pystra.reliability.form.FORM

Use it: User guide · Worked example · Theory

class FORM(model, limit_state, *, options=None, on_failure='raise')

Bases: AnalysisObject

Find a FORM design point and approximate the failure probability.

Parameters:

• model (StochasticModel) – Named physical variables and their joint probability model.

• limit_state (LimitState) – Physical response, with negative values indicating failure.

• options (FORMOptions | None) – Iteration, convergence, differentiation and transformation settings.

• on_failure (Literal['raise', 'return']) – On nonconvergence, raise AnalysisError, which carries the unconverged record in .result, or return that record with a warning.

Notes

Call run() to obtain an immutable FORMResult. Check its converged flag before using its probability or design point. Numerical convergence does not establish the accuracy of the local boundary approximation or exclude competing failure regions.

Independent standard normal coordinates are the usual choice. Explicit spherical Student-t generalized Nataf instead uses a Student-t half-space tail. FORMResult.design_index is the signed geometric distance and FORMResult.beta the normal-equivalent index.

See Running FORM and SORM for usage and FORM, SORM and design-point diagnostics for the formulation and references.

Everyday operations

FORM.run()

Execute FORM and return an immutable FORMResult snapshot.

Return type:

FORMResult