You're reading the documentation for a development version. For the latest stable documentation, please have a look at v5.1.X. For the latest pre-release documentation, please have a look at v6.0.0-alpha.1.
Flux reconstruction methods¶
Group /input/model/unit_XXX/discretization¶
GRID_FACES
Optional, required only for non-equidistant grids. An array specifying the coordinates of all grid faces, which are used for reconstruction. The array length must equal the number of grid faces, i.e., number of grid cells plus one, and the column boundaries must be included.
Note
large ratios of adjacent cells sizes may reduce reconstruction accuracy. Recommendation \(r = max(Δx_{i+1}/Δx_i, Δx_i/Δx_{i+1}) < 3\).
Avoid large ratios \(Δx_{max} / Δx_{min}\), which might cause stiff ODE systems and thus slow, unstable time integration.
Type: double
Range: \([0, L]\)
Length: \(NCOL + 1\)
Group /input/model/unit_XXX/discretization/weno - RECONSTRUCTION = WENO¶
BOUNDARY_MODEL
Boundary model type: 0. Lower WENO order (stable) 1. Zero weights (unstable for small \(D_{\mathrm{ax}}\)) 2. Zero weights for \(p \neq 0\) (less stable)
Type: int
Range: \(\{0, 1, 2\}\)
Length: 1
WENO_EPS
WENO \(\varepsilon\), a small regularization parameter added to the nonlinear weight denominator to prevent division by zero and improve numerical condition for small smoothness indicators. A default value \(1e-10\) often works well.
Type: double
Range: \(>0\)
Length: 1
WENO_ORDER
WENO order, also called WENO \(k\):
Standard upwind scheme (order 1)
WENO 2 (order 3, 2nd order at boundaries)
WENO 3 (order 5, 2nd order at boundaries)
Type: int
Range: \(\{1, 2, 3\}\)
Length: 1
Group /input/model/unit_XXX/discretization/koren - RECONSTRUCTION = KOREN¶
The Koren scheme implemented in CADET intrinsically uses a van Leer flux limiter and has a theoretical convergence order between 1 and 2 depending on the smoothness of the solution.
KOREN_EPS
Sets \(\varepsilon\) in the van Leer flux limiter, a small numerical parameter added to the slope-ratio denominator to prevent division by zero and improve numerical condition when gradients are very small. A default value \(1e-10\) often works well.
Type: double
Range: \(> 0\)
Length: 1