A probabilistic reliability technique that refines FORM by including second‑order (quadratic) curvature information of the limit‑state surface at the design point. SORM approximates the local geometry by a quadratic surface (characterized by principal curvatures or the Hessian of the limit‑state in transformed standard‑normal space) and modifies the failure probability estimate to account for the effect of curvature on the Gaussian probability mass near the design point.

Principle

Accounting for principal curvatures of the limit‑state at the most probable point corrects the first‑order half‑space approximation by quantifying how local convexity or concavity concentrates or disperses probability mass, producing a more accurate Pf when curvature is non‑negligible.

Demonstration

Illustrative scenario → The same beam with uncertain load and capacity exhibits a noticeably curved limit‑state near the FORM design point. Recognition → Compute gradients and Hessian of the transformed limit‑state at the design point and determine principal curvatures. Action → Apply a SORM correction formula (e.g., Breitung) to the FORM index and evaluate Pf with curvature terms. Consequence → The SORM estimate adjusts Pf upward or downward relative to FORM depending on curvature signs and magnitudes, yielding improved agreement with sampling methods when curvature is moderate.

Misapplication

Using SORM without ensuring the Hessian is accurately computed, or applying it where the limit‑state is highly non‑quadratic, discontinuous, or where multiple design points contribute appreciably. The semantic error is assuming second‑order correction suffices whenever FORM is imperfect; in strongly nonlinear or multimodal cases it may still be inadequate.

Consequence

When applied to differentiable limit‑states with moderate curvature and a single dominant design point, SORM typically improves Pf estimates and informs on curvature effects; misapplication can produce numerically unstable corrections, misleading confidence in Pf, or worse estimates if curvature is not well represented by a quadratic approximation.

Reversal

If curvature is negligible, SORM offers little improvement over FORM and adds computational cost; if the limit‑state possesses multiple comparable minima or strong non‑quadratic behavior, full simulation or advanced sampling techniques are preferable.

Boundary

Clearly within: smooth, twice‑differentiable limit‑states with moderate curvature and a single dominant design point. Boundary case: higher curvature or mild multimodality where SORM may help but should be validated against simulation. Clearly outside: nondifferentiable, discontinuous, highly nonlinear limit‑states, or problems with multiple separated failure regions.

Semantic Tension

Trade‑off between improved accuracy (by capturing curvature) and increased mathematical and numerical complexity (Hessian evaluation, possible instability); tension between using analytic curvature corrections and resorting to computationally expensive but robust simulation.

Synthesis

SORM extends FORM by using local quadratic geometry to correct the linear approximation; it is an intermediate‑cost refinement that can significantly improve probability estimates for moderate curvature but must be validated and cannot substitute for simulation in strongly nonlinear or multimodal problems.