Skip to content
SEINA-3D

What the module models

Long-duration three-dimensional premixing pours—lasting several seconds from initial jet entry until melt quenching and floor accumulation—require strict global conservation of mass and energy alongside detailed tracking of hundreds of thousands of solidifying melt particles. In SEINA-3D v2, the dedicated premixing code couples a finite-volume compressible liquid–gas carrier on block-structured AMReX Cartesian grids with Lagrangian melt parcels and a bottom porous debris-bed model.

Governing equations

Liquid water (l\mathrm{l}) and the gas phase (g\mathrm{g}, pure steam or a steam–argon headspace mixture obeying Dalton’s partial-pressure law p=pv+ρ~aRArTgp = p_{\mathrm{v}} + \tilde{\rho}_{\mathrm{a}} R_{\mathrm{Ar}} T_{\mathrm{g}}) evolve in conservative finite-volume form over the fluid pore volume fraction εf=1−αm\varepsilon_{\mathrm{f}} = 1 - \alpha_{\mathrm{m}}:

∂ρ~c∂t+∇⋅(ρ~cuc)=Γc,c∈{l,v,a}\frac{\partial \tilde{\rho}_c}{\partial t} + \nabla \cdot (\tilde{\rho}_c \mathbf{u}_c) = \Gamma_c, \qquad c \in \{\mathrm{l}, \mathrm{v}, \mathrm{a}\}

Dispersed melt is represented by Lagrangian parcels pp, each carrying total mass mpm_p, representative particle count NpN_p, diameter DpD_p, position xp\mathbf{x}_p, velocity up\mathbf{u}_p, and total enthalpy HpH_p:

dxpdt=up,mpdupdt=mpg−Vp∇p+Fd,pl+Fd,pv,dHpdt=−Qp→c−Qp,rad\begin{aligned} \frac{\mathrm{d}\mathbf{x}_p}{\mathrm{d}t} &= \mathbf{u}_p, \\ m_p \frac{\mathrm{d}\mathbf{u}_p}{\mathrm{d}t} &= m_p \mathbf{g} - V_p \nabla p + \mathbf{F}_{\mathrm{d},p\mathrm{l}} + \mathbf{F}_{\mathrm{d},p\mathrm{v}}, \\ \frac{\mathrm{d}H_p}{\mathrm{d}t} &= -Q_{p\to\mathrm{c}} - Q_{p,\mathrm{rad}} \end{aligned}

Drag impulses and film-boiling plus grey-body radiative heat losses from all parcels inside a cell enter the local liquid and gas control volumes as equal and opposite interphase source terms, bounded by available carrier mass, momentum, and heat capacity. Parcels that settle onto the vessel floor transition into column- or cell-grouped porous debris-bed enthalpy states, continuing to exchange heat and vapor with overlying and interstitial coolant.

Numerical method

On block-structured AMReX Eulerian field arrays and distributed Lagrangian parcel containers, the solver advances each time step from a verified start-of-step state. Within each step, nozzle feed, parcel motion, wall collisions, paired interphase drag and phase change, compressible common-pressure solution, and equation-of-state inversion preserve discrete mass and total energy even across disappearing or trace phases. Trial steps that violate phase positivity or pressure-iteration tolerances repeat from the start-of-step state with a reduced Δt\Delta t. Multi-second runs support binary checkpoint restart across distributed-memory message-passing (MPI) decompositions.

Verification

  • Conservation, domain decomposition, and restart verification: automated unit and integration benchmarks verified machine-precision mass and total-energy conservation across disappearing-phase states, single- and multi-rank AMReX domain decompositions, and multi-segment checkpoint restarts.
  • Three-dimensional FARO L-28 simulation: demonstrated on the 8.0 s8.0\ \mathrm{s} FARO L-28 melt-jet premixing case (175 kg175\ \mathrm{kg} UO2–ZrO2\mathrm{UO_2\text{--}ZrO_2} melt at 3053 K3053\ \mathrm{K}, 44 mm44\ \mathrm{mm} jet pouring into 1.44 m1.44\ \mathrm{m} of saturated water at 0.51 MPa0.51\ \mathrm{MPa}), capturing two-phase pool swelling, upward steam-chimney transport, fragment quenching, and floor debris-bed buildup. Earlier Eulerian premixing validation was reported in Five-field multiphase framework (SEINA-3D v1).