global_reliability

Global reliability methods

Topics

uncertainty_quantification, reliability_methods

Specification

• Alias: nond_global_reliability

• Arguments: None

Child Keywords:

Required/Optional	Description of Group	Dakota Keyword	Dakota Keyword Description
Optional		initial_samples	Initial number of samples for sampling-based methods
Required (Choose One)	Approximation	x_gaussian_process	Create GP surrogate in x-space
		u_gaussian_process	Create GP surrogate in u-space
Optional (Choose One)	GP Implementation	surfpack	Use the Surfpack version of Gaussian Process surrogates
		dakota	Select the built in Gaussian Process surrogate
		experimental	Use the experimental Gaussian Process surrogate
Optional		import_build_points_file	File containing points you wish to use to build a surrogate
Optional		export_approx_points_file	Output file for surrogate model value evaluations
Optional		use_derivatives	Use derivative data to construct surrogate models
Optional		seed	Seed of the random number generator
Optional		rng	Selection of a random number generator
Optional		response_levels	Values at which to estimate desired statistics for each response
Optional		probability_levels	Specify probability levels at which to estimate the corresponding response value
Optional		gen_reliability_levels	Specify generalized relability levels at which to estimate the corresponding response value
Optional		distribution	Selection of cumulative or complementary cumulative functions
Optional		max_iterations	Number of iterations allowed for optimizers and adaptive UQ methods
Optional		convergence_tolerance	Stopping criterion based on objective function or statistics convergence
Optional		model_pointer	Identifier for model block to be used by a method

Description

These methods do not support forward/inverse mappings involving reliability_levels, since they never form a reliability index based on distance in u-space. Rather they use a Gaussian process model to form an approximation to the limit state (based either in x-space via the x_gaussian_process specification or in u-space via the u_gaussian_process specification), followed by probability estimation based on multimodal adaptive importance sampling (see [BES+07]) and [BES+08]). These probability estimates may then be transformed into generalized reliability levels if desired. At this time, inverse reliability analysis (mapping probability or generalized reliability levels into response levels) is not implemented.

The Gaussian process model approximation to the limit state is formed over the aleatory uncertain variables by default, but may be extended to also capture the effect of design, epistemic uncertain, and state variables. If this is desired, one must use the appropriate controls to specify the active variables in the variables specification block. Refer to Variable Support for additional information on supported variable types.