cosapp.drivers.nonlinearsolver

Classes

BaseSolverBuilder(solver)	Base interface for numerical system building strategy used by NonLinearSolver.
MultipointSolverBuilder(solver, points)	System building strategy for solvers with one or more RunSingleCase sub-drivers.
NonLinearSolver(name[, owner, method])	Solve mathematical problem with algebraic variables.
StandaloneSolverBuilder(solver)	System building strategy for solvers with no RunSingleCase sub-drivers.

class cosapp.drivers.nonlinearsolver.BaseSolverBuilder(solver: AbstractSolver)

Bases: ABC

Base interface for numerical system building strategy used by NonLinearSolver. Implementations will differ for single- or multi-point design.

abstract build_system() → None

abstract update_residues() → None

update_unknowns(x: ndarray) → None

class cosapp.drivers.nonlinearsolver.MultipointSolverBuilder(solver: NonLinearSolver, points: List[RunSingleCase])

Bases: BaseSolverBuilder

System building strategy for solvers with one or more RunSingleCase sub-drivers. In this case, the actual mathematical problem is a combination of unknowns and equations declared at solver level, and of local design and off-design problems declared at case level.

build_system()

update_residues()

class cosapp.drivers.nonlinearsolver.NonLinearSolver(name: str, owner: cosapp.systems.System | None = None, method: NonLinearMethods | str = NonLinearMethods.NR, **kwargs)

Bases: AbstractSolver

Solve mathematical problem with algebraic variables.

compute_jacobian

Should the Jacobian matrix be computed? ; default True

Type:

bool

jac_lup

LU decomposition of latest Jacobian matrix (if available, None otherwise)

Type:

ndarray, optional

jac

Latest Jacobian matrix computed (if available, None otherwise)

Type:

ndarray, optional

add_child(child: AnyDriver, execution_index: int | None = None, desc='') → AnyDriver

Add a child Driver to the current Driver.

When adding a child Driver, it is possible to specified its position in the execution order.

Child Port, inwards and outwards can also be pulled at the parent level by providing either the name of the port/inward/outward or a list of them or the name mapping of the child element (dictionary keys) to the parent element (dictionary values). If the argument is not a dictionary, the name in the parent system will be the same as in the child.

Parameters:

• child (-) – Driver to add to the current Driver

• execution_index (-) – Index of the execution order list at which the Module should be inserted; default latest.

• [str (- desc) – Sub-driver description in the context of its parent driver.

• optional] – Sub-driver description in the context of its parent driver.

Notes

The added child will have its owner set to match the one of the current driver.

add_equation(equation: str | Iterable[dict | str], *args, **kwargs) → MathematicalProblem

Add off-design equation(s).

More details in cosapp.core.MathematicalProblem.add_equation.

Parameters:

• rhs'] (- equation [str or Iterable of str of the kind 'lhs ==) – Equation or collection of equations to be added.

• *args (-) –

  Additional arguments forwarded to MathematicalProblem.add_equation.

• **kwargs – Additional arguments forwarded to MathematicalProblem.add_equation.

Returns:

The updated problem.

Return type:

• MathematicalProblem

add_target(expression: str | Iterable[str], *args, **kwargs) → MathematicalProblem

Add deferred off-design equation(s).

More details in cosapp.core.MathematicalProblem.add_target.

Parameters:

• expression (-) – Targetted expression

• *args (-) –

• **kwargs (Forwarded to MathematicalProblem.add_target)

Returns:

The modified mathematical problem

Return type:

MathematicalProblem

add_unknown(name: str | Iterable[dict | str], *args, **kwargs) → MathematicalProblem

Add design unknown(s).

More details in cosapp.core.MathematicalProblem.add_unknown.

Parameters:

• str] (- name [str or Iterable of dictionary or) – Name of the variable or collection of variables to be added.

• *args (-) –

  Additional arguments forwarded to MathematicalProblem.add_unknown.

• **kwargs – Additional arguments forwarded to MathematicalProblem.add_unknown.

Returns:

The updated problem.

Return type:

• MathematicalProblem

compute() → None

Run the resolution method to find free vars values that zero out residues

compute_jacobian: bool

extend(problem: MathematicalProblem, *args, **kwargs) → MathematicalProblem

Extend solver inner problem.

Parameters:

• [MathematicalProblem] (- problem) – Source mathematical problem.

• *args (-) –

  Additional arguments forwarded to MathematicalProblem.extend.

• **kwargs – Additional arguments forwarded to MathematicalProblem.extend.

Returns:

The extended problem.

Return type:

• MathematicalProblem

is_standalone() → bool

Is this Driver able to solve a system?

Returns:

Ability to solve a system or not.

Return type:

bool

jac: numpy.ndarray | None

jac_lup: Tuple[numpy.ndarray | None, numpy.ndarray | None]

log_debug_message(handler: HandlerWithContextFilters, record: LogRecord, format: LogFormat = LogFormat.RAW) → bool

Callback method on the driver to log more detailed information.

This method will be called by the log handler when log_context() is active if the logging level is lower or equals to VERBOSE_LEVEL. It allows the object to send additional log message to help debugging a simulation.

Parameters:

• handler (HandlerWithContextFilters) – Log handler on which additional message should be published.

• record (logging.LogRecord) – Log record

• format (LogFormat) – Format of the message

Returns:

Should the provided record be logged?

Return type:

bool

property method: NonLinearMethods

Selected solver algorithm.

Type:

NonLinearMethods

reset_problem() → None

Reset mathematical problem

resolution_method(fresidues: Callable[[Sequence[float], float | str, bool], ndarray], x0: Sequence[float], args: Tuple[float | str] = (), options: Dict[str, Any] | None = None) → SolverResults

Function call to cancel the residues.

Parameters:

• fresidues (Callable[[Sequence[float], Union[float, str]], numpy.ndarray]) – Residues function taking two parameters (evaluation vector, time/ref) and returning the residues

• x0 (Sequence[float]) – The initial values vector to converge to the solution

• args (Tuple[Union[float, str], bool], optional) – A tuple of additional argument for fresidues starting with the time/ref parameter and the need to update residues reference

• options (OptionsDictionary, optional) – Options for the numerical resolution method

Returns:

Solution container

Return type:

SolverResults

property results: SolverResults

structure containing solver results, together with additional detail.

Type:

SolverResults

set_iteratives(x: Sequence[float]) → None

setup_run() → None

Set up the mathematical problem.

class cosapp.drivers.nonlinearsolver.StandaloneSolverBuilder(solver: AbstractSolver)

Bases: BaseSolverBuilder

System building strategy for solvers with no RunSingleCase sub-drivers. In this case, the actual mathematical problem is entirely defined by unknowns and equations declared at solver level.

build_system()

update_residues()