Spaces:
Sleeping
Sleeping
|
Download docs/parameters.md from ilyaRozhkovA/LAMPSUI: direct link, hf CLI and curl.
- Browser
- Download file 3.93 kB
-
https://huggingface.co/spaces/ilyaRozhkovA/LAMPSUI/resolve/main/docs/parameters.md
- Command line
-
hf download hf://spaces/ilyaRozhkovA/LAMPSUI/docs/parameters.md
-
curl -L -o parameters.md https://huggingface.co/spaces/ilyaRozhkovA/LAMPSUI/resolve/main/docs/parameters.md
3.93 kB
| # Parameter reference | |
| These are the variables the UI exposes. They live in each example's | |
| `in.sdpd_phase.2d` as `variable NAME equal X` lines and are patched at run | |
| time when the user (or agent) supplies overrides. | |
| ## Run | |
| | Name | Type | Meaning | Sensible range | | |
| |------|------|---------|----------------| | |
| | `prodRun` | int | Production-phase timesteps | 100 (smoke) — 250 000 (publication) | | |
| | `preRun` | int | Equilibration timesteps before production | 0 — 20 000 | | |
| | `preRunTemp` | int | Initial temperature-equilibration steps | 0 — 1000 | | |
| | `dt` | float | Integration timestep | 1e-5 — 1e-3, smaller for stiffer EOS / smaller dh | | |
| ## Output | |
| | Name | Type | Meaning | | |
| |------|------|---------| | |
| | `fqTher` | int | Print thermo line every N steps | | |
| | `fqDump` | int | Write trajectory dump every N steps | | |
| | `fqRes` | int | Write restart file every N steps | | |
| | `fqV` | int | Re-apply velocity boundary conditions every N steps | | |
| For a 100-step smoke test use `fqTher=10, fqDump=50` so you actually get | |
| some output. For a 250 000-step production run, `fqTher=1000, fqDump=5000` | |
| keeps disk usage in check. | |
| ## Geometry | |
| | Name | Type | Meaning | | |
| |------|------|---------| | |
| | `dim` | int | 2 or 3 (every bundled example is 2) | | |
| | `dh` | float | Particle spacing in lattice units. Smaller = finer = slower | | |
| | `coreL` | float | Edge length of the simulation box | | |
| | `dhN` | int | dh × dhN = sphere/droplet radius | | |
| | `diaNx`, `diaNy` | int | Droplet diameter multipliers in X/Y | | |
| | `squareL` | float | Half-length of a square sub-region | | |
| | `lenNx` | int | Length-times multiplier for square regions | | |
| The kernel cutoff `h = 4·dh` is hard-coded. If you halve `dh` (×2 resolution) | |
| expect ~16× more particles in 2D and 4× more neighbours per particle. | |
| ## Fluid properties | |
| | Name | Type | Meaning | | |
| |------|------|---------| | |
| | `rho` | float | Reference fluid density | | |
| | `T` | float | kBT — thermal noise. Validation cases use ~1e-18 (effectively zero) | | |
| | `Tbuf` | float | Temperature in buffer / reservoir regions | | |
| ## Multi-phase / Surface tension | |
| | Name | Type | Meaning | | |
| |------|------|---------| | |
| | `sigmao` | float | Reference interfacial tension for the multi-phase term | | |
| | `F` | int (1, 2, or 3) | Selector for the multi-phase free-energy functional | | |
| | `fluid_type` | int | Phase identifier passed to the SDPD-mf pair style | | |
| | `f2Frac` | float (0–1) | Volume fraction of phase 2 (the dispersed phase) | | |
| | `randPos` | int (0/1/2) | 0=single droplet, 1=random, 2=square block | | |
| For a Young–Laplace droplet test, vary `sigmao` and the droplet size | |
| (`dhN`/`diaNx`); the pressure jump should follow Δp = σ/r in 2D. | |
| ## Boundary / driving | |
| | Name | Type | Meaning | | |
| |------|------|---------| | |
| | `dVx`, `dVy` | float | Imposed wall velocities (Couette/Poiseuille setup) | | |
| | `dSSx`, `dSSy` | float | Imposed shear strains | | |
| ## What's NOT exposed in the UI yet | |
| These live in each example's `micro` file (or `eta_input` / `pair_input` / | |
| `s_input` / `thixo_input` for cases that read them from data files): | |
| - `eta11`, `eta22`, `eta12` — viscosity per type-pair | |
| - `ksai11`, `ksai12` — multi-phase coupling constants | |
| - `s11`, `s22`, `s12` — strength factors | |
| - `wall_index*` — wall non-slip flag per type-pair | |
| - `a_thixo`, `b_thixo`, `alpha_thixo`, `f0_thixo` — thixotropy coefficients | |
| To change these, edit `Numerical examples/<case>/micro` directly (or its | |
| `*_input` files) before running. | |
| ## What to vary in an optimization | |
| Good choices for sweeps: | |
| - `dh` (resolution) — affects accuracy and runtime | |
| - `sigmao` (surface tension) — drives droplet shape recovery | |
| - `dt` (timestep) — affects stability margin | |
| - `f2Frac`, `randPos` — initial phase distribution | |
| Not great choices: | |
| - `prodRun` — varying it changes runtime, not physics | |
| - `dim`, `fluid_type`, `F` — categorical selectors, only meaningful at | |
| specific values; use them as `categorical` not `int` in search space | |