windse.OptimizationManager

The OptimizationManager submodule contains all the required function for optimizing via dolfin-adjoint. To use dolfin-adjoin set:

general:
    dolfin_adjoint: True

in the param.yaml file.

Todo

• Read through an update the docstrings for these functions.
• Create specific optimization classes.

class windse.OptimizationManager.ConsComp(**kwargs)

Bases: openmdao.core.explicitcomponent.ExplicitComponent

OpenMDAO component to wrap the constraint computation.

A small wrapper used on the fenics methods for computing constraint and Jacobian values using the OpenMDAO syntax.

compute(inputs, outputs)

Compute outputs given inputs. The model is assumed to be in an unscaled state.

Parameters:

• inputs (Vector) – unscaled, dimensional input variables read via inputs[key]
• outputs (Vector) – unscaled, dimensional output variables read via outputs[key]
• discrete_inputs (dict or None) – If not None, dict containing discrete input values.
• discrete_outputs (dict or None) – If not None, dict containing discrete output values.

compute_partials(inputs, partials)

Compute sub-jacobian parts. The model is assumed to be in an unscaled state.

Parameters:

• inputs (Vector) – unscaled, dimensional input variables read via inputs[key]
• partials (Jacobian) – sub-jac components written to partials[output_name, input_name]
• discrete_inputs (dict or None) – If not None, dict containing discrete input values.

initialize()

Perform any one-time initialization run at instantiation.

setup()

Declare inputs and outputs.

Available attributes:

name pathname comm options

class windse.OptimizationManager.ObjComp(**kwargs)

Bases: openmdao.core.explicitcomponent.ExplicitComponent

OpenMDAO component to wrap the objective computation from dolfin.

Specifically, we use the J and dJ (function and Jacobian) methods to compute the function value and derivative values as needed by the OpenMDAO optimizers.

compute(inputs, outputs)

Compute outputs given inputs. The model is assumed to be in an unscaled state.

Parameters:

• inputs (Vector) – unscaled, dimensional input variables read via inputs[key]
• outputs (Vector) – unscaled, dimensional output variables read via outputs[key]
• discrete_inputs (dict or None) – If not None, dict containing discrete input values.
• discrete_outputs (dict or None) – If not None, dict containing discrete output values.

compute_partials(inputs, partials)

Compute sub-jacobian parts. The model is assumed to be in an unscaled state.

Parameters:

• inputs (Vector) – unscaled, dimensional input variables read via inputs[key]
• partials (Jacobian) – sub-jac components written to partials[output_name, input_name]
• discrete_inputs (dict or None) – If not None, dict containing discrete input values.

initialize()

Perform any one-time initialization run at instantiation.

setup()

Declare inputs and outputs.

Available attributes:

name pathname comm options

class windse.OptimizationManager.Optimizer(solver)

Bases: object

A GenericProblem contains on the basic functions required by all problem objects.

Parameters:dom (windse.DomainManager.GenericDomain()) – a windse domain object.

Gradient()

Returns a gradient of the objective function

windse.OptimizationManager.gather(m)

Helper function to gather constraint Jacobians. Adapated from fenics.

windse.OptimizationManager.om_wrapper(J, initial_DVs, dJ, H, bounds, **kwargs)

Custom optimization wrapper to use OpenMDAO optimizers with dolfin-adjoint.

Follows the API as defined by dolfin-adjoint.

Parameters:

• J (object) – Function to compute the model analysis value at a design point.
• initial_DVs (array) – The initial design variables so we can get the array sizing correct for the OpenMDAO implementation.
• dJ (object) – Function to compute the Jacobian at a design point.
• H (object) – Function to compute the Hessian at a design point (not used).
• bounds (array) – Array of lower and upper bound values for the design variables.

Returns:

DVs – The optimal design variable values.

Return type:

array

Classes

class windse.OptimizationManager.ConsComp(**kwargs)

Bases: openmdao.core.explicitcomponent.ExplicitComponent

OpenMDAO component to wrap the constraint computation.

A small wrapper used on the fenics methods for computing constraint and Jacobian values using the OpenMDAO syntax.

compute(inputs, outputs)

Compute outputs given inputs. The model is assumed to be in an unscaled state.

Parameters:

• inputs (Vector) – unscaled, dimensional input variables read via inputs[key]
• outputs (Vector) – unscaled, dimensional output variables read via outputs[key]
• discrete_inputs (dict or None) – If not None, dict containing discrete input values.
• discrete_outputs (dict or None) – If not None, dict containing discrete output values.

compute_partials(inputs, partials)

Compute sub-jacobian parts. The model is assumed to be in an unscaled state.

Parameters:

• inputs (Vector) – unscaled, dimensional input variables read via inputs[key]
• partials (Jacobian) – sub-jac components written to partials[output_name, input_name]
• discrete_inputs (dict or None) – If not None, dict containing discrete input values.

initialize()

Perform any one-time initialization run at instantiation.

setup()

Declare inputs and outputs.

Available attributes:

name pathname comm options

class windse.OptimizationManager.MinimumDistanceConstraint(m_pos, min_distance=200)

Bases: object

class windse.OptimizationManager.ObjComp(**kwargs)

Bases: openmdao.core.explicitcomponent.ExplicitComponent

OpenMDAO component to wrap the objective computation from dolfin.

Specifically, we use the J and dJ (function and Jacobian) methods to compute the function value and derivative values as needed by the OpenMDAO optimizers.

compute(inputs, outputs)

Compute outputs given inputs. The model is assumed to be in an unscaled state.

Parameters:

• inputs (Vector) – unscaled, dimensional input variables read via inputs[key]
• outputs (Vector) – unscaled, dimensional output variables read via outputs[key]
• discrete_inputs (dict or None) – If not None, dict containing discrete input values.
• discrete_outputs (dict or None) – If not None, dict containing discrete output values.

compute_partials(inputs, partials)

Compute sub-jacobian parts. The model is assumed to be in an unscaled state.

Parameters:

• inputs (Vector) – unscaled, dimensional input variables read via inputs[key]
• partials (Jacobian) – sub-jac components written to partials[output_name, input_name]
• discrete_inputs (dict or None) – If not None, dict containing discrete input values.

initialize()

Perform any one-time initialization run at instantiation.

setup()

Declare inputs and outputs.

Available attributes:

name pathname comm options

class windse.OptimizationManager.Optimizer(solver)

Bases: object

A GenericProblem contains on the basic functions required by all problem objects.

Parameters:dom (windse.DomainManager.GenericDomain()) – a windse domain object.

Gradient()

Returns a gradient of the objective function

Functions

windse.OptimizationManager.gather(m)

Helper function to gather constraint Jacobians. Adapated from fenics.

windse.OptimizationManager.om_wrapper(J, initial_DVs, dJ, H, bounds, **kwargs)

Custom optimization wrapper to use OpenMDAO optimizers with dolfin-adjoint.

Follows the API as defined by dolfin-adjoint.

Parameters:

• J (object) – Function to compute the model analysis value at a design point.
• initial_DVs (array) – The initial design variables so we can get the array sizing correct for the OpenMDAO implementation.
• dJ (object) – Function to compute the Jacobian at a design point.
• H (object) – Function to compute the Hessian at a design point (not used).
• bounds (array) – Array of lower and upper bound values for the design variables.

Returns:

DVs – The optimal design variable values.

Return type:

array