Second-Order Reliability Method (SORM)
Also known as: SORM, Second-order approximation
The Second-Order Reliability Method (SORM) is an extension of FORM that improves failure probability estimates by accounting for the curvature of the limit-state surface at the design point. Introduced by Fiessler, Neumann, and Rackwitz in 1979, SORM provides more accurate approximations for nonlinear failure surfaces while remaining computationally efficient. It has become the standard refinement when FORM accuracy is insufficient.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use SORM when FORM results are suspect due to nonlinearity, or when you need higher accuracy without Monte Carlo verification. SORM is ideal for moderate-to-strong nonlinearity and when the failure surface has significant curvature. It is especially valuable for optimization studies where small changes in design lead to large probability changes. Assume computational budget allows a few dozen more function evaluations than FORM; avoid problems with discontinuous or non-smooth limit-state functions.
Strengths & limitations
- Higher accuracy: curvature terms correct FORM errors for nonlinear limit-state functions, often yielding failure probabilities within 5-10% of Monte Carlo reference.
- Computationally efficient: still much faster than Monte Carlo, requiring only marginal additional cost beyond FORM (mainly Hessian computation).
- Multiple curvature formulas: Breitung, Hohenbichler-Rackwitz, and Tvedt corrections available; users can choose based on problem characteristics.
- Gradient information provided: design point sensitivity remains available for importance ranking of variables.
- Curvature analysis is intensive: computing and analyzing the Hessian is more expensive than FORM, and numerical Hessians can be noisy.
- Local approximation: applies curvature correction only at the design point; does not capture complex surface behavior away from it.
- Multiple failure modes: like FORM, SORM finds only the most probable failure mode; additional critical surfaces must be analyzed separately.
- Assumes smooth surface: requires continuous second derivatives; problems with kinks or discontinuities may yield unreliable curvature estimates.
Frequently asked
When should I use SORM instead of just running FORM and checking with Monte Carlo?
SORM is valuable in two cases: (1) you need an accurate analytical approximation without Monte Carlo (e.g., for a large design study), or (2) you want to understand the nonlinearity contribution numerically. If you can afford Monte Carlo verification, use both: FORM for speed, SORM for understanding the curvature effect, and Monte Carlo for final validation.
What is the 'principal curvature' and how does it affect failure probability?
At the design point, the limit-state surface curves with different radii in different directions. The principal curvatures are the maximum and minimum of these radii. If the surface curves outward (convex), it reduces the failure probability compared to FORM; if inward (concave), it increases it. SORM formulas combine these curvatures to correct the failure probability.
Which SORM correction formula should I use?
Breitung's formula is simplest and works well for moderate nonlinearity. Hohenbichler-Rackwitz is better for intermediate cases. Tvedt's correction handles high nonlinearity better but is more complex. Start with Breitung; if the FORM and SORM results differ by more than 50%, try Tvedt or validate with Monte Carlo.
Can SORM handle multiple failure modes?
No, SORM (like FORM) identifies the single most probable failure mode. If your limit-state function has multiple critical surfaces (e.g., tension and compression failures), you must run separate SORM analyses for each and combine results using system reliability formulas (series/parallel systems).
How do I compute the Hessian matrix if my limit-state function is only given as a black-box simulator?
Use numerical differentiation: finite differences approximate the second derivatives by evaluating the function at perturbed points. This requires O(n^2) function calls (where n is the number of variables) but is simple to implement. Alternatively, use automatic differentiation if your simulator supports it. Be aware that noisy simulators require careful step-size selection to balance truncation and round-off errors.
Sources
- Fiessler, B., Neumann, H. J., & Rackwitz, R. (1979). Quadratic limit states in structural reliability. Journal of the Engineering Mechanics Division, 105(4), 661-676. DOI: 10.1061/jmcea3.0002512 ↗
- Breitung, K. (1984). Asymptotic approximations for multinormal integrals. Journal of Engineering Mechanics, 110(3), 357-366. DOI: 10.1061/(ASCE)0733-9399(1984)110:3(357) ↗
- Hohenbichler, M., & Rackwitz, R. (1988). Improvement of second-order reliability estimates by importance sampling. Journal of Engineering Mechanics, 114(12), 2195-2199. DOI: 10.1061/(ASCE)0733-9399(1988)114:12(2195) ↗
- Melchers, R. E. (2002). Structural Reliability Analysis and Prediction (2nd ed.). John Wiley & Sons. link ↗
How to cite this page
ScholarGate. (2026, June 3). Second-Order Reliability Method (SORM). ScholarGate. https://scholargate.app/en/reliability-engineering/second-order-reliability-method
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- First-Order Reliability MethodReliability Engineering↔ compare
- Highly Accelerated Life TestingReliability Engineering↔ compare
- Rainflow CountingReliability Engineering↔ compare
- Response Surface Desirability FunctionReliability Engineering↔ compare