"""
================================================================
Discretization (:mod:`CADETProcess.processModel.discretization`)
================================================================
The following module is a collection of classes that define different discretization
schemes for :mod:`UnitOperations <CADETProcess.processModel.unitOperation>` for the
CADETProcess software package.
The :class:`~CADETProcess.processModel.DiscretizationParametersBase` class is the base
class for all other classes in this module and defines some common parameters.
Specific parameters for each scheme are defined as attributes of each class.
""" # noqa
from CADETProcess.dataStructure import (
Bool,
Constant,
RangedInteger,
SizedRangedList,
Structure,
Switch,
UnsignedFloat,
UnsignedInteger,
frozen_attributes,
)
__all__ = [
"DiscretizationParametersBase",
"NoDiscretization",
"LRMDiscretizationFV",
"LRMDiscretizationDG",
"LRMPDiscretizationFV",
"LRMPDiscretizationDG",
"GRMDiscretizationFV",
"GRMDiscretizationDG",
"WenoParameters",
"ConsistencySolverParameters",
"DGMixin",
"MCTDiscretizationFV",
]
[docs]
@frozen_attributes
class DiscretizationParametersBase(Structure):
"""
Base class for storing discretization parameters.
Attributes
----------
dimensionality : dict
Dimensionality of the parameters
weno_parameters : WenoParameters
Parameters for the WENO scheme.
consistency_solver: ConsistencySolverParameters
Consistency solver parameters for Cadet.
"""
_dimensionality = []
def __init__(self) -> None:
"""Initialize a new DiscretizationParametersBase instance."""
self.weno_parameters = WenoParameters()
self.consistency_solver = ConsistencySolverParameters()
super().__init__()
@property
def dimensionality(self) -> dict:
"""dict: Dimensionality of the parameters."""
dim = {}
for d in self._dimensionality:
v = getattr(self, d)
if v is None:
continue
dim[d] = v
return dim
@property
def parameters(self) -> dict:
"""dict: Dictionary with parameter values."""
parameters = super().parameters
parameters["weno"] = self.weno_parameters.parameters
parameters["consistency_solver"] = self.consistency_solver.parameters
return parameters
@parameters.setter
def parameters(self, parameters: dict) -> None:
try:
self.weno_parameters.parameters = parameters.pop("weno")
except KeyError:
pass
try:
self.consistency_solver.parameters = parameters.pop("consistency_solver")
except KeyError:
pass
super(DiscretizationParametersBase, self.__class__).parameters.fset(
self, parameters
)
[docs]
class NoDiscretization(DiscretizationParametersBase):
"""Class for unit operations without spatial discretization."""
pass
[docs]
class DGMixin(DiscretizationParametersBase):
"""Mixin DG discretization parameters."""
pass
[docs]
class LRMDiscretizationFV(DiscretizationParametersBase):
"""
Discretization parameters of the FV version of the LRM.
Attributes
----------
ncol : UnsignedInteger, optional
Number of axial column discretization cells. Default is 100.
use_analytic_jacobian : Bool, optional
If True, use analytically computed Jacobian matrix (faster).
If False, use Jacobians generated by algorithmic differentiation (slower).
Default is True.
reconstruction : Switch, optional
Method for spatial reconstruction. Valid values are 'WENO' (Weighted
Essentially Non-Oscillatory). Default is 'WENO'.
"""
spatial_method = Constant(value="FV")
ncol = UnsignedInteger(default=100)
use_analytic_jacobian = Bool(default=True)
reconstruction = Switch(default="WENO", valid=["WENO"])
_parameters = DiscretizationParametersBase._parameters + [
"ncol",
"use_analytic_jacobian",
"reconstruction",
]
_dimensionality = ["ncol"]
[docs]
class LRMDiscretizationDG(DGMixin):
"""
Discretization parameters of the DG version of the LRM.
Attributes
----------
nelem : UnsignedInteger, optional
Number of axial column elements. Default is 16.
use_analytic_jacobian : Bool, optional
If True, use analytically computed Jacobian matrix (faster).
If False, use Jacobians generated by algorithmic differentiation (slower).
Default is True.
polydeg : UnsignedInteger, optional
Degree of the polynomial used for axial discretization. Default is 4.
exact_integration : Bool, optional
Whether to use exact integration for the axial discretization.
Default is False.
See Also
--------
CADETProcess.processModel.LRMPDiscretizationFV
CADETProcess.processModel.LumpedRateModelWithPores
"""
spatial_method = Constant(value="DG")
nelem = UnsignedInteger(default=16)
use_analytic_jacobian = Bool(default=True)
polydeg = UnsignedInteger(default=4)
exact_integration = Bool(default=False)
_parameters = DiscretizationParametersBase._parameters + [
"nelem",
"use_analytic_jacobian",
"polydeg",
"exact_integration",
]
_dimensionality = ["axial_dof"]
@property
def axial_dof(self) -> int:
"""int: Number of degrees of freedom in the axial discretization."""
return self.nelem * (self.polydeg + 1)
[docs]
class LRMPDiscretizationFV(DiscretizationParametersBase):
"""
Discretization parameters of the FV version of the LRMP.
Attributes
----------
ncol : UnsignedInteger, optional
Number of axial column discretization cells. Default is 100.
par_geom : Switch, optional
The geometry of the particles in the model.
Valid values are 'SPHERE', 'CYLINDER', and 'SLAB'.
Default is 'SPHERE'.
use_analytic_jacobian : Bool, optional
If True, use analytically computed Jacobian matrix (faster).
If False, use Jacobians generated by algorithmic differentiation (slower).
Default is True.
reconstruction : Switch, optional
Method for spatial reconstruction. Valid values are 'WENO' (Weighted
Essentially Non-Oscillatory). Default is 'WENO'.
gs_type : Bool, optional
Type of Gram-Schmidt orthogonalization.
If 0, use classical Gram-Schmidt.
If 1, use modified Gram-Schmidt.
The default is 1.
max_krylov : UnsignedInteger, optional
Size of the Krylov subspace in the iterative linear GMRES solver.
If 0, max_krylov = NCOL * NCOMP * NPARTYPE is used.
The default is 0.
max_restarts : UnsignedInteger, optional
Maximum number of restarts to use for the GMRES method. Default is 10.
schur_safety : UnsignedFloat, optional
Safety factor for the Schur complement solver. Default is 1.0e-8.
See Also
--------
CADETProcess.processModel.LRMPDiscretizationDG
CADETProcess.processModel.LumpedRateModelWithPores
"""
spatial_method = Constant(value="FV")
ncol = UnsignedInteger(default=100)
par_geom = Switch(default="SPHERE", valid=["SPHERE", "CYLINDER", "SLAB"])
use_analytic_jacobian = Bool(default=True)
reconstruction = Switch(default="WENO", valid=["WENO"])
gs_type = Bool(default=True)
max_krylov = UnsignedInteger(default=0)
max_restarts = UnsignedInteger(default=10)
schur_safety = UnsignedFloat(default=1.0e-8)
_parameters = DiscretizationParametersBase._parameters + [
"ncol",
"par_geom",
"use_analytic_jacobian",
"reconstruction",
"gs_type",
"max_krylov",
"max_restarts",
"schur_safety",
]
_dimensionality = ["ncol"]
[docs]
class LRMPDiscretizationDG(DGMixin):
"""
Discretization parameters of the DG version of the LRMP.
Attributes
----------
nelem : UnsignedInteger, optional
Number of axial column elements. Default is 16.
par_geom : Switch, optional
The geometry of the particles in the model.
Valid values are 'SPHERE', 'CYLINDER', and 'SLAB'.
Default is 'SPHERE'.
use_analytic_jacobian : Bool, optional
If True, use analytically computed Jacobian matrix (faster).
If False, use Jacobians generated by algorithmic differentiation (slower).
Default is True.
polydeg : UnsignedInteger, optional
Degree of the polynomial used for spatial discretization. Default is 4.
exact_integration : Bool, optional
Whether to use exact integration for the spatial discretization.
Default is False.
See Also
--------
CADETProcess.processModel.LRMPDiscretizationFV
CADETProcess.processModel.LumpedRateModelWithPores
"""
spatial_method = Constant(value="DG")
nelem = UnsignedInteger(default=16)
par_geom = Switch(default="SPHERE", valid=["SPHERE", "CYLINDER", "SLAB"])
use_analytic_jacobian = Bool(default=True)
polydeg = UnsignedInteger(default=4)
exact_integration = Bool(default=False)
_parameters = DiscretizationParametersBase._parameters + [
"nelem",
"par_geom",
"use_analytic_jacobian",
"polydeg",
"exact_integration",
]
_dimensionality = ["axial_dof"]
@property
def axial_dof(self) -> int:
"""int: Number of axial degrees of freedom in the spatial discretization."""
return self.nelem * (self.polydeg + 1)
[docs]
class GRMDiscretizationFV(DiscretizationParametersBase):
"""
Discretization parameters of the FV version of the LRMP.
Attributes
----------
ncol : UnsignedInteger, optional
Number of axial column discretization cells. Default is 100.
npar : UnsignedInteger, optional
Number of discretization cells in the radial direction. Default is 5.
par_geom : Switch, optional
The geometry of the particles in the model.
Valid values are 'SPHERE', 'CYLINDER', and 'SLAB'.
Default is 'SPHERE'.
par_disc_type : Switch, optional
Discretization scheme inside the particles for all or each particle type.
Valid values are 'EQUIDISTANT_PAR', 'EQUIVOLUME_PAR', and 'USER_DEFINED_PAR'.
Default is 'EQUIDISTANT_PAR'.
par_disc_vector : SizedRangedList, optional
Node coordinates for the cell boundaries
(ignored if `par_disc_type` != `USER_DEFINED_PAR).
The coordinates are relative and have to include the endpoints `0` and `1`.
They are later linearly mapped to the true radial range.
The coordinates for each particle type are appended to one long vector in
type-major ordering.
Default is a uniformly spaced vector with `npar+1` points between 0 and 1.
par_boundary_order : RangedInteger, optional
The order of the boundary scheme used to discretize the particles.
Valid values are 1 (first order) and 2 (second order).
Default is 2.
use_analytic_jacobian : Bool, optional
If True, use analytically computed Jacobian matrix (faster).
If False, use Jacobians generated by algorithmic differentiation (slower).
Default is True.
reconstruction : Switch, optional
Method for spatial reconstruction. Valid values are 'WENO' (Weighted
Essentially Non-Oscillatory). Default is 'WENO'.
gs_type : Bool, optional
Type of Gram-Schmidt orthogonalization.
If 0, use classical Gram-Schmidt.
If 1, use modified Gram-Schmidt.
The default is 1.
max_krylov : UnsignedInteger, optional
Size of the Krylov subspace in the iterative linear GMRES solver.
If 0, max_krylov = ncol * ncomp * npar is used, where ncomp is the
number of components in the model.
Default is 0.
max_restarts : UnsignedInteger, optional
Maximum number of restarts to use for the GMRES method. Default is 10.
schur_safety : UnsignedFloat, optional
Safety factor for the Schur complement solver. Default is 1.0e-8.
fix_zero_surface_diffusion : Bool, optional
If True, fix the surface diffusion coefficient of particles with zero
surface diffusion to a small positive value. Default is False.
See Also
--------
CADETProcess.processModel.LRMPDiscretizationDG
CADETProcess.processModel.LumpedRateModelWithPores
"""
spatial_method = Constant(value="FV")
ncol = UnsignedInteger(default=100)
npar = UnsignedInteger(default=5)
par_geom = Switch(default="SPHERE", valid=["SPHERE", "CYLINDER", "SLAB"])
par_disc_type = Switch(
default="EQUIDISTANT_PAR",
valid=["EQUIDISTANT_PAR", "EQUIVOLUME_PAR", "USER_DEFINED_PAR"],
)
par_disc_vector = SizedRangedList(
lb=0, ub=1, size="par_disc_vector_length", is_optional=True
)
par_boundary_order = RangedInteger(lb=1, ub=2, default=2)
use_analytic_jacobian = Bool(default=True)
reconstruction = Switch(default="WENO", valid=["WENO"])
gs_type = Bool(default=True)
max_krylov = UnsignedInteger(default=0)
max_restarts = UnsignedInteger(default=10)
schur_safety = UnsignedFloat(default=1.0e-8)
fix_zero_surface_diffusion = Bool(default=False)
_parameters = DiscretizationParametersBase._parameters + [
"ncol",
"npar",
"par_geom",
"par_disc_type",
"par_disc_vector",
"par_boundary_order",
"use_analytic_jacobian",
"reconstruction",
"gs_type",
"max_krylov",
"max_restarts",
"schur_safety",
"fix_zero_surface_diffusion",
]
_required_parameters = [
"ncol",
"npar",
"par_geom",
"par_disc_type",
"par_boundary_order",
"use_analytic_jacobian",
"reconstruction",
"gs_type",
"max_krylov",
"max_restarts",
"schur_safety",
"fix_zero_surface_diffusion",
]
_dimensionality = ["ncol", "npar"]
@property
def par_disc_vector_length(self) -> int:
"""int: Number of entries in the particle discretization vector."""
return self.npar + 1
[docs]
class GRMDiscretizationDG(DGMixin):
"""
Discretization parameters of the DG version of the GRM.
Attributes
----------
nelem : UnsignedInteger, optional
Number of axial column elements. Default is 16.
par_nelem : UnsignedInteger, optional
Number of particle (radial) discretization cells for each particle type.
Default is 1.
par_geom : Switch, optional
The geometry of the particles in the model.
Valid values are 'SPHERE', 'CYLINDER', and 'SLAB'.
Default is 'SPHERE'.
use_analytic_jacobian : Bool, optional
If True, use analytically computed Jacobian matrix (faster).
If False, use Jacobians generated by algorithmic differentiation (slower).
Default is True.
polydeg : UnsignedInteger, optional
Degree of the polynomial used for axial discretization. Default is 4.
par_polydeg : UnsignedInteger, optional
Degree of the polynomial used for particle radial discretization. Default is 4.
exact_integration : Bool, optional
Whether to use exact integration for the axial discretization.
Default is False.
par_exact_integration : Bool, optional
Whether to use exact integration for the particle radial discretization.
Default is True.
fix_zero_surface_diffusion : Bool, optional
Whether to fix zero surface diffusion for particles. Default is False.
If True, the parameters must not become non-zero during this or subsequent
simulation runs. The internal data structures are optimized for a more efficient
simulation.
Default is False (optimization disabled in favor of flexibility).
See Also
--------
CADETProcess.processModel.GRMDiscretizationDG
CADETProcess.processModel.GeneralRateModel
"""
spatial_method = Constant(value="DG")
nelem = UnsignedInteger(default=16)
par_nelem = UnsignedInteger(default=1)
par_geom = Switch(default="SPHERE", valid=["SPHERE", "CYLINDER", "SLAB"])
par_disc_type = Switch(
default="EQUIDISTANT_PAR",
valid=["EQUIDISTANT_PAR", "EQUIVOLUME_PAR", "USER_DEFINED_PAR"],
)
par_disc_vector = SizedRangedList(lb=0, ub=1, size="par_disc_vector_length")
par_boundary_order = RangedInteger(lb=1, ub=2, default=2)
use_analytic_jacobian = Bool(default=True)
polydeg = UnsignedInteger(default=4)
par_polydeg = UnsignedInteger(default=4)
exact_integration = Bool(default=False)
par_exact_integration = Bool(default=True)
fix_zero_surface_diffusion = Bool(default=False)
_parameters = DiscretizationParametersBase._parameters + [
"nelem",
"par_nelem",
"par_geom",
"par_disc_type",
"par_disc_vector",
"par_boundary_order",
"use_analytic_jacobian",
"polydeg",
"par_polydeg",
"exact_integration",
"par_exact_integration",
"fix_zero_surface_diffusion",
]
_dimensionality = ["axial_dof", "par_dof"]
@property
def axial_dof(self) -> int:
"""int: Number of axial degrees of freedom in the axial discretization."""
return self.nelem * (self.polydeg + 1)
@property
def par_dof(self) -> int:
"""int: Number of particle degrees of freedom in the axial discretization."""
return self.axial_dof * self.par_disc_vector_length
@property
def par_disc_vector_length(self) -> int:
"""int: Number of entries in the particle discretization vector."""
return self.par_nelem * (self.par_polydeg + 1)
[docs]
@frozen_attributes
class WenoParameters(Structure):
"""
Discretization parameters for the WENO scheme.
Attributes
----------
boundary_model : Switch, optional
Specifies the method for dealing with boundary cells.
Valid values are:
0: Lower WENO order (stable)
1: Zero weights (unstable for small `D_ax`)
2: Zero weights for p != 0 (stable?)
3: Large ghost points
Default is 0.
weno_eps : UnsignedFloat, optional
A small positive number used to avoid division by zero in the WENO scheme.
Default is 1e-10.
weno_order : UnsignedInteger, optional
Order of the WENO scheme. Valid values are:
1: Standard upwind scheme (order 1)
2: WENO 2 (order 3)
3: WENO 3 (order 5)
Default is 3.
See Also
--------
Structure
"""
boundary_model = UnsignedInteger(default=0, ub=3)
weno_eps = UnsignedFloat(default=1e-10)
weno_order = UnsignedInteger(default=3, ub=3)
_parameters = ["boundary_model", "weno_eps", "weno_order"]
[docs]
@frozen_attributes
class ConsistencySolverParameters(Structure):
"""
A class for defining the consistency solver parameters for Cadet.
Parameters
----------
solver_name : Switch, optional
Name of the solver.
Valid values are 'LEVMAR', 'ATRN_RES', 'ATRN_ERR', and 'COMPOSITE'.
The default is 'LEVMAR'
init_damping : UnsignedFloat, optional
The initial damping parameter. Default is 0.01.
min_damping : UnsignedFloat, optional
The minimum damping parameter. Default is 0.0001.
max_iterations : UnsignedFloat, optional
The maximum number of iterations. Default is 50.
subsolvers : Switch, optional
Vector with names of solvers for the composite solver
(only required for composite solver).
Valid values are 'LEVMAR', 'ATRN_RES', 'ATRN_ERR', and 'COMPOSITE'.
The default is 'LEVMAR'
See Also
--------
Structure
"""
solver_name = Switch(
default="LEVMAR", valid=["LEVMAR", "ATRN_RES", "ARTN_ERR", "COMPOSITE"]
)
init_damping = UnsignedFloat(default=0.01)
min_damping = UnsignedFloat(default=0.0001)
max_iterations = UnsignedInteger(default=50)
subsolvers = Switch(default="LEVMAR", valid=["LEVMAR", "ATRN_RES", "ARTN_ERR"])
_parameters = [
"solver_name",
"init_damping",
"min_damping",
"max_iterations",
"subsolvers",
]
[docs]
class MCTDiscretizationFV(DiscretizationParametersBase):
"""
Discretization parameters of the FV version of the MCT.
Attributes
----------
ncol : UnsignedInteger, optional
Number of axial column discretization cells. Default is 100.
use_analytic_jacobian : Bool, optional
If True, use analytically computed Jacobian matrix (faster).
If False, use Jacobians generated by algorithmic differentiation (slower).
Default is True.
reconstruction : Switch, optional
Method for spatial reconstruction. Valid values are 'WENO' (Weighted
Essentially Non-Oscillatory). Default is 'WENO'.
"""
spatial_method = Constant(value="FV")
ncol = UnsignedInteger(default=100)
use_analytic_jacobian = Bool(default=True)
reconstruction = Switch(default="WENO", valid=["WENO"])
_parameters = ["ncol", "use_analytic_jacobian", "reconstruction"]
_dimensionality = ["ncol"]
