pith. sign in

arxiv: 2605.23631 · v1 · pith:3OEHCL42new · submitted 2026-05-22 · 📊 stat.CO

Directional subset simulation method for reliability analysis

Oindrila Kanjilal (TUM) , Julien Bect (L2S , RT-UQ) This is my paper

Pith reviewed 2026-05-25 02:31 UTC · model grok-4.3

classification 📊 stat.CO
keywords subset simulationreliability analysisrare event estimationMarkov chain Monte Carlodirectional samplingmulti-modal failure domainsfailure probability
0
0 comments X

The pith

Directional subset simulation selects intermediate domains to keep Markov chain samples active in multiple directions, avoiding trapping for multi-modal failure regions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Standard subset simulation estimates rare failure probabilities by breaking them into a chain of larger conditional probabilities, but its Markov chain sampling can become trapped in one mode when the failure domain has several disconnected regions. The directional subset simulation method changes how the intermediate domains are chosen at each level, drawing on directional sampling to ensure that samples remain spread across different directions in parameter space. This preserves the ability of the chains to reach all relevant failure modes without extra tuning. A sympathetic reader cares because many physical systems exhibit multiple distinct ways to fail, and inaccurate probability estimates can affect safety and design decisions.

Core claim

The dSS method evaluates the small failure probability as a product of conditional probabilities by sampling a sequence of nested sub-domains, but replaces the standard choice of those sub-domains with a directional selection rule that preserves samples in several directions of the parameter space at each intermediate level, thereby preventing the Markov chains from becoming confined to a single mode.

What carries the argument

Directional selection of intermediate failure domains, which uses concepts from directional sampling to propagate samples toward failure while maintaining coverage across multiple directions.

If this is right

  • Failure probability estimates become reliable for systems whose failure domains contain multiple separated modes.
  • The Markov chain samples at each level of the hierarchy explore a wider portion of the relevant parameter space.
  • The overall computational structure of subset simulation remains unchanged while its range of applicability increases.
  • Numerical examples confirm that the method handles cases where standard subset simulation produces inaccurate results.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same directional selection idea might be transferred to other adaptive Monte Carlo schemes that encounter mode-trapping in multimodal settings.
  • Combining the directional rule with existing variance-reduction techniques could further lower the variance of the probability estimator.
  • Scalability tests on higher-dimensional engineering models would reveal whether the directional overhead remains modest as dimension grows.

Load-bearing premise

The directional selection of intermediate domains can be performed without introducing bias into the conditional probability estimates or requiring problem-specific tuning that limits general applicability.

What would settle it

Apply both standard subset simulation and dSS to a known multi-modal test case with an independently computable exact failure probability, then check whether dSS recovers the correct value while standard SS underestimates due to trapping.

Figures

Figures reproduced from arXiv: 2605.23631 by Julien Bect (L2S, Oindrila Kanjilal (TUM), RT-UQ).

Figure 1
Figure 1. Figure 1: Contour plot of the piecewise linear function in Example 4.1. The blue line [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Progression of samples in a typical run of (a) SS and (b–d) Case 1–3 of the [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Histogram of the log10 𝜋ˆ𝐹 estimates for the piecewise linear LSF in Example 4.1 obtained from 104 independent runs of the SS and Case 1 of the dSS methods, with (a) 𝑁 = 500 and (b) 𝑁 = 4000 samples per level and level probability of 𝜌 = 0.2. The red line shows the reference value log10 𝜋ˆ𝐹,ref [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Accuracy of the algorithms, as measured by the metric [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Progression of samples in a typical run of (a) SS and (b) Case 1 of the dSS [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
read the original abstract

Estimating the probabilities of rare failure events is a key challenge in the reliability analysis of physical systems. Subset simulation (SS) is a very popular adaptive Monte Carlo method for this problem. In SS, the small failure probability is evaluated as a product of larger conditional probabilities by iteratively sampling a sequence of nested sub-domains of the parameter space, encompassing the target failure domain of interest, using Markov chain Monte Carlo methods. For failure domains with multiple modes, the Markov chain samples used to explore the intermediate levels of SS can be trapped in a confined region of the input parameter space, leading to inaccurate failure probability estimates. In this contribution, we propose the directional subset simulation (dSS) method for this problem, which uses concepts from directional sampling to informedly propagate samples towards failure. This is accomplished through a novel selection of the intermediate failure domains, which preserves samples in several directions in the parameter space in each intermediate level. The merits of the dSS method are illustrated through a selection of numerical examples.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The paper proposes the directional subset simulation (dSS) method as an extension of standard subset simulation (SS) for estimating small failure probabilities. Standard SS uses MCMC to sample nested intermediate failure domains but can trap chains in local modes for multi-modal failure regions. dSS incorporates directional sampling ideas to select intermediate domains that retain samples across multiple directions in parameter space at each level, with the goal of maintaining an unbiased product-of-conditional-probabilities estimator while improving exploration.

Significance. If the directional selection rule preserves the required conditional probabilities without bias or problem-specific tuning, dSS would provide a targeted fix for a documented limitation of SS in multi-modal settings, with direct applicability to reliability analysis of engineering systems. The manuscript supplies numerical examples as empirical support; explicit credit is due for framing the modification as preserving the estimator by algorithmic construction rather than ad-hoc adjustment.

minor comments (3)
  1. The description of how the directional intermediate domains are constructed (e.g., the precise criterion for retaining directional diversity while keeping the sets nested) would benefit from an explicit algorithmic pseudocode block or numbered steps to allow direct reproduction.
  2. Numerical examples should report the effective sample size or chain mixing diagnostics alongside the failure probability estimates to substantiate that the directional selection indeed mitigates trapping relative to standard SS.
  3. A short discussion of computational overhead (extra directional evaluations per level) relative to the accuracy gain would help readers assess practical trade-offs.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of the directional subset simulation (dSS) method, recognition of its potential to address multi-modal failure domains in subset simulation, and recommendation for minor revision. We appreciate the acknowledgment that the approach is framed as preserving the unbiased product-of-conditional-probabilities estimator through algorithmic construction.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces an algorithmic variant of subset simulation (dSS) that redefines intermediate failure domains using directional sampling concepts to mitigate MCMC trapping in multi-modal cases. No derivation chain, equations, or fitted parameters are presented that reduce by construction to prior inputs or self-citations. The unbiasedness of the product-of-conditionals estimator is asserted to hold by the algorithmic construction itself, with numerical examples provided as external validation rather than internal fitting. No load-bearing self-citations, uniqueness theorems, or ansatzes are invoked. The central claim remains an independent algorithmic proposal.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only; no explicit free parameters, axioms, or invented entities are stated in the provided text.

pith-pipeline@v0.9.0 · 5709 in / 974 out tokens · 26956 ms · 2026-05-25T02:31:25.985886+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

  1. [1]

    S. K. Au and J. L. Beck. Estimation of small failure probabilities in high dimensions by subset simulation. Probabilistic Engineering Mechanics, 16(4): 263–277, 2001

    work page 2001

  2. [2]

    Breitung

    K. Breitung. 40 years FORM: Some new aspects? Probabilistic Engineering Mechanics, 42:71–77, 2015

    work page 2015

  3. [3]

    Breitung

    K. Breitung. The geometry of limit state function graphs and subset simulation: Counterexamples. Reliability Engineering & System Safety, 182:98–106, 2019

    work page 2019

  4. [4]

    C ´erou, P

    F. C ´erou, P. Del Moral, T. Furon, and A. Guyader. Sequential Monte Carlo for rare event estimation. Statistics and Computing, 22:795–808, 2012

    work page 2012

  5. [5]

    Ditlevsen, R

    O. Ditlevsen, R. E. Melchers, and H. Gluver. General multi-dimensional prob- ability integration by directional simulation. Computers & Structures , 36(2): 355–368, 1990

    work page 1990

  6. [6]

    Douc and O

    R. Douc and O. Capp ´e. Comparison of resampling schemes for particle filter- ing. In ISPA 2005. Proceedings of the 4th International Symposium on Image and Signal Processing and Analysis, 2005 , pages 64–69, 2005

    work page 2005

  7. [7]

    W. R. Gilks, S. Richardson, and D. J. Spiegelhalter.Markov chain Monte Carlo in practice. Chapman and Hall/CRC, 1996

    work page 1996

  8. [8]

    Kanjilal and C

    O. Kanjilal and C. S. Manohar. Markov chain splitting methods in structural reliability integral estimation.Probabilistic Engineering Mechanics, 40:42–51, 2015

    work page 2015

  9. [9]

    H. J. Kinnear and F. A. DiazDelaO. Niching subset simulation. Probabilistic Engineering Mechanics, 79:103729, 2025

    work page 2025

  10. [10]

    Der Kiureghian and T

    A. Der Kiureghian and T. Dakessian. Multiple design points in first and second- order reliability. Structural Safety, 20:37–49, 1998

    work page 1998

  11. [11]

    Kurtz and J

    N. Kurtz and J. Song. Cross-entropy-based adaptive importance sampling using Gaussian mixture. Structural Safety, 42:35–44, 2013. Directional subset simulation method for reliability analysis 15

    work page 2013

  12. [12]

    R. E. Melchers and A. T. Beck. Structural reliability analysis and prediction . John Wiley & Sons, 2018

    work page 2018

  13. [13]

    Nie and B

    J. Nie and B. R. Ellingwood. A new directional simulation method for sys- tem reliability. Part I: application of deterministic point sets. Probabilistic Engineering Mechanics, 19:425–436, 2004

    work page 2004

  14. [14]

    Papaioannou and D

    I. Papaioannou and D. Straub. Combination line sampling for structural relia- bility analysis. Structural Safety, 88:102025, 2021

    work page 2021

  15. [15]

    Papaioannou, W

    I. Papaioannou, W. Betz, K. Zwirglmaier, and D. Straub. MCMC algorithms for subset simulation. Probabilistic Engineering Mechanics, 41:89–103, 2015

    work page 2015

  16. [16]

    Papaioannou, C

    I. Papaioannou, C. Papadimitriou, and D. Straub. Sequential importance sam- pling for structural reliability analysis. Structural Safety, 62:66–75, 2016

    work page 2016

  17. [17]

    H. J. Pradlwarter, G. I. Schu ¨eller, P. S. Koutsourelakis, and D. C. Charmpis. Application of line sampling simulation method to reliability benchmark prob- lems. Structural Safety, 29(3):208–221, 2007

    work page 2007

  18. [18]

    M. Rakshi. SESC: A new subset simulation method for rare-events estimation. Mechanical Systems and Signal Processing, 150(107139), 2021

    work page 2021

  19. [19]

    Sharma and C

    A. Sharma and C. S. Manohar. Modified replica exchange-based MCMC algorithm for estimation of structural reliability based on particle splitting method. Probabilistic Engineering Mechanics, 72:103448, 2023

    work page 2023

  20. [20]

    Song and R

    C. Song and R. Kawai. Monte Carlo and variance reduction methods for struc- tural reliability analysis: A comprehensive review. Probabilistic Engineering Mechanics, 73:103479, 2023

    work page 2023

  21. [21]

    Tabandeh, G

    A. Tabandeh, G. Jia, and P. Gardoni. A review and assessment of importance sampling methods for reliability analysis. Structural Safety, 97:102216, 2022

    work page 2022

  22. [22]

    Xian and Z

    J. Xian and Z. Wang. Relaxation-based importance sampling for structural reliability ana. Structural Safety, 106:102393, 2024

    work page 2024

  23. [23]

    Zhang, Z

    A. Zhang, Z. Chen, Q. Pan, X. Li, P. Feng, X. Gan, G. Chen, and L. Gao. Reliability analysis method for multiple failure modes with overlapping failure domains. Probabilistic Engineering Mechanics, 79:103741, 2025

    work page 2025

  24. [24]

    Zhang, Z

    X. Zhang, Z. Lu, and K. Cheng. Cross-entropy-based directional importance sampling with von Mises-Fisher mixture model for reliability analysis. Relia- bility Engineering & System Safety, 220:108306, 2022

    work page 2022

  25. [25]

    K. M. Zuev, J. L. Beck, S. K. Au, and L. S. Katafygiotis. Bayesian post- processor and other enhancements of subset simulation for estimating failure probabilities in high dimensions. Computers & Structures , 92–93:283–296, 2012

    work page 2012