pystra.results#

Use it: User guide · Worked example · Theory

Immutable records returned by every reliability analysis.

Each analysis’s run() returns one of these frozen records. A record is a snapshot: rerunning the analysis, or changing its model, does not alter it. Vector fields are read-only NumPy arrays in variable_names order, and mapping fields are read-only. Records compare equal when their values are equal.

Every record has method, status, message, n_limit_state_evaluations, variable_names and options, the converged property and summary(). Records of probability estimates add failure_probability and beta, the normal-equivalent reliability index \(-\Phi^{-1}(p_f)\).

status is one of:

"converged"

A design-point method (FORM, SORM, system FORM, sensitivity) met its convergence criteria.

"not_converged"

It did not, or SORM’s formula is undefined at the fitted curvatures. The estimate is unavailable or unreliable. Unless constructed with on_failure="return", the analysis raises AnalysisError carrying the record.

"completed"

A simulation or diagnostic ran to completion.

"precision_not_met"

A simulation used its whole sample budget before reaching its target coefficient of variation. The estimate is still reported.

n_limit_state_evaluations counts the limit-state evaluations made by the method itself. A nested FORM record (form, component_results or the form diagnostic) reports that FORM analysis’s own count; system FORM and sensitivity analysis, which consist of FORM runs, report their total. options holds the frozen settings the analysis used.

Classes

DistributionAnalysisResult

Record of a distribution analysis: Monte Carlo samples of the model.

FORMResult

Record of a FORM analysis.

SORMResult

Record of a SORM analysis.

SensitivityResult

Record of a sensitivity analysis of the FORM reliability index.

SimulationResult

Record of a simulation estimate.

StrongMaximumResult

Record of a Strong Maximum Test, a diagnostic of a FORM design point.

SystemFORMResult

Record of a system FORM approximation.