pith. sign in

arxiv: 2508.18170 · v2 · submitted 2025-08-25 · 💻 cs.CE

Balancing the exploration-exploitation trade-off in active learning for surrogate model-based reliability analysis via multi-objective optimization

Jonathan A. Moran , Pablo G. Morato This is my paper

Pith reviewed 2026-05-18 21:05 UTC · model grok-4.3

classification 💻 cs.CE
keywords active learningsurrogate modelsreliability analysismulti-objective optimizationexploration-exploitation trade-offfailure probabilitylimit-state function
0
0 comments X

The pith

Casting acquisition as multi-objective optimization lets adaptive rules balance exploration against exploitation for surrogate reliability analysis.

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

The paper establishes that single-score acquisition functions conceal the exploration-exploitation trade-off. It instead casts next-sample selection as an explicit multi-objective optimization problem whose solution is a compact Pareto set of candidates. Adaptive selection rules, such as a scheduled shift from exploration to exploitation and reliability-aware picking, then choose the evaluation point from that set. Tests across diverse limit-state functions show these strategies deliver robust rankings, reach strict failure-probability error targets, and improve sample efficiency compared with conventional methods. The result matters because expensive high-fidelity models become practical for reliability assessment when the number of required evaluations drops while accuracy targets remain fixed.

Core claim

By formulating sample acquisition as a multi-objective optimization problem in which exploration (global uncertainty reduction) and exploitation (accuracy near the failure boundary) are explicit competing objectives, the method produces a Pareto set that quantifies the trade-off; principled criteria and adaptive rules, including a scheduled exploration-to-exploitation shift and a reliability-aware selection rule, then select the next sample, yielding strategies whose relative failure-probability error trajectories, sample-efficiency comparisons, and global rankings demonstrate robust overall performance that consistently meets strict error targets across diverse limit-state functions.

What carries the argument

The multi-objective optimization formulation of the acquisition step that returns a Pareto set of candidate samples from which adaptive trade-off rules choose the next limit-state evaluation.

If this is right

  • Relative failure-probability error trajectories converge to targets with fewer high-fidelity evaluations than single-score baselines.
  • Sample-efficiency rankings remain favorable across varied limit-state functions without per-problem retuning.
  • Strict error targets are met consistently by the scheduled-shift and reliability-aware selection rules.
  • The Pareto-set representation supplies an explicit, quantifiable view of the exploration-exploitation balance at each iteration.

Where Pith is reading between the lines

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

  • The same Pareto-based acquisition could be ported to other surrogate tasks such as Bayesian optimization or global sensitivity analysis that also face exploration-exploitation decisions.
  • Higher-dimensional input spaces may require new normalization or selection heuristics to keep the Pareto front computationally tractable.
  • Comparing the overhead of Pareto-front generation against portfolio-based acquisition would clarify whether the added machinery pays off in wall-clock time on very expensive models.

Load-bearing premise

The proposed adaptive trade-off rules will select samples from the Pareto set in a way that generalizes to unseen limit-state functions and surrogate models without extensive per-problem tuning.

What would settle it

A new limit-state function or surrogate model on which the adaptive MOO strategies either fail to reach the target error level or require substantially more evaluations than the conventional single-score methods.

Figures

Figures reproduced from arXiv: 2508.18170 by Jonathan A. Moran, Pablo G. Morato.

Figure 1
Figure 1. Figure 1: Multi-objective optimization framework for sample acquisition in active learning for reliability analysis. The framework integrates [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Pareto front in the exploration-exploitation objective space. The bi-objective space is defined by [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Knee point and compromise solution on the normalized Pareto front. The normalized bi-objective space is defined by exploitation [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Adaptive Pareto-based acquisition strategy based on reliability predictions (MOO-R). (Left) Normalized Pareto front [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Active learning curves across benchmark limit-state functions. The evolution of the relative [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Sample efficiency of active learning strategies across benchmark limit-state functions. The number of training samples required to achieve stable δβtarget, defined in [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Experiments not meeting the relative β-error targets within the acquisition budget. For each acquisition strategy (MOO-R, MOO-K, MOO-C, EFF, and U), the bars show the total number of experiments where the relative β-error target, δβtarget, is not reached within the allocated budget of 190 active samples. Results are aggregated over all limit-state functions. Colors within each bar identify the limit￾state … view at source ↗
Figure 9
Figure 9. Figure 9: Exploration-exploitation preference and relative [PITH_FULL_IMAGE:figures/full_fig_p021_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Pareto fronts and selected samples for the four-branch ( [PITH_FULL_IMAGE:figures/full_fig_p022_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Acquired samples in the input space for the four-branch ( [PITH_FULL_IMAGE:figures/full_fig_p022_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Exploration-exploitation preference and estimated failure probability for the four-branch ( [PITH_FULL_IMAGE:figures/full_fig_p023_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Exploration-exploitation preference and estimated failure probability for the nonlinear oscillator limit-state function. Results are [PITH_FULL_IMAGE:figures/full_fig_p024_13.png] view at source ↗
read the original abstract

Reliability assessment of engineering systems often requires repeated evaluations of limit-state functions that may rely on computationally expensive high-fidelity models, rendering direct sampling-based reliability analysis impractical. An effective solution is to approximate the limit-state function with a surrogate model that can be iteratively refined through active learning, thereby reducing the number of model evaluations. At each iteration, an acquisition strategy selects the next sample for evaluation by balancing two competing objectives: exploration, to reduce global predictive uncertainty, and exploitation, to improve accuracy near the failure boundary. Conventional strategies such as the U-function, EFF, ERF, REIF, and portfolio-based schemes encode this balance through single pointwise scores, concealing the underlying trade-off. In this work, we formulate sample acquisition as a multi-objective optimization (MOO) problem in which exploration and exploitation are explicit competing objectives, yielding a compact Pareto set that provides a quantifiable trade-off representation. To select samples from the Pareto set, we investigate principled MOO criteria and propose adaptive trade-off rules, including a scheduled exploration-to-exploitation shift and a reliability-aware selection rule. Across diverse limit-state functions, we evaluate all tested strategies through relative failure-probability error trajectories, sample-efficiency comparisons, and global rankings, showing that the adaptive MOO-based strategies achieve robust overall performance while consistently meeting strict error targets.

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

2 major / 2 minor

Summary. The paper formulates sample acquisition in active learning for surrogate-based reliability analysis as a multi-objective optimization (MOO) problem with explicit exploration (global uncertainty reduction) and exploitation (accuracy near the failure boundary) objectives. It generates a Pareto set of candidate samples and proposes adaptive selection rules, including a scheduled exploration-to-exploitation shift and a reliability-aware criterion, to choose the next evaluation point. Evaluations across multiple limit-state functions compare these strategies to conventional single-score methods (U-function, EFF, etc.) using relative failure-probability error trajectories, sample-efficiency metrics, and global rankings, claiming that the adaptive MOO approaches deliver robust performance and consistently meet strict error targets.

Significance. If the adaptive MOO rules generalize without per-problem retuning, the work offers a more transparent and quantifiable representation of the exploration-exploitation trade-off than single-score heuristics. This could improve sample efficiency and reliability in engineering applications involving expensive high-fidelity models, particularly when the Pareto front provides interpretable options for balancing the two objectives.

major comments (2)
  1. The central claim of robust generalization across diverse limit-state functions rests on evaluations using a finite collection of analytic benchmarks. No cross-validation on held-out functions or additional test cases with differing smoothness, dimensionality, or discontinuity characteristics is described, leaving open whether the observed rankings and error trajectories are driven by the adaptive rules themselves or by characteristics shared with the benchmark suite.
  2. The proposed adaptive trade-off rules (scheduled exploration-to-exploitation shift and reliability-aware Pareto selection) contain scheduling hyperparameters. The manuscript does not report sensitivity analysis on these parameters or demonstrate that performance remains stable under modest perturbations, which is necessary to substantiate that the adaptivity, rather than benchmark-specific tuning, produces the reported gains in sample efficiency and error-target compliance.
minor comments (2)
  1. Clarify the exact definition and implementation of the reliability-aware selection rule when multiple points lie on the Pareto front; a short pseudocode or explicit tie-breaking procedure would improve reproducibility.
  2. Ensure that all reported error trajectories include uncertainty bands (e.g., standard deviation over repeated runs) so that apparent differences between strategies can be assessed for statistical significance.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive comments, which help clarify the scope and limitations of our evaluation. We address each major comment below and outline revisions that will be incorporated into the next version of the manuscript.

read point-by-point responses

  1. Referee: The central claim of robust generalization across diverse limit-state functions rests on evaluations using a finite collection of analytic benchmarks. No cross-validation on held-out functions or additional test cases with differing smoothness, dimensionality, or discontinuity characteristics is described, leaving open whether the observed rankings and error trajectories are driven by the adaptive rules themselves or by characteristics shared with the benchmark suite.

    Authors: The benchmark functions were deliberately chosen from the standard set used in surrogate-based reliability analysis to span a range of dimensionalities, degrees of smoothness, and discontinuity features. The adaptive MOO strategies exhibited consistent ranking and error-target compliance across this collection without requiring function-specific retuning. We agree that a broader set of test cases would further strengthen the generalization claim. In the revision we will add results on two additional limit-state functions with higher dimensionality and more pronounced discontinuities, together with an expanded discussion of benchmark selection criteria and acknowledged limitations for functions outside the tested class. revision: yes

  2. Referee: The proposed adaptive trade-off rules (scheduled exploration-to-exploitation shift and reliability-aware Pareto selection) contain scheduling hyperparameters. The manuscript does not report sensitivity analysis on these parameters or demonstrate that performance remains stable under modest perturbations, which is necessary to substantiate that the adaptivity, rather than benchmark-specific tuning, produces the reported gains in sample efficiency and error-target compliance.

    Authors: The scheduling parameters were selected according to a principled linear decay of the exploration weight as surrogate accuracy improves, rather than by exhaustive per-benchmark optimization. While internal checks during development indicated stability, we acknowledge that a formal sensitivity study was omitted from the submitted manuscript. The revised version will include a new subsection that perturbs the scheduling and reliability-aware thresholds by ±20 % and reports the resulting variation in sample-efficiency metrics and failure-probability error trajectories for all benchmarks, thereby confirming that the reported gains arise from the adaptive structure itself. revision: yes

Circularity Check

0 steps flagged

No circularity: formulation and adaptive rules are presented as independent proposals evaluated empirically

full rationale

The paper formulates acquisition as an explicit MOO problem yielding a Pareto set, then proposes new adaptive selection rules (scheduled shift and reliability-aware) that are not derived from or fitted to the target performance metrics. Evaluation proceeds via direct empirical comparison of error trajectories and rankings on benchmark limit-state functions. No load-bearing step reduces by the paper's equations to a self-defined quantity, fitted input renamed as prediction, or self-citation chain; the central claims rest on the proposed heuristics and their observed behavior rather than tautological re-expression of inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach rests on standard multi-objective optimization concepts (Pareto dominance) and active-learning machinery already present in the literature; the abstract introduces no new free parameters, axioms beyond standard math, or invented entities.

axioms (1)
  • standard math Pareto optimality defines the set of non-dominated solutions for the two objectives
    Invoked when the paper states that sample acquisition yields a compact Pareto set.

pith-pipeline@v0.9.0 · 5770 in / 1284 out tokens · 41762 ms · 2026-05-18T21:05:32.912865+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

51 extracted references · 51 canonical work pages

  1. [1]

    Teixeira, M

    R. Teixeira, M. Nogal, A. O’Connor, Adaptive approaches in metamodel-based reliability analysis: A review, Structural Safety 89 (2021) 102019

    work page 2021

  2. [2]

    Teixeira, B

    R. Teixeira, B. Martinez-Pastor, M. Nogal, A. O’Connor, Reliability analysis using a multi-metamodel complement-basis approach, Reliability Engineering & System Safety 205 (2021) 107248

    work page 2021

  3. [3]

    Moustapha, S

    M. Moustapha, S. Marelli, B. Sudret, Active learning for structural reliability: Survey, general framework and benchmark, Structural Safety 96 (2022) 102174

    work page 2022

  4. [4]

    Sudret, Global sensitivity analysis using polynomial chaos expansions, Reliability Engineering & System Safety 93 (7) (2008) 964–979

    B. Sudret, Global sensitivity analysis using polynomial chaos expansions, Reliability Engineering & System Safety 93 (7) (2008) 964–979

    work page 2008

  5. [5]

    Bichon, S

    B. Bichon, S. Mahadevan, M. Eldred, Reliability-based design optimization using e fficient global reliability analysis, in: 50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference 17th AIAA/ASME/AHS Adaptive Structures Conference 11th AIAA No, 2009, p. 2261

    work page 2009

  6. [6]

    M. Ehre, I. Papaioannou, B. Sudret, D. Straub, Sequential active learning of low-dimensional model repre- sentations for reliability analysis, SIAM Journal on Scientific Computing 44 (3) (2022) B558–B584

    work page 2022

  7. [7]

    P. G. Morato, K. G. Papakonstantinou, C. P. Andriotis, J. S. Nielsen, P. Rigo, Optimal inspection and main- tenance planning for deteriorating structural components through dynamic Bayesian networks and Markov decision processes, Structural Safety 94 (2022) 102140

    work page 2022

  8. [8]

    Moran A., P

    J. Moran A., P. G. Morato, P. Rigo, Active learning for structural reliability analysis with multiple limit-state functions through variance-enhanced PC-Kriging surrogate models, in: Proceedings of the 14th International Conference on Applications of Statistics and Probability in Civil Engineering (ICASP14), 2023, pp. 1–8

    work page 2023

  9. [9]

    Zhang, L

    X. Zhang, L. Wang, J. D. Sørensen, REIF: A novel active-learning function toward adaptive Kriging surrogate models for structural reliability analysis, Reliability Engineering & System Safety 185 (2019) 440–454

    work page 2019

  10. [10]

    Marelli, B

    S. Marelli, B. Sudret, An active-learning algorithm that combines sparse polynomial chaos expansions and bootstrap for structural reliability analysis, Structural Safety 75 (2018) 67–74

    work page 2018

  11. [11]

    Schobi, B

    R. Schobi, B. Sudret, J. Wiart, Polynomial-chaos-based Kriging, International Journal for Uncertainty Quan- tification 5 (2) (2015)

    work page 2015

  12. [12]

    Y . Bao, Z. Xiang, H. Li, Adaptive subset searching-based deep neural network method for structural reliability analysis, Reliability Engineering & System Safety 213 (2021) 107778

    work page 2021

  13. [13]

    M. S. Chauhan, M. Ojeda-Tuz, R. A. Catarelli, K. R. Gurley, D. Tsapetis, M. D. Shields, On active learning for Gaussian process-based global sensitivity analysis, Reliability Engineering & System Safety 245 (2024) 109945

    work page 2024

  14. [14]

    Moustapha, P

    M. Moustapha, P. Parisi, S. Marelli, B. Sudret, Reliability analysis of arbitrary systems based on active learn- ing and global sensitivity analysis, Reliability Engineering & System Safety 248 (2024) 110150

    work page 2024

  15. [15]

    S. L. Dhulipala, M. D. Shields, P. Chakroborty, W. Jiang, B. W. Spencer, J. D. Hales, V . M. Laboure, Z. M. Prince, C. Bolisetti, Y . Che, Reliability estimation of an advanced nuclear fuel using coupled active learning, multifidelity modeling, and subset simulation, Reliability Engineering & System Safety 226 (2022) 108693

    work page 2022

  16. [16]

    Gaspar, A

    B. Gaspar, A. P. Teixeira, C. G. Soares, Adaptive surrogate model with active refinement combining Kriging and a trust region method, Reliability Engineering & System Safety 165 (2017) 277–291. 27

    work page 2017

  17. [17]

    B. J. Bichon, M. S. Eldred, L. P. Swiler, S. Mahadevan, J. M. McFarland, E fficient global reliability analysis for nonlinear implicit performance functions, AIAA Journal 46 (10) (2008) 2459–2468

    work page 2008

  18. [18]

    Echard, N

    B. Echard, N. Gayton, M. Lemaire, AK-MCS: An active learning reliability method combining Kriging and Monte Carlo Simulation, Structural Safety 33 (2) (2011) 145–154

    work page 2011

  19. [19]

    I. Das, J. E. Dennis, A closer look at drawbacks of minimizing weighted sums of objectives for Pareto set generation in multicriteria optimization problems, Structural Optimization 14 (1) (1997) 63–69

    work page 1997

  20. [20]

    R. T. Marler, J. S. Arora, The weighted sum method for multi-objective optimization: new insights, Structural and Multidisciplinary Optimization 41 (2010) 853–862

    work page 2010

  21. [21]

    Pareto, Manuale di economia politica, Piccola biblioteca scientifica, Societa Editrice, 1906

    V . Pareto, Manuale di economia politica, Piccola biblioteca scientifica, Societa Editrice, 1906

    work page 1906

  22. [22]

    R. T. Marler, J. S. Arora, Survey of multi-objective optimization methods for engineering, Structural and Multidisciplinary Optimization 26 (2004) 369–395

    work page 2004

  23. [23]

    Miettinen, Nonlinear multiobjective optimization, V ol

    K. Miettinen, Nonlinear multiobjective optimization, V ol. 12, Springer Science & Business Media, 1999

    work page 1999

  24. [24]

    Y . Shi, C. Lin, Y . Ma, J. Shen, Active learning Kriging-based multi-objective modeling and optimization for system reliability-based robust design, Reliability Engineering & System Safety 245 (2024) 110007

    work page 2024

  25. [25]

    Z. Lv, L. Wang, Z. Han, J. Zhao, W. Wang, Surrogate-assisted particle swarm optimization algorithm with Pareto active learning for expensive multi-objective optimization, IEEE /CAA Journal of Automatica Sinica 6 (3) (2019) 838–849

    work page 2019

  26. [26]

    M. Xiao, J. Zhang, L. Gao, A system active learning Kriging method for system reliability-based design optimization with a multiple response model, Reliability Engineering & System Safety 199 (2020) 106935

    work page 2020

  27. [27]

    J. Chen, Z. Chen, W. Jiang, H. Guo, L. Chen, A reliability-based design optimization strategy using quantile surrogates by improved PC-Kriging, Reliability Engineering & System Safety 253 (2025) 110491

    work page 2025

  28. [28]

    Dubourg, B

    V . Dubourg, B. Sudret, J.-M. Bourinet, Reliability-based design optimization using Kriging surrogates and subset simulation, Structural and Multidisciplinary Optimization 44 (2011) 673–690

    work page 2011

  29. [29]

    R. Liu, G. Ma, F. Kong, Z. Ai, K. Xiong, W. Zhou, X. Wang, X. Chang, Pareto-guided active learning for accelerating surrogate-assisted multi-objective optimization of arch dam shape, Engineering Structures 326 (2025) 119541

    work page 2025

  30. [30]

    R. S. de Oliveira, M. F. de L.O. Santos, S. M. Afonso, R. de S. Motta, Successive Pareto simulation method for efficient structural reliability analysis, Probabilistic Engineering Mechanics 81 (2025) 103819

    work page 2025

  31. [31]

    Bechikh, L

    S. Bechikh, L. Ben Said, K. Gh ´edira, Searching for knee regions in multi-objective optimization using mobile reference points, in: Proceedings of the 2010 ACM symposium on applied computing, 2010, pp. 1118–1125

    work page 2010

  32. [32]

    Zeleny, Compromise programming, Multiple Criteria Decision Making (1973)

    M. Zeleny, Compromise programming, Multiple Criteria Decision Making (1973)

    work page 1973

  33. [33]

    D. D. Cox, S. John, A statistical method for global optimization, in: IEEE international conference on systems, man, and cybernetics, IEEE, 1992, pp. 1241–1246

    work page 1992

  34. [34]

    Ranjan, D

    P. Ranjan, D. Bingham, G. Michailidis, Sequential experiment design for contour estimation from complex computer codes, Technometrics 50 (4) (2008) 527–541. 28

    work page 2008

  35. [35]

    Branke, K

    J. Branke, K. Deb, H. Dierolf, M. Osswald, Finding knees in multi-objective optimization, in: Parallel Prob- lem Solving from Nature-PPSN VIII: 8th International Conference, Birmingham, UK, September 18-22,

    work page

  36. [36]

    Proceedings 8, Springer, 2004, pp. 722–731

    work page 2004

  37. [37]

    K. Deb, S. Gupta, Understanding knee points in bicriteria problems and their implications as preferred solution principles, Engineering optimization 43 (11) (2011) 1175–1204

    work page 2011

  38. [38]

    Rasmussen, C

    C. Rasmussen, C. Williams, Gaussian processes for machine learning, Adaptive Computation and Machine Learning series, MIT Press, 2005

    work page 2005

  39. [39]

    Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound constrained optimization

    J. L. Morales, J. Nocedal, Remark on “Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound constrained optimization”, ACM Transactions on Mathematical Software (TOMS) 38 (1) (2011) 1–4

    work page 2011

  40. [40]

    M. D. McKay, R. J. Beckman, W. J. Conover, A comparison of three methods for selecting values of input variables in the analysis of output from a computer code, Technometrics 21 (2) (1979) 239–245

    work page 1979

  41. [41]

    Consortium des ´Equipements de Calcul Intensif (C ´ECI), C ´ECI High-performance computing clusters, https://www.ceci-hpc.be/clusters.html, accessed: August 21, 2025 (2025)

    work page 2025

  42. [42]

    A. B. Yoo, M. A. Jette, M. Grondona, SLURM: Simple Linux Utility for Resource Management, in: D. Fei- telson, L. Rudolph, U. Schwiegelshohn (Eds.), Job Scheduling Strategies for Parallel Processing, Springer Berlin Heidelberg, Berlin, Heidelberg, 2003, pp. 44–60

    work page 2003

  43. [43]

    Waarts, Structural reliability using finite element methods: an appraisal of directional adaptive response surface sampling (DARS), Ph.D

    P. Waarts, Structural reliability using finite element methods: an appraisal of directional adaptive response surface sampling (DARS), Ph.D. thesis, Technische Universiteit Delft (2000)

    work page 2000

  44. [44]

    Schueremans, D

    L. Schueremans, D. Van Gemert, Benefit of splines and neural networks in simulation based structural relia- bility analysis, Structural Safety 27 (3) (2005) 246–261

    work page 2005

  45. [45]

    Schueremans, D

    L. Schueremans, D. Van Gemert, Use of Kriging as meta-model in simulation procedures for structural relia- bility, in: 9th International Conference on Structural Safety and Reliability, 2005, pp. 2483–2490

    work page 2005

  46. [46]

    D. M. Himmelblau, et al., Applied nonlinear programming, McGraw-Hill, 2018

    work page 2018

  47. [47]

    K. G. Papakonstantinou, H. Nikbakht, E. Eshra, Hamiltonian MCMC methods for estimating rare events probabilities in high-dimensional problems, Probabilistic Engineering Mechanics 74 (2023) 103485

    work page 2023

  48. [48]

    Echard, N

    B. Echard, N. Gayton, M. Lemaire, N. Relun, A combined importance sampling and Kriging reliability method for small failure probabilities with time-demanding numerical models, Reliability Engineering & System Safety 111 (2013) 232–240

    work page 2013

  49. [49]

    M. H. F. Nielsen, Benefit of splines and neural networks in simulation based structural reliability analysis, Structural Safety 27 (2005) 403–404

    work page 2005

  50. [50]

    Gayton, J

    N. Gayton, J. M. Bourinet, M. Lemaire, CQ2RS: A new statistical approach to the response surface method for reliability analysis, Structural Safety 25 (1) (2003) 99–121

    work page 2003

  51. [51]

    Rackwitz, Reliability analysis — A review and some perspectives, Structural Safety 23 (4) (2001) 365–395

    R. Rackwitz, Reliability analysis — A review and some perspectives, Structural Safety 23 (4) (2001) 365–395. 29 Appendix A. Proofs Appendix A.1. Proof of Proposition 1 by contradiction Proof. Suppose xU is the minimizer of U(x) but is not Pareto optimal, then: ∃ x ∈ X : h ∀i ∈ {µ, σ}, fi(x) ≥ fi(xU) i ∧ h ∃i ∈ {µ, σ}, fi(x) > fi(xU) i . (A.1) By explicitl...

    work page 2001