmark, not independent fire, wall or blowdown validation.

Retained-phase finite-rate inventory

The dynamic capability resolver classifies NonEquilibriumReleaseInventory as DYNAMIC_LUMPED for both ProcessSystem and ProcessModel, including when its release opening is closed. This describes stored-state ownership; runtime activation remains UNVERIFIED and source-frame evidence remains UNQUALIFIED under the dynamic capability contract.

NonEquilibriumReleaseInventory is a separate opt-in model for exactly one gas and one oil, generic-liquid or aqueous phase. It does not call a TP or VU equilibrium flash after a component transfer. Instead, it retains gas and liquid component masses separately and evaluates forced homogeneous EOS phases at one common pressure and temperature. A coupled Newton solve enforces

\[V_g(P,T,\mathbf{m}_g)+V_l(P,T,\mathbf{m}_l)=V_{vessel}\]

and

\[U_g(P,T,\mathbf{m}_g)+U_l(P,T,\mathbf{m}_l)=U_{target}\]

after every candidate substep. The state is committed only after per-component, total-mass, volume and energy residuals close. The selected withdrawal phase supplies the source composition and stagnation enthalpy. The other phase remains in the inventory rather than being re-partitioned by an equilibrium flash.

ReleaseInventoryInterphaseTransfer requires component coefficients in m/s, a donor-fraction bound and a non-empty provenance string. For live interface area $A$ in m2, gas molar concentration $c_g$ in mol/m3, component molar mass $M_i$ in kg/mol and explicit timestep $\Delta t$ in s, signed liquid-to-gas transfer is

\[\Delta m_i=k_i A c_g M_i\left(x_i\frac{\phi_i^l}{\phi_i^g}-y_i\right)\Delta t\]

Each component is bounded independently to the configured fraction of its donor inventory. Positive values evaporate and negative values condense. The phase-enthalpy difference associated with the transfer is retained as Balance.getInterphaseLatentEnergyJ() and deterministic frame provenance. It is an internal diagnostic, not an energy source: the total-energy solve already retains the complete thermodynamic effect and the global balance remains $U+H_{released}-Q_{external}=U_0$.

The model assumes thermal equilibrium between phases, prescribed film coefficients and one live cylindrical interface. It fails closed on phase disappearance, an opening inside the exclusion band, an unsupported phase assemblage, invalid fugacity/property data or a non-converged state solve. It does not predict separate phase temperatures, interfacial heat-transfer coefficients, nucleation delay, wall inertia, entrainment, droplets, population balance, pipe waves or solids. Its current evidence level is UNQUALIFIED.

The executable construction and both process-container paths are covered by NonEquilibriumReleaseInventoryTest:

Map<String, Double> coefficientsMS = new TreeMap<String, Double>();
coefficientsMS.put("methane", 2.0e-3);
coefficientsMS.put("n-butane", 2.0e-3);
ReleaseInventoryInterphaseTransfer film = new ReleaseInventoryInterphaseTransfer(
    coefficientsMS, 0.25, "synthetic-screening-input:v1");
ReleaseInventoryPhaseGeometry vesselGeometry = new ReleaseInventoryPhaseGeometry(
    ReleaseInventoryPhaseGeometry.Orientation.VERTICAL,
    vesselDiameterM, vesselLengthM, gasOpeningElevationM, interfaceExclusionHalfBandM);
NonEquilibriumReleaseInventory finiteRate = new NonEquilibriumReleaseInventory(
    "finite-rate-inventory", flashedFluid, vesselVolumeM3,
    openingDiameterM, 0.70, receiverPressurePa,
    selectedReleaseModel, maximumSubstepS, PhaseType.GAS,
    vesselGeometry, film);
process.add(finiteRate);
session.addInventorySource("opening", "finite-rate-inventory");
SourceTermFrame frame = session.runSteadyState().get(0);
frame = session.step(0.10).get(0);
NonEquilibriumReleaseInventory.Balance balance = finiteRate.getBalance();

getThermoSystem() returns a defensive clone of the selected withdrawal phase because a single SystemInterface cannot represent two retained compositions without reflashing them. Whole- inventory consumers must use getGasPhase(), getLiquidPhase() and getBalance().

The general VU solver refreshes caloric derivatives on its final TP state before returning, so inventory checks use current enthalpy and internal energy. Its update equations and stopping policy remain unchanged. Caloric-state consistency alone does not establish the stricter inventory closure; the same $10^{-11}$ gates and 32-solve bound remain mandatory.

The release model is explicitly selected. Its upstream pressure, temperature and enthalpy must match the selected phase and its upstream/exit composition must match that phase’s withdrawal. Exit static enthalpy plus specific kinetic energy must equal upstream stagnation enthalpy. Legacy SCREENING_ONLY or unresolved-station results are rejected. No alternate release model, property default or flash recovery is substituted.

Executable process example

The corresponding construction, registration, stepping, balance access and isolation calls are exercised by ReleaseInventoryTest.documentationExampleAndPhysicalIsolationWorkForBothContainers. Use imports from neqsim.process.safety.release, neqsim.process.processmodel, neqsim.thermo.system and java.util.List.

SystemInterface gas = new SystemSrkEos(300.0, 2.0); // K and bara
gas.addComponent("nitrogen", 1.0);
gas.setMixingRule("classic");
ReleaseInventory vessel = new ReleaseInventory("inventory", gas,
    1.0, 0.01, 0.7, 10000.0, new IdealGasReleaseModel(), 0.1);
ProcessSystem process = new ProcessSystem();
process.add(vessel);
SourceTermSession session = new SourceTermSession("inventory-study", process);
session.addInventorySource("opening", "inventory");
List<SourceTermFrame> initial = session.runSteadyState(); // no depletion
List<SourceTermFrame> next = session.step(0.5);
ReleaseInventory.Balance accounting = vessel.getBalance();
double emittedMassKg = accounting.getReleasedMassKg();
double outgoingEnthalpyJ = accounting.getReleasedEnergyJ();
vessel.setReleaseEnabled(false); // physically close this opening
List<SourceTermFrame> isolated = session.step(0.5);

For an equilibrium gas-over-liquid inventory, select the gas boundary explicitly:

ReleaseInventory phaseSelected = new ReleaseInventory("two-phase", flashedFluid,
    1.0, 0.01, 0.62, 101325.0, new HomogeneousEquilibriumReleaseModel(), 0.02,
    PhaseType.GAS);

Use the actual equilibrated phase type when selecting a hydrocarbon liquid (OIL or LIQUID). Construction rejects an absent or unsupported phase. If re-equilibration removes the selected phase during a transient call, the call fails atomically with INVENTORY_SELECTED_PHASE_ABSENT; callers must split the regime or select another assessed model.

For an assessed gas-to-liquid transition, declare both phases and the inventory mass-fraction boundary explicitly:

List<PhaseType> withdrawalPlan = Arrays.asList(PhaseType.GAS, liquidType);
ReleaseInventory staged = new ReleaseInventory("staged", flashedFluid,
    1.0, 0.01, 0.62, 101325.0, new HomogeneousEquilibriumReleaseModel(), 0.02,
    withdrawalPlan, 1.0e-4);

This does not model a vessel level, interface geometry, entrainment or interfacial transfer rate. Use a threshold justified for the modeled inventory and verify sensitivity around it.

Geometry-gated phase withdrawal

For a vertical or horizontal cylindrical vessel, supply internal diameter, cylinder length, opening-centre elevation from the vessel bottom, and an assessed interface exclusion half-band. The cylinder capacity must match volumeM3:

ReleaseInventoryPhaseGeometry geometry = new ReleaseInventoryPhaseGeometry(
    ReleaseInventoryPhaseGeometry.Orientation.VERTICAL,
    0.50, 2.0, 1.70, 0.05); // m, m, m, m
ReleaseInventory geometryGated = new ReleaseInventory("geometry-gated", flashedFluid,
    Math.PI * 0.50 * 0.50 / 4.0 * 2.0,
    0.01, 0.62, 101325.0, new HomogeneousEquilibriumReleaseModel(), 0.02,
    PhaseType.GAS, geometry);

Vertical level is liquid volume divided by cylinder cross-sectional area. Horizontal level is the bounded inverse of the circular-segment area, and interface area is chord length times vessel length. The implementation supports one gas phase above at most one oil, generic-liquid or aqueous phase. Multiple material liquid phases, an opening in the exclusion band, a mismatch between opening phase and selected phase, or loss of phase-volume closure fail the complete inventory call without committing state.

Frames retain inventoryLiquidLevelM, inventoryInterfacialAreaM2, opening and cylinder geometry, and inventoryGeometryStatus=OPENING_PHASE_CONFIRMED_ENTRAINMENT_BAND_CLEAR. This status means only that the opening is outside the declared unresolved band. It does not qualify an entrainment rate, interface wave model, droplet population, slip, or finite-rate heat/mass transfer.

For an area-based model, add process under "gas-area", construct the session from that ProcessModel, and use session.addInventorySource("opening", "gas-area", "inventory"). Keep dynamic mode enabled; this unit is constructed with setCalculateSteadyState(false). A steady run checks the present state without resetting the inventory or accounting. Repeated native evaluations with the same transient UUID do not withdraw twice; a different duration with that same UUID is rejected.

getBalance() provides immutable initial, remaining and released component masses in kg, initial/current internal energy and released enthalpy in J, fixed volume, simulation time and the last substep count. These integrated quantities are the balance evidence. A frame’s rate is the instantaneous rate at its simulation time, not a timestep average. Do not apply the frame rate as an additional withdrawal from this unit.

Frame and failure semantics

addInventorySource binds the inventory’s own geometry and release model. It rejects a second physical-source registration for the same unit. Existing addSource remains a hypothetical sample and does not enable inventory coupling.

The finite-pipe ReleaseInventory constructor adds flow-path length and specified Darcy friction before the model argument. That geometry is retained as inventory pressure and composition evolve and is exported by addInventorySource. It enables quasi-steady coupling to the ideal-gas and single-equilibrium-phase real-gas Fanno models; it does not add transient pipe-wave storage to the lumped inventory. Schema v1 accepts the paired optional flowPathLength and darcyFrictionFactor source fields.

Transient single-gas pipe line packing

IdealGasPipeDecompression is a separate process unit for finite-speed wave propagation and line-pack discharge. It solves the one-dimensional conservative mass, momentum and total-energy equations on a caller-selected finite-volume grid. The upstream end is closed; the downstream end is connected to a constant absolute receiver pressure. A local Lax–Friedrichs numerical flux and CFL-limited substeps advance the state, while specified Darcy friction changes momentum without removing adiabatic total energy.

IdealGasPipeDecompression pipe = new IdealGasPipeDecompression(
    "ruptured-line", gas, 100.0, 0.20, 1.0, 101325.0, 0.012, 80, 0.45);
process.add(pipe);

SourceTermSession session = new SourceTermSession("linepack-case", process);
session.addInventorySource("rupture", "ruptured-line");
SourceTermFrame frame = session.step(0.05).get(0);

The committed frame rate is the exact downstream mass flux used by the conservative update. Frame provenance includes pipe length/diameter, cell count, CFL, friction factor, integrator identity, last substep count, remaining clock, and cumulative discharged mass and total energy. Both ProcessSystem and ProcessModel paths use the same coupled-source registration; no second hypothetical withdrawal is calculated.

The model requires one nonreacting equilibrium gas phase, freezes composition and calorically perfect properties at the initial state, and remains UNQUALIFIED. It excludes real-gas property evolution, heat transfer, pipe elasticity, an upstream vessel, two-sided rupture, phase change, slip, entrainment and solid-bearing flow. Refine the grid and compare retained pressure/rate histories before using a result even inside this applicability boundary.

RealGasPipeDecompression uses the same geometry, process-container registration and conservative accounting, but recovers each cell through a NeqSim volume/internal-energy flash. Pressure, temperature and acoustic speed therefore evolve with the selected EOS. It fails the whole caller step atomically if any cell or the receiver boundary ceases to be one equilibrium gas phase. This adds real-gas transient property evolution, not heat transfer, wall elasticity, finite upstream vessel or two-sided coupling, non-equilibrium multiphase transport, solids, or qualification.

Compatibility frames carry releaseBasis=COUPLED_RIGID_ADIABATIC_GAS_INVENTORY. Explicit phase-selected frames carry releaseBasis=COUPLED_RIGID_ADIABATIC_PHASE_SELECTED_INVENTORY and inventoryWithdrawalPhase. Both carry the integrator identity, inventory volume/time, cumulative released mass/enthalpy, and rateTimeBasis=INSTANTANEOUS_AT_FRAME_TIME in provenance. Evidence stays UNQUALIFIED. Process, equipment, calculation UUID and session sequence semantics are retained. The configured maximum substep and last substep count are also recorded in provenance. Ordered-plan frames instead carry releaseBasis=COUPLED_RIGID_ADIABATIC_PHASE_TRANSITION_INVENTORY, the complete ordered plan, the exhaustion threshold, active phase, last-call transition count and cumulative transition count. All remain string-valued provenance in schema v1; SI source quantities are unchanged.

vessel.setReleaseEnabled(false) stops physical withdrawal and yields an INVENTORY_OPENING_CLOSED disabled frame after the next successful process call. It invalidates the prior equipment calculation identity. In contrast, session.setEnabled(sourceId, false) disables export only; process physics continue. Opening/closure changes occur at step boundaries; split steps at known event times.

Condition	Behavior
Receiving pressure at or above initial inventory pressure	Valid zero flow; no reverse flow modeled.
Substep would reduce pressure below receiving pressure	Bounded event location lands on receiving pressure; provenance records the physical release duration.
Selected phase is absent initially	Construction rejects the configuration.
Selected phase disappears after re-equilibration	INVENTORY_SELECTED_PHASE_ABSENT; the unit call commits no state.
Active planned phase reaches its explicit exhaustion threshold	Advances only to the immediately next usable caller-declared phase and records the transition.
Immediately next planned phase is absent or exhausted	INVENTORY_WITHDRAWAL_PHASE_PLAN_EXHAUSTED; the unit call commits no state.
Reactions, forced phases, solids or hydrates	INVENTORY_REGIME_UNSUPPORTED; no fallback physics is invented.
Invalid model, mismatched upstream state/composition, or screening result	Fail closed without committing inventory.
Volume, component or energy closure failure	Fail closed, with the corresponding closure diagnostic.
Native process step failure	Session emits INVALID without numeric source payload and becomes faulted.

Updates are transactional within this unit: failure preserves its pre-call fluid, accounting, clock and UUID. Other process units and the process clock may already have advanced. Follow the session recovery guidance; a steady rerun is not a rollback of the surrounding process. Caller-owned custom release models must be deterministic and must not mutate themselves or the process during evaluation.

Validation and applicability

ReleaseInventoryCpaIsolationTest exercises ordinary SRK-CPA methane/water inventories (90/10 mol%, mixing rule 10, 60 bara, 298.15 and 308.15 K, 0.02 m3) after the phase-workspace clone repairs. A test-only prescribed 0.001 kg/s flux isolates inventory bookkeeping from physical discharge-rate prediction. Gas and aqueous withdrawal must retain named-component, internal-energy/outgoing-enthalpy and fixed-volume closure. The checks compare caller state, phase compositions and fugacity derivatives before and after successful withdrawal, an exported copy’s reflash, and rejection after a completed VU substep. A retry must reproduce a fresh control.

Native sessions exercise both process containers and emit four additional schema fixtures. The retained target/source-term-benchmarks/cpa-inventory-refinement.csv records three substep sizes (0.4, 0.2 and 0.1 s) over 2 s for both withdrawal phases at both temperatures. This is numerical/conservation evidence for persistent gas/aqueous phases, not experimental CPA or release-rate qualification. It does not resolve the remove/add physical-slot lifecycle tracked in #4310, prove phase disappearance or reappearance, or implement finite-rate inventory heat/mass transfer. Those boundaries remain required before broader inventory claims in #3860.

ReleaseInventoryTest checks multicomponent gas depletion, phase-selected gas-over-liquid and liquid withdrawal using the homogeneous-equilibrium release model. A controlled deterministic release model isolates explicit gas-to-liquid exhaustion transition, conservation and timestep/ threshold refinement from release-model solver noise. The suite also checks component/energy/volume closure, cooling and decreasing release rate, input immutability, cloning, zero flow, isolation/reopening, invalid steps, pressure-boundary rollback, injected mid-step failure, and both process containers. Actual frames, including disabled and failed cases, are validated against the bundled JSON Schema.

Phase-selected source frames are schema-validated from both process containers. A three-level substep refinement checks convergent pressure for gas withdrawal from a methane/n-butane two-phase inventory, while released component mass verifies selection of the gas composition. For 0.04, 0.02 and 0.01 s maximum substeps over 0.2 s, the final pressures are 998944.206, 998944.262 and 998944.290 Pa; the fine/coarse pressure-difference ratio is 0.500. CI retains the component withdrawals and energy residuals in target/source-term-benchmarks/inventory-phase-selected-convergence.csv. The capability regression checks both process containers and preserves the distinction between audited state ownership and unverified runtime activation. The focused safety workflow also runs DynamicCapabilityBuiltInInventoryTest to catch missing built-in registrations.

For an independent temporal reference, the constant-gamma ideal-gas choked, rigid adiabatic solution follows by combining the balance equations with $U=mc_vT$ and the choked discharge relation. With $a=\dot m_0/m_0$ in 1/s:

\[\frac{m(t)}{m_0}=\left[1+\frac{\gamma-1}{2}at\right]^{-2/(\gamma-1)}\] \[\frac{T(t)}{T_0}=\left(\frac{m(t)}{m_0}\right)^{\gamma-1},\qquad \frac{p(t)}{p_0}=\left(\frac{m(t)}{m_0}\right)^\gamma\]

The test uses nitrogen at 300 K, initially 2.0 and 2.2 bara, 1 m3, a 0.01 m opening, $C_d=0.7$, receiving pressure 0.1 bara and 10 s duration. It compares three substep limits (0.2, 0.1 and 0.05 s) to the closed-form mass, pressure and temperature and checks first-order refinement. The reference uses the initial NeqSim molar mass/gamma; its temporal solution is independent, while the implemented inventory retains the real EOS and variable properties. The test tolerance is 0.3% of initial values and is not an experimental accuracy claim. CI retains target/source-term-benchmarks/inventory-convergence.csv.

The local 0.05 s results at 10 s are:

Initial pressure (Pa)	Remaining mass (kg)	Pressure (Pa)	Closed-form pressure (Pa)	Absolute energy residual (J)
200000	2.00989363	171096.452	171105.421	0.000224
220000	2.21087835	188197.626	188206.827	0.000243

The fine/coarse pressure-difference ratio is approximately 0.499 in both cases. These results separate numerical refinement from the small real-EOS/constant-gamma reference difference.

The transition benchmark spans thresholds 0.026175–0.026185 and maximum substeps 0.15–0.0375 s. All nine cases conserve 0.0003 kg release, commit exactly one transition and retain component, energy and volume closure. The finest nearby-threshold pressure spread is below 0.032%; successive step refinement is non-increasing within the asserted numerical tolerance.

This increment covers equilibrium, phase-selected withdrawal, explicit ordered phase-exhaustion transitions, and opt-in live cylindrical phase geometry with a fail-closed entrainment/interface band. The opt-in external energy boundary covers only caller-imposed constant power or lumped conductance. Predictive entrainment and slip, finite-rate interfacial heat/latent/component transfer, qualified fire/wall heat-transfer physics, real-gas transient decompression, non-equilibrium transfer, solid-bearing flow physics and experimental qualification remain separate work in #3860. The numerical limit of the selected release model still applies. No facility qualification or independent safety/domain review is implied. \n### Receiving-pressure event provenance\n\ninventoryPressureEquilibrationEvent records whether the last successful transient call landed on the no-flow boundary. inventoryReleaseDurationS is the physical discharge duration within that caller timestep. Event trials use the same EOS, component removal and enthalpy balance as ordinary substeps; failed event location commits no state.\n