solve_flow

solve_flow(
    im,
    direction='x',
    n_steps=15000,
    nu=1.0 / 6.0,
    log_every=500,
    verbose=True,
    sparse=False,
    tol=0.001,
)

Run a pressure-driven single-phase LBM simulation to steady state.

Parameters

Name	Type	Description	Default
im	(np.ndarray, shape(nx, ny, nz))	Binary image of the pore space. 1 (or True) = pore, 0 (or False) = solid. This matches the PoreSpy convention.	required
direction	(x, y, z)	Axis along which the pressure gradient is applied. Default `'x'`.	`'x'`
n_steps	int	Number of LBM time steps to run. Default 15000.	`15000`
nu	float	Kinematic viscosity in lattice units. Default 1/6.	`1.0 / 6.0`
log_every	int	Print a progress line every this many steps. Default 500.	`500`
verbose	bool	Print progress to stdout. Default True.	`True`
sparse	bool	If True, use Taichi sparse (pointer-backed) storage for the distribution fields. Only pore cells are allocated, which can significantly reduce memory on images with high solid fractions. Default False (dense storage).	`False`
tol	float or None	Convergence tolerance. The simulation stops early when the relative change in the total velocity magnitude between log intervals falls below this value: `delta\\|v\\| / \\|v\\| < tol`. Set to `None` to always run the full `n_steps`. Default 1e-3.	`0.001`

Name	Type	Description
result	FlowResult	Result object containing `solid`, `rho`, `velocity`, `direction`, and `nu` as numpy arrays/values. Pass directly to `compute_permeability()` or `compute_hydraulic_conductance()`, or call `result.export_to_vtk(prefix)` to save a VTR file.

Taichi must be initialized by the caller before invoking this function:

import taichi as ti
ti.init(arch=ti.cpu)