Pith. sign in

REVIEW 4 major objections 8 minor 50 references

Chance-Constrained Optimal Design of Porous Thermal Insulation Systems Under Spatially Correlated Uncertainty

T0 review · 4 major / 8 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Chance-constrained insulation design becomes tractable at large scale via a quadratic Taylor control variate and sigmoid-smoothed chance constraints.

desk verdict Solid application of Taylor-based UQ to aerogel thermal break design, but the chance constraint is never verified against true Monte Carlo — a needed revision rather than a rejection. read the letter →

arxiv 2501.02077 v2 pith:HIIQ4GNK submitted 2025-01-03 cs.CE

Pratyush Kumar Singh , Danial Faghihi This is my paper · ORCID
classification cs.CE
keywords thermalinsulationchance-constrainedoptimizationspatiallycorrelateduncertaintycontrolvariatequadraticTaylorapproximationtopologyundervonMisesstresssilicaaerogel
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that optimal designs for silica-aerogel thermal breaks, balancing insulation quality against the risk of stress failure, can be computed under spatially correlated manufacturing uncertainty even when the design parameter space contains hundreds of thousands of unknowns. The method models porosity as a random field, expresses the insulation objective through mean and variance, and enforces mechanical safety by a chance constraint that limits the probability that the von Mises stress p-norm crosses a critical value. To keep cost low, a second-order Taylor expansion in the uncertain parameters is used as a control variate for Monte Carlo estimates, and a sigmoid-smoothed indicator function makes the chance constraint differentiable for gradient-based optimization. The paper demonstrates the framework on two- and three-dimensional beam-insulator models, including a design problem with 919,464 parameters, and reports that the final design satisfies the critical chance constraint.

What carries the argument

The load-bearing object is the pair of quadratic Taylor surrogates $Q_{QUAD}(m)$ and $f_{QUAD}(m)$ for the insulation objective and the stress margin, together with their use as control variates in Monte Carlo moment estimation. Their expectations and variances are expressed through the covariance-preconditioned Hessian $H_k = K \bar{Q}_{mm}$, whose trace is estimated by a randomized eigensolver on a generalized eigenvalue problem; the same eigenvalue structure and the surrogate $f_{QUAD}$ are used inside the chance constraint. A sigmoid $l_\omega(x) = 1/(1 + e^{-2\omega x})$ smooths the discontinuous failure indicator, and a continuation scheme gradually increases $\omega$ and the quadratic penalty weight $\gamma$ inside an inexact Newton conjugate-gradient optimizer. The p-norm von Mises stress aggregation $T_{pn} = (\int_\Omega T_{VM}^p \, d\Omega)^{1/p}$ is the smooth stand-in for peak stress that makes the chance constraint computable.

What would settle it

Run a high-fidelity Monte Carlo evaluation of the optimized 3D design from Section 5.5 with tens of thousands of samples of the original finite element model, count the fraction of realizations with $T_{pn}$ exceeding $T_{cr} = 12.5$ MPa, and compare it to $\alpha_c = 0.1$; if the measured fraction clearly exceeds 0.1, the surrogate-based chance constraint is satisfied only approximately and the paper's central scalability claim would not carry over to true safety.

Watch

Extended reading notes

Core claim

The central claim is that second-order Taylor expansions of both the design objective and the stress chance constraint, evaluated at the mean of a Gaussian random field, are enough to drive large-scale PDE-constrained optimization under uncertainty, provided the expansion is used as a control variate rather than as a direct replacement. The porosity field is parametrized as $\phi_f = \mathrm{sigmoid}(d + m)$, with $m$ a Mat\'ern random field, so the design variable $d$ is high-dimensional and spatially correlated uncertainties enter multiplicatively. The objective is mean-variance thermal compliance, and the constraint is $P(T_{cr} - T_{pn}(m,d) \ge 0) \le \alpha_c$, with $T_{pn}$ the p-norm of von Mises stress. The chance constraint is evaluated on the quadratic surrogate $f_{QUAD}$, smoothed by a sigmoid, and converted to a quadratic penalty inside an inexact Newton conjugate-gradient loop. Numerical experiments on 2D and 3D thermal-break systems claim converged designs at design dimensions up to 919,464, with the 3D case reaching the critical chance threshold $\alpha_c = 0.1$.

Load-bearing premise

The entire feasibility guarantee rests on the second-order Taylor expansion of the von Mises stress p-norm being accurate in the low-probability tail near the stress threshold, because the paper evaluates the chance constraint and its convergence check on that surrogate and never verifies the final design's failure probability with high-fidelity Monte Carlo.

Editorial extensions

If this is right

  • If the claims hold, chance-constrained design of additively manufactured insulation becomes feasible at full finite-element resolution rather than only on coarsened models.
  • The reported convergence behavior implies that refining the mesh and increasing the number of design parameters does not change the number of optimizer iterations or the eigenvalue decay, so the method is effectively dimension-agnostic.
  • Using the quadratic expansion as a control variate lowers the Monte Carlo sample count needed for accurate mean-variance estimates; the paper presents cases where 100 control-variate samples behave like 500 standard Monte Carlo samples under high variance.
  • Tuning the critical chance $\alpha_c$ and critical stress $T_{cr}$ produces a design trade-off curve between insulation performance and mechanical safety, with lower $\alpha_c$ yielding more low-porosity support material.
  • The 3D result that different random initial fields converge to the same final pattern suggests that the continuation scheme reaches a design that is canonical for the tested load and boundary conditions.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • My inference: the paper's reliability story is only as strong as the quadratic surrogate in the failure tail, because the final designs are accepted on $f_{QUAD}$; a high-fidelity Monte Carlo check of the optimized 3D design is the natural next experiment.
  • My inference: the same control-variate-plus-sigmoid recipe should transfer to other spatially correlated manufacturing uncertainties, such as layer thickness, residual stress, or thermal conductivity fields, whenever the response is sufficiently smooth in the uncertain parameter.
  • My inference: the Gaussian random field assumption for porosity could understate extreme events; replacing the Mat\'ern noise with heavier-tailed or non-Gaussian fields and re-running the framework would reveal how much of the design is driven by the Gaussian tail behavior.
  • My inference: a cheaper validation could use importance sampling with the quadratic surrogate as the proposal to estimate the true failure probability at a fraction of the cost of brute-force Monte Carlo.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 8 minor

Summary. This paper proposes a computational framework for the risk-averse, chance-constrained optimal design of silica aerogel thermal break components under spatially correlated manufacturing uncertainty. The design variable is the spatial porosity field; the uncertainty is modeled as a Matérn Gaussian random field; the objective is thermal compliance with a mean-plus-variance risk term and Tikhonov regularization; and mechanical stability is enforced through a chance constraint on the p-norm of the von Mises stress. To make the problem tractable, the paper uses a second-order Taylor expansion of the objective as a Monte Carlo control variate, a second-order expansion of the stress constraint as a surrogate for the failure probability, a sigmoid-smoothed quadratic penalty with a continuation scheme, and an inexact Newton-CG optimizer. Numerical experiments include 2D parameter studies in the critical chance αc and critical stress Tcr, plus a 3D design problem with 919,464 design parameters.

Significance. The paper extends the Taylor-expansion-based methodology for PDE-constrained optimization under uncertainty (Alexanderian et al. 2017; Chen et al. 2019; Chen and Ghattas 2021; Tan and Faghihi 2024) in nontrivial ways: the control-variate formulation of Section 3.2 and the complete design gradient derived through the Lagrangian construction of Appendix B are substantial technical pieces, and the 3D demonstration at 919,464 design parameters is at the scale of the most ambitious stress-constrained designs under uncertainty reported in the literature. The public code release (Section 5) and the careful convergence studies for the objective functional (Figures 5-7) are concrete strengths. However, the central claim that the framework enforces mechanical stability is compromised, as written, by an event-direction error in the chance constraint (Eq. 10, Major Comment 1), and the reported failure and safety probabilities are computed exclusively from the quadratic surrogate (Major Comment 2). Both issues are fixable but require re-examination of the reported designs and qualitative conclusions.

major comments (4)
  1. [§2.3, Eqs. (9)-(11); §4.2, Eq. (30); §5.3-5.5] The chance constraint as written constrains the wrong event relative to the paper's stated objective. With f = Tcr - Tpn (Eq. 9) and the indicator χ[0,∞)(f) (Eq. 11), the event {f ≥ 0} is the safe event (the stress p-norm is below the critical value), and Eq. (10) imposes P(f ≥ 0) ≤ αc. The abstract states that the framework enforces 'mechanical stability by restricting the probability of exceeding critical stress thresholds,' which requires P(f < 0) ≤ αc, i.e., P(f ≥ 0) ≥ 1 - αc. The penalty in Eq. (30), with E[lω(fQUAD)] ≈ P(f ≥ 0), consequently drives the safe probability down to αc ∈ {0.05, 0.08, 0.1}; Figure 8(a) is described in Section 5.3 as showing 'rapid decay of the critical chance,' and Section 5.5 reports that 'the probability in (10) approaches' αc = 0.1 from above. Under the formulation as written, the final 3D design is therefore expected to satisfy the critical stress constraint with only about 10% probability and to violate it with about 90% probability, which contradicts the central stability claim. The correction is local (e.g., apply the indicator to -f, or redefine f = Tpn - Tcr), but it is not cosmetic: the parametric behaviors in Section 5.4 (increasing αc expands the mechanically stronger regions) and the decay in Figure 8(a) are consistent with the inverted constraint, so the numerical studies and the qualitative conclusions must be re-examined under the corrected formulation.
  2. [§3.1, Eq. (20); §4.2, Eq. (30); Algorithm 2; §5.5] The probability reported in the chance-constraint results is never verified against the true forward model. In Eq. (20), the probability in Eq. (10) is replaced by a Monte Carlo estimate of the indicator applied to the quadratic surrogate fQUAD; the same fQUAD enters the penalty term in Eq. (30), and Algorithm 2 tracks it as the 'approximate chance' f̂, so Figure 8(a) and the 3D feasibility claim in Section 5.5 quantify the surrogate probability, not the probability of the exact thermo-mechanical response. Section 5.2 validates the quadratic approximation and control variate only for the objective functional Q (Figures 5-7); no experiment evaluates f = Tcr - Tpn with the full finite element solve at any optimized design. The p-norm of the von Mises stress is strongly nonlinear in the porosity field, so a second-order expansion around the mean can be accurate for the central distribution while misestimating the low-probability tail that determines the constraint at αc ∈ {0.05, 0.1}. The authors should add a post-optimality verification: draw samples of m, solve the full forward model, evaluate the exact indicator in Eq. (11), and report the empirical probability at each reported optimal design alongside the fQUAD estimate. Without this check, the central claim that the final 3D design satisfies the probabilistic stress constraint is not established.
  3. [§2.3, Eq. (8); §4.1-4.3, Algorithm 2; §5.3-5.5] Several quantities that determine the meaning and reproducibility of the chance-constraint results are not reported. First, the stress aggregation exponent p in Eq. (8) is never stated; since Tpn is a one-parameter family of relaxations of the maximum von Mises stress, both the interpretation of Tcr and the tail distribution of f depend on p. Second, the continuation schedule (ω0, γ0, Λω, Λγ) of Algorithm 2 is never specified; the convergence shown in Figure 8(a) and the claimed feasibility of the 3D design depend on ω being large enough for the sigmoid in Eq. (27) to approximate the indicator and on γ being large enough for the quadratic penalty in Eq. (29) to be binding, and without the schedule the reader cannot tell whether the final iterate satisfies the original constraint (10) or only the smoothed, penalized surrogate. Third, the sample count Nmc in Eq. (20) and the number of retained eigenpairs Nf_eig for the constraint surrogate are not stated for any experiment. These values should be reported in Sections 5.3-5.5 or in Table 1.
  4. [§5.3, §5.5] The scalability claim is only partially evidenced. The dimension-independence study in Figure 8 is carried out on the 2D problem corresponding to Figure 6(c), whereas for the 3D problem with 919,464 design parameters the paper reports only the final design patterns (Figures 12-13) and gives no iteration counts, total numbers of PDE solves, eigenvalue counts, or wall-clock times; the abstract's claim that the framework is 'agnostic to the dimensionality of the stochastic design space' is therefore not directly supported at the scale where it matters most. In addition, the statement in Section 5.5 that 'we achieve global optima' because several initial conditions produced the same pattern overstates what a small number of initializations can establish for a nonconvex optimization problem; this should be rephrased as convergence to the same local solution for the tested initializations.
minor comments (8)
  1. [§3.1, Eq. (20)] Equation (20) is typeset as 'P (f (m, d)) ≥ 0) = ...'; the opening parenthesis is misplaced and the formula should read P(f(m,d) ≥ 0) = ... .
  2. [§2.3, Eq. (15)] In Eq. (15) the indicator is written as χ[0,∞)](·) with a stray closing bracket; use the notation χ[0,∞)(·) defined in Eq. (11) consistently.
  3. [§5.1-5.5, figure captions] The mean of the uncertain parameter is denoted m̄ in the text and m̃ in several figure captions (Figures 4, 9, 10, 12); unify the notation.
  4. [§5.5] The 3D experiment introduces 'a uniform circular load ... simulating the influence of additional building envelope assemblies,' but the abstract formulation (1)-(3) does not include external mechanical tractions (the traction term ⟨t, wu⟩ appears only in the weak form (4)); the load magnitude, area, and direction should be specified.
  5. [§5.4, Figures 9-10] The color conventions in Figures 9-10 are described only in prose ('mechanically stronger material (highlighted in red)', 'stronger insulating material (depicted in blue)'); state explicitly in the captions that red corresponds to low porosity (strong, less insulating) and blue to high porosity (weak, more insulating).
  6. [Appendix C, Algorithm 2] The loop condition 'while ||dk - dk-1|| ≥ ϵout or k < kM' is ambiguous; presumably an 'and' is intended so that the iteration stops when the design update is small or the maximum iteration count is reached, and the roles of ϵout and kM should be clarified.
  7. [§5.2, Figure 5] The box plots in Figure 5 are said to show 'lower error bounds,' but the error metric and the reference solution used for the comparison are not defined; specify the ground truth (presumably a large-sample Monte Carlo estimate) and the error definition.
  8. [§3.1, §1] There are several typographical errors: 'the the mean and variance' in Section 3.1, 'probabilistic constrained' in the Introduction, and 'the the' in the sentence introducing the trace estimator; these should be corrected.

Circularity Check

1 steps flagged · score 4.0 of 10

Chance feasibility is reported from the same quadratic surrogate fQUAD that is minimized, so the true failure probability is never independently verified.

  1. other [Section 3.1, Eq. (20); Section 4.3, Eq. (30); Appendix C, Algorithm 2]
    "The probability P (f (., d) ≥ 0) by Monte Carlo estimator (15) with the quadratic approximation yields, P (f (m, d)) ≥ 0) = 1/Nmc Σ χ[0,∞)(fQUAD(mi, d)). ... JQUAD(d) = ... + τγ(E[lω(fQUAD(m, d))] − αc). ... Algorithm 2: ... Evaluate approximate chance ˆfk+1 at dk+1."

    The chance-constraint value reported as P(f(m,d)≥0) is not the original FE probability of Eq. (10); it is the Monte Carlo estimate of the quadratic surrogate fQUAD from Eq. (20). The same fQUAD appears in the quadratic penalty of the optimized objective in Eq. (30), and the outer continuation loop in Algorithm 2 terminates using this approximate chance. Thus the feasibility claim is evaluated with the same approximation that was minimized, so the reported probability is forced by construction to match the surrogate. No high-fidelity FE Monte Carlo check of f is performed, so the true tail probability is never independently predicted.

full rationale

The scalability framework itself has independent content: mean and variance estimates of Q are compared against standard Monte Carlo in Figure 5, and the dimension studies in Figures 8 and the 3D experiment support the algorithmic scalability claim. The citations to prior work, including [24] by one of the authors, are legitimate reuse of published Taylor/control-variate derivations rather than load-bearing self-citations or imported uniqueness theorems. The only notable circularity is the chance-constraint enforcement loop: Eq. (20) defines the reported probability through fQUAD, Eq. (30) places that same fQUAD in the penalty objective, and Algorithm 2 checks convergence with the same approximate chance. Because the optimization and the reported feasibility share the same surrogate, the final statement that the design satisfies P(f(m,d)≥0)≤αc is unverified for the true finite-element stress response. This is a partial self-confirming loop in the constraint enforcement, while the central scalability claim remains independent; hence a moderate score of 4 rather than a higher one.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central method postulates no new physical entities. The free parameters are user-set algorithmic and scenario choices; none are fitted to experimental data. The key load-bearing assumptions are the accuracy of the multiphase material model, the Gaussian random-field representation of manufacturing variability, and the fidelity of the quadratic surrogate for the stress chance constraint, the last of which is not validated with high-fidelity Monte Carlo.

free parameters (6)
  • βV (variance weight in objective) = 0.1
    User-set coefficient for V[Q] in Eq. (7); affects the balance between mean performance and risk, not fitted to data.
  • βRG (regularization weight) = 1e-5
    User-set Tikhonov regularization coefficient in Eq. (7); controls interface thickness.
  • Continuation schedule (ω0, γ0, Λω, Λγ) = not reported
    Algorithmic parameters for sigmoid smoothing and quadratic penalty in Appendix C; no numerical values are given, preventing exact replication.
  • Number of dominant eigenvalues N_eig = 25 (Figs. 5-7); not stated for 3D
    Truncation order for the randomized trace estimator in Eq. (23); controls the accuracy of the Taylor-based moments.
  • Stress p-norm exponent p = not reported
    Exponent in Eq. (8) for the von Mises stress aggregation; required to compute the chance constraint but never specified.
  • Uncertainty field parameters (σ, LCR, anisotropy) = σ=0.25/0.5/0.12, LCR=0.05/0.25, ϑx=1, ϑy=1e-4
    Scenario choices for the Matérn random field in Section 5.1; not fitted but define the cases studied.
assumptions (5)
  • domain assumption The multiphase continuum mixture model (Eqs. 1-2) represents the thermomechanical behavior of silica aerogel accurately.
    Adopted from prior work [44] with parameters from the literature; no experimental validation is presented in this paper.
  • domain assumption Manufacturing porosity uncertainty is a Gaussian random field with Matérn covariance and strong anisotropy along the deposition direction.
    Sections 2.2 and 5.1; the entire chance-constrained formulation relies on this stochastic model.
  • domain assumption The p-norm of the von Mises stress (Eq. 8) is a satisfactory surrogate for the maximum stress in the constraint.
    Standard in stress-constrained topology optimization, but the p value is not stated.
  • ad hoc to paper The second-order Taylor expansion of the chance constraint function f remains accurate in the tail region relevant to αc.
    Core to the method: Eq. (20) uses fQUAD for P(f>=0), but no tail-accuracy check or high-fidelity Monte Carlo verification is reported.
  • ad hoc to paper The sigmoid smoothing and quadratic penalty continuation converge to the feasible set of the original chance-constrained problem.
    Algorithm 2 relies on increasing ω and γ; convergence to the true chance constraint is assumed and not demonstrated against full Monte Carlo.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Chance-Constrained Optimal Design of Porous Thermal Insulation Systems Under Spatially Correlated Uncertainty." pith.science (2026). https://pith.science/paper/HIIQ4GNK

@misc{pith2026250102077,
  author       = {Pith},
  title        = {Pith review of: Chance-Constrained Optimal Design of Porous Thermal Insulation Systems Under Spatially Correlated Uncertainty},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HIIQ4GNK}},
  note         = {Machine review of arXiv:2501.02077}
}
read the original abstract

This paper presents a computationally efficient method for the optimal design of silica aerogel porous material systems, balancing thermal insulation performance with mechanical stability under stress concentrations. The proposed approach explicitly accounts for additive manufacturing uncertainties by modeling material porosity as a spatially correlated stochastic field within a multiphase finite element formulation. A risk-averse objective function, incorporating statistical moments of the design objective, is employed in conjunction with chance constraints that enforce mechanical stability by restricting the probability of exceeding critical stress thresholds. To mitigate the prohibitively high computational cost associated with the large-dimensional uncertainty space and Monte Carlo estimations of the objective function's statistical moments, a second-order Taylor expansion is utilized as a control variate. Furthermore, a continuation-based smoothing strategy is introduced to address the non-differentiability of the chance constraints, ensuring compatibility with gradient-based optimization. The resulting framework achieves computational scalability, remaining agnostic to the dimensionality of the stochastic design space. The effectiveness of the method is demonstrated through numerical experiments on two- and three-dimensional thermal break systems for building insulation. The results highlight the framework's capability to solve large-scale, chance-constrained optimal design problems governed by finite element models with uncertain design parameter spaces reaching dimensions in the hundreds of thousands.

Figures

Figures reproduced from arXiv: 2501.02077 by the authors.

Figure 1
Figure 1. The domain of a beam-insulator system used for numerical experiments, indicating both [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Finite element solution of the thermomechanical (forward) model for the beam-insulator [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Approximation of the discontinuous indicator function [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Samples of uncertain parameter m and corresponding aerogel porosity field ϕf (x) for cor￾relation length LCR = 0.25 and variance of σ 2 = 0.5 2 . The mean of the uncertain parameter is ˜m = 0 and anisotropy is ϑx = 1, ϑy = 1 × 10−4 for both cases. 5.1. Spatially-correl…
Figure 5
Figure 5. Figure 5: Convergence plots corresponding to scenario in Figure 6 for the mean and variance of the [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: The top row shows the optimal design obtained for [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: The top row shows the optimal design obtained for [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Optimization solution corresponding to Figure 6 (c) for different dimensions of design [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: Effect of the critical chance αc on optimal design. The top row shows the optimal design results obtained at the mean of the uncertain sample ˜m = 0 for different values of critical chance: (a) αc = 0.05 (b) αc = 0.08 and (c) αc = 0.1. The bottom row represents the von…
Figure 10
Figure 10. Figure 10: Effect of the limiting critical stress Tcr on optimal design. The top row shows the optimal design results obtained at the mean of the uncertain sample ˜m = 0 for different values of limiting critical stress: (a) Tcr = 20 MPa (b) Tcr = 22.5 MPa and (c) Tcr = 25 MPa. T…
Figure 11
Figure 11. Figure 11: Porosity and states of the scenario at the optimal design shown in Figure 9 (b) obtained [PITH_FULL_IMAGE:figures/full_fig_p019_11.png]
Figure 12
Figure 12. Figure 12: Evolution of design pattern with continuation scheme. Optimal design obtained at each [PITH_FULL_IMAGE:figures/full_fig_p020_12.png]
Figure 13
Figure 13. Figure 13: Optimal design patterns for the 3D scenario. Final design obtained after five steps of the [PITH_FULL_IMAGE:figures/full_fig_p021_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 47 canonical work pages

  1. [1]

    L. An, B. Liang, Z. Guo, J. Wang, C. Li, Y. Huang, Y. Hu, Z. Li, J. N. Arm- strong, C. Zhou, et al., Wearable aramid–ceramic aerogel composite for harsh environment, Advanced Engineering Materials 23 (3) (2021) 2001169

    work page 2021

  2. [2]

    Z. Guo, R. Yang, T. Wang, L. An, S. Ren, C. Zhou, Cost-Effective Additive Man- ufacturing of Ambient Pressure-Dried Silica Aerogel, Journal of Manufacturing Science and Engineering 143 (1) (2020) 011011. doi:10.1115/1.4048740

    work page doi:10.1115/1.4048740 2020

  3. [3]

    L. An, M. Di Luigi, J. Tan, D. Faghihi, S. Ren, Flexible percolation fibrous thermal insulating composite membranes for thermal management, Materials Advances 4 (1) (2023) 284–290

    work page 2023

  4. [4]

    Woischwill, I

    C. Woischwill, I. Y. Kim, Multimaterial multijoint topology optimization, Interna- tional Journal for Numerical Methods in Engineering 115 (13) (2018) 1552–1579

    work page 2018

  5. [5]

    Guilleminot, A

    J. Guilleminot, A. Asadpoure, M. Tootkaboni, Topology optimization under topo- logically dependent material uncertainties, Structural and Multidisciplinary Op- timization 60 (2019) 1283–1287

    work page 2019

  6. [6]

    Keshavarzzadeh, R

    V. Keshavarzzadeh, R. M. Kirby, A. Narayan, Stress-based topology optimization under uncertainty via simulation-based gaussian process, Computer Methods in Applied Mechanics and Engineering 365 (2020) 112992

    work page 2020

  7. [7]

    Keshavarzzadeh, F

    V. Keshavarzzadeh, F. Fernandez, D. A. Tortorelli, Topology optimization under uncertainty via non-intrusive polynomial chaos expansion, Computer Methods in Applied Mechanics and Engineering 318 (2017) 120–147

    work page 2017

  8. [8]

    L. W. Ng, K. E. Willcox, Multifidelity approaches for optimization under uncer- tainty, International Journal for numerical methods in Engineering 100 (10) (2014) 746–772

    work page 2014

Show all 50 references
  1. [9]

    Zheng, Z

    J. Zheng, Z. Luo, C. Jiang, J. Wu, Level-set topology optimization for robust de- sign of structures under hybrid uncertainties, International Journal for Numerical Methods in Engineering 117 (5) (2019) 523–542

    2019

  2. [10]

    G. A. da Silva, N. Aage, A. T. Beck, O. Sigmund, Three-dimensional manufac- turing tolerant topology optimization with hundreds of millions of local stress constraints, International Journal for Numerical Methods in Engineering 122 (2) (2021) 548–578

    2021

  3. [11]

    R. B. dos Santos, A. J. Torii, A. A. Novotny, Reliability-based topology optimiza- tion of structures under stress constraints, International journal for numerical methods in engineering 114 (6) (2018) 660–674. 30

    2018

  4. [12]

    W. Ding, H. Liao, X. Yuan, Multi-material topology optimization considering arbitrary strength and yield criteria constraints with single-variable interpolation, International Journal for Numerical Methods in Engineering 125 (19) (2024) e7561

    2024

  5. [13]

    Alexanderian, N

    A. Alexanderian, N. Petra, G. Stadler, O. Ghattas, Mean-variance risk-averse optimal control of systems governed by pdes with random parameter fields us- ing quadratic approximations, SIAM/ASA Journal on Uncertainty Quantification 5 (1) (2017) 1166–1192

    2017

  6. [14]

    P. Chen, U. Villa, O. Ghattas, Taylor approximation and variance reduction for pde-constrained optimal control under uncertainty, Journal of Computational Physics 385 (2019) 163–186

    2019

  7. [15]

    P. Chen, M. R. Haberman, O. Ghattas, Optimal design of acoustic metamaterial cloaks under uncertainty, Journal of Computational Physics 431 (2021) 110114

    2021

  8. [16]

    J. C. Kr¨ uger, M. Kranz, T. Schmidt, R. Seifried, B. Kriegesmann, An efficient and non-intrusive approach for robust design optimization with the first-order second- moment method, Computer methods in applied mechanics and engineering 414 (2023) 116136

    2023

  9. [17]

    Maute, Topology optimization under uncertainty, in: Topology optimization in structural and continuum mechanics, Springer, 2014, pp

    K. Maute, Topology optimization under uncertainty, in: Topology optimization in structural and continuum mechanics, Springer, 2014, pp. 457–471

    2014

  10. [18]

    D. P. Kouri, T. M. Surowiec, Risk-averse pde-constrained optimization using the conditional value-at-risk, SIAM Journal on Optimization 26 (1) (2016) 365–396

    2016

  11. [19]

    D. Luo, T. O’Leary-Roseberry, P. Chen, O. Ghattas, Efficient pde-constrained optimization under high-dimensional uncertainty using derivative-informed neural operators, arXiv preprint arXiv:2305.20053 (2023)

    2023 arXiv

  12. [20]

    K.-W. Liao, N. I. D. Biton, A heuristic moment-based framework for optimization design under uncertainty, Engineering with Computers 36 (2020) 1229–1242

    2020

  13. [21]

    P. Chen, O. Ghattas, Taylor approximation for chance constrained optimization problems governed by partial differential equations with high-dimensional random parameters, SIAM/ASA Journal on Uncertainty Quantification 9 (4) (2021) 1381– 1410

    2021

  14. [22]

    Kranz, J

    M. Kranz, J. K. L¨ udeker, B. Kriegesmann, A generalized approach for robust topology optimization using the first-order second-moment method for arbitrary response functions, Structural and multidisciplinary optimization 66 (5) (2023) 98

    2023

  15. [23]

    Kriegesmann, J

    B. Kriegesmann, J. K. L¨ udeker, Robust compliance topology optimization us- ing the first-order second-moment method, Structural and Multidisciplinary Op- timization 60 (2019) 269–286. 31

    2019

  16. [24]

    J. Tan, D. Faghihi, A scalable framework for multi-objective pde-constrained de- sign of building insulation under uncertainty, Computer Methods in Applied Me- chanics and Engineering 419 (2024) 116628

    2024

  17. [25]

    Y. Wang, O. Sigmund, Multi-material topology optimization for maximizing struc- tural stability under thermo-mechanical loading, Computer Methods in Applied Mechanics and Engineering 407 (2023) 115938

    2023

  18. [26]

    C. Wang, B. Xu, Z. Duan, J. Rong, Stress-related topology optimization for castable design, International Journal for Numerical Methods in Engineering 122 (21) (2021) 6203–6233

    2021

  19. [27]

    G. A. da Silva, A. T. Beck, O. Sigmund, Stress-constrained topology optimization considering uniform manufacturing uncertainties, Computer Methods in Applied Mechanics and Engineering 344 (2019) 512–537

    2019

  20. [28]

    K. Li, G. Cheng, Structural topology optimization of elastoplastic continuum un- der shakedown theory, International Journal for Numerical Methods in Engineer- ing 123 (19) (2022) 4459–4482

    2022

  21. [29]

    D. M. De Leon, J. Alexandersen, J. S. O. Fonseca, O. Sigmund, Stress-constrained topology optimization for compliant mechanism design, Structural and Multidis- ciplinary Optimization 52 (2015) 929–943

    2015

  22. [30]

    R. Zhao, J. Zhao, C. Wang, Stress-constrained multiscale topology optimization with connectable graded microstructures using the worst-case analysis, Interna- tional Journal for Numerical Methods in Engineering 123 (8) (2022) 1882–1906

    2022

  23. [31]

    Duysinx, M

    P. Duysinx, M. P. Bendsøe, Topology optimization of continuum structures with local stress constraints, International journal for numerical methods in engineering 43 (8) (1998) 1453–1478

    1998

  24. [32]

    D. Yang, H. Liu, W. Zhang, S. Li, Stress-constrained topology optimization based on maximum stress measures, Computers & Structures 198 (2018) 23–39

    2018

  25. [33]

    C. Le, J. Norato, T. Bruns, C. Ha, D. Tortorelli, Stress-based topology opti- mization for continua, Structural and Multidisciplinary Optimization 41 (2010) 605–620

    2010

  26. [34]

    T. T. Banh, Q. X. Lieu, J. Kang, Y. Ju, S. Shin, D. Lee, A novel robust stress- based multimaterial topology optimization model for structural stability frame- work using refined adaptive continuation method, Engineering with Computers 40 (2) (2024) 677–713

    2024

  27. [35]

    Y. Guo, S. Cheng, Y. Wang, X. Lai, L. Chen, Topology optimization for transient thermoelastic structures under time-dependent loads, Engineering with Comput- ers 40 (3) (2024) 1677–1693. 32

    2024

  28. [36]

    Herrero-P´ erez, S

    D. Herrero-P´ erez, S. G. Pic´ o-Vicente, H. Mart ´ ınez-Barber´ a, Adaptive density- based robust topology optimization under uncertain loads using parallel comput- ing, Engineering with Computers 40 (1) (2024) 21–43

    2024

  29. [37]

    Hoang, T

    V.-N. Hoang, T. Pham, D. Ho, H. Nguyen-Xuan, Robust multiscale design of in- compressible multi-materials under loading uncertainties, Engineering with Com- puters (2022) 1–16

    2022

  30. [38]

    Y. Wu, H. Hu, J. Zheng, Y. Zhang, E. Li, Z. He, Robust topology optimization with interval field model: on the spatially varied non-probabilistic uncertainty of material property, loading and geometry, Engineering with Computers 40 (2) (2024) 1093–1109

    2024

  31. [39]

    H. Deng, A. C. To, Topology optimization based on deep representation learn- ing (drl) for compliance and stress-constrained design, Computational Mechanics 66 (2) (2020) 449–469

    2020

  32. [40]

    Cheng, B

    C. Cheng, B. Yang, X. Wang, I. Lee, Uncertainty-oriented thermoelastic topology optimization with stress constraint, International Journal for Numerical Methods in Engineering 125 (10) (2024) e7441

    2024

  33. [41]

    M. K. Oh, D. S. Lee, J. Yoo, Stress constrained topology optimization simulta- neously considering the uncertainty of load positions, International Journal for Numerical Methods in Engineering 123 (2) (2022) 339–365

    2022

  34. [42]

    Mart ´ ınez-Frutos, R

    J. Mart ´ ınez-Frutos, R. Ortigosa, Risk-averse approach for topology optimization of fail-safe structures using the level-set method, Computational Mechanics 68 (5) (2021) 1039–1061

    2021

  35. [43]

    Y. Liu, L. Wang, D. Liu, Uncertain dynamic topology optimization based on the interval reliability evaluation and equivalent static loads (esls) algorithm, Engi- neering with Computers 38 (6) (2022) 5657–5674

    2022

  36. [44]

    J. Tan, U. Villa, N. Shamsaei, S. Shao, H. M. Zbib, D. Faghihi, A predictive discrete-continuum multiscale model of plasticity with quantified uncertainty, In- ternational Journal of Plasticity 138 (2021) 102935

    2021

  37. [45]

    Roininen, J

    L. Roininen, J. M. J. Huttunen, S. Lasanen, Whittle-mat´ ern priors for bayesian statistical inversion with applications in electrical impedance tomography, Inverse Problems and Imaging 8 (2) (2014) 561–586. doi:10.3934/ipi.2014.8.561. URL https://www.aimsciences.org/article/i...

    2014 doi

  38. [46]

    Villa, N

    U. Villa, N. Petra, O. Ghattas, HIPPYlib: An Extensible Software Framework for Large-Scale Inverse Problems Governed by PDEs: Part I: Deterministic Inversion and Linearized Bayesian Inference, ACM Trans. Math. Softw. 47 (2) (Apr. 2021). 33 doi:10.1145/3428447. URL https://doi....

    2021 doi

  39. [47]

    Nocedal, S

    J. Nocedal, S. J. Wright, Quadratic programming, Numerical optimization (2006) 448–492

    2006

  40. [48]

    M. S. Alnæs, J. Blechta, J. Hake, A. Johansson, B. Kehlet, A. Logg, C. Richardson, J. Ring, M. E. Rognes, G. N. Wells, The FEniCS project version 1.5, Archive of Numerical Software 3 (100) (2015). doi:10.11588/ans.2015.100.20553

    2015 doi

  41. [49]

    D. Luo, P. Chen, T. O’Leary-Roseberry, U. Villa, O. Ghattas, Soupy: Stochas- tic pde-constrained optimization under high-dimensional uncertainty in python, Journal of Open Source Software 9 (99) (2024) 6101. doi:10.21105/joss.06101. URL https://doi.org/10.21105/joss.06101

    2024 doi

  42. [50]

    P. K. Singh, K. A. Farrell-Maupin, D. Faghihi, A framework for strategic discovery of credible neural network surrogate models under uncertainty, Computer Methods in Applied Mechanics and Engineering 427 (2024) 117061. 34

    2024

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.