|
PICurv 0.1.0
A Parallel Particle-In-Cell Solver for Curvilinear LES
|
|
Loading...
Searching...
No Matches
|
PICurv 0.1.0
A Parallel Particle-In-Cell Solver for Curvilinear LES
|
|
PICurv-specific terms, and the CFD terms this documentation uses in a particular way. Deprecated spellings are listed with their canonical replacements so a search for the old name lands here.
**Ucont** — the contravariant volume fluxes stored on cell faces. This is the quantity the solver evolves. Curvilinear codes evolve fluxes rather than Cartesian velocities because the flux formulation keeps mass conservation exact on a distorted grid.
**Ucat** — the Cartesian velocity at cell centres, reconstructed from Ucont. Derived, not evolved. A visualization file showing Ucat is showing a reconstruction of the solver's state, not the state itself.
**Ucat_nodal** — Ucat interpolated to grid nodes for visualization.
**Aj** — the inverse Jacobian of the curvilinear mapping, cell-centred.
**Csi / Eta / Zet** — face area vectors for the three computational directions \(\xi, \eta, \zeta\).
**Nvert** — per-cell classification marking blanked, immersed, or boundary-adjacent cells.
Shifted index architecture — cell-centred data for geometric cell i is stored at array index i+1. This is deliberate: it makes boundary handling symmetric. See Grid, Cell, and Variable Architecture Guide.
Pseudo-time / dual time — an inner iteration driving each physical timestep to convergence. Dual Time Picard Jameson RK iterates in pseudo-time; Explicit RK4 does not.
Pseudo-CFL — the dimensionless Courant number governing the pseudo-time step, \(\Delta\tau = \text{pseudo\_cfl} / \lambda_{max}\). Exclusive to the Picard-Jameson solver. Not checkpointed: a restart begins from the configured initial value.
Accepted versus attempted trial — the pseudo-time controller may reject a trial and roll back. max_iterations caps accepted iterations; a separate hard cap of 3 x max_iterations bounds total attempts.
Fractional step — the split that makes each timestep tractable: solve momentum for an intermediate velocity, then project it onto a divergence-free field with a pressure solve. See Fractional-Step (Projection) Method.
Projection — the step that enforces incompressibility. The momentum solve alone does not produce a divergence-free field.
Type versus handler — type says what kind of boundary a face is (wall, inlet, …); handler says how it is computed (noslip, parabolic, …). Not every handler is legal for every type. See 2. Boundary Types.
Driven periodic — a periodic pair plus a body force sustaining the flow against wall friction. Without it, periodic wall-bounded flow decays.
**target_flux** — a volumetric flux (bulk velocity times cross-sectional area), given in physical m³/s in YAML. Not a velocity. picurv divides it by \(U_{ref} L_{ref}^2\) before the runtime sees it.
Seam plane — the single plane where a periodic pair joins. enforce_seam_flux adds a local correction there, on top of the global body force.
Latent option — present in the C enum but not exposed through validation. Not selectable; listed so its absence is explicit rather than mysterious.
Swarm — the PETSc DMSwarm holding the Lagrangian particles.
Settling — locating each particle's host cell and migrating it to the owning MPI rank after advection.
Walking search — the cell-location algorithm that walks from a particle's previous host cell toward its new one, rather than searching globally.
Scatter — writing particle-carried quantities back onto the Eulerian grid. In the current implementation the scatter path targets the particle-derived scalar Psi; it does not feed particle momentum back into Ucat or Ucont. "Two-way coupling" is therefore accurate for the scalar field and not established for momentum — prefer the specific statement over the general phrase.
Advect-then-interpolate — the ordering of the particle step: a particle moves using the velocity interpolated at the previous step, and interpolation closes the step rather than opening it. See Particle Model and Coupling Overview.
Subgrid stress — the momentum transfer done by eddies smaller than a grid cell. It involves the unfiltered velocity, which the simulation never had, so it is modelled rather than computed. Closing it is the entire content of an LES model.
**Cs, and the coefficient C** — Cs is the Smagorinsky constant of the classical form, near 0.16 for isotropic turbulence. C is the coefficient that multiplies Delta^2 |S| in the eddy viscosity, which is Cs^2. The distinction matters here because the CS field stores C, not Cs: the dynamic procedure produces C as a ratio, and diagnostics convert to Cs only where a human reads the number.
Test filter — a second, wider filter applied to the already-filtered field. The stress carried between the grid filter and the test filter is computable from resolved data, which is what lets the dynamic procedure measure a coefficient instead of assuming one.
Germano identity — the exact algebraic relation between the stresses at the grid filter, the test filter, and the band in between. It contains no modelling. See 3. The Germano Identity.
**Leonard stress (L_ij)** — the computable stress in that band; the measurement the dynamic procedure fits its coefficient to.
Backscatter — energy flowing from unresolved scales back to resolved ones, which appears as a negative coefficient and a negative eddy viscosity. It is physical, and common pointwise in developed turbulence. Clipping modes discard it; clipping.mode: none keeps it and bounds the total viscosity instead.
Homogeneous direction — an axis along which the flow is statistically invariant. Averaging the dynamic coefficient over such directions is what Lilly's least-squares closure assumes, and averaging.mode: homogeneous derives them from the case's periodicity.
**Filter width (Delta)** — the length scale separating resolved from modelled motions, derived per cell from its metrics. Selectable, because the cube root of the cell volume is exact only for a cube.
Gradient (Clark) model — a tensor-diffusivity term added to the viscous flux, the leading term of a Taylor expansion of the Leonard stress. Not purely dissipative; combined with an eddy-viscosity closure it forms the classical mixed model.
**k_sgs** — the modelled subgrid kinetic energy. Reported as a diagnostic, never conflated with the resolved turbulent kinetic energy that the field-statistics pipeline derives from velocity fluctuations.
Capability entry (Tier 2) — the standard eight-part description of one selectable value. Defined in Documentation Extension Framework.
Evidence facet — an independent claim about what has been verified. Facets do not form a ladder. See Capability Status Vocabulary.
Accepted spelling versus deprecated alias — a spelling is a synonym with no migration story (smagorinsky for constant_smagorinsky); a deprecated alias is a retired name that should be migrated away from (Dual Time Picard RK4).
| You may see | Current name | Notes |
|---|---|---|
Dual Time Picard RK4 | Dual Time Picard Jameson RK | Deprecated alias; still parses |
rk4_residual_noise_allowance_factor | jameson_residual_noise_allowance_factor | Rename only |
apply_trim | enforce_seam_flux | The new name says what it holds on target |
params.vector, params.velocity | vx, vy, vz | Rejected, not silently accepted |
58_Turbulence_Statistics_Pipeline_Specification | 58_Field_Statistics | Page renamed |
/picurv-docs/<page>.html | /docs/picurv/<page>.html | Old page URLs do not resolve |
1.9.8