Pith. sign in

REVIEW 4 major objections 7 minor 1 cited by

Inverse Design of Nonlinear Mechanics of Bio-inspired Materials Through Interface Engineering and Bayesian Optimization

T0 review · 4 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Bayesian optimization with an expandable design space recovers interface parameters that reproduce a target nonlinear stress-strain curve in nacre-inspired composites, and finds multiple distinct designs that achieve the same curve…

desk verdict Solid BO-for-inverse-design paper with a clean validation experiment; the geometry-sensitivity claim overreaches beyond the tested parameter regime but the method deserves peer review. read the letter →

arxiv 2412.14071 v1 pith:3CEVBLMO submitted 2024-12-18 cond-mat.mtrl-sci physics.app-ph

Wei Zhang , Mingjian Tang , Haoxuan Mu , Xingzi Yang , Xiaowei Zeng , Rui Tuo , Wei (Wayne) Chen , Wei Gao This is my paper · ORCID
classification cond-mat.mtrl-sciphysics.app-ph MSC 74P1074S0562F15
keywords Bayesianoptimizationinversedesignnacre-inspiredcompositeinterfaceengineeringtraction-separationlawstress-straincurvespaceexpansionnon-uniquesolutions
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

Given a desired nonlinear stress-strain response, the paper asks whether the soft-interface parameters of a nacre-inspired composite can be recovered computationally. It answers yes: a Bayesian optimization loop, operating on a five-parameter bilinear traction-separation interface law and guided by a pointwise curve-difference metric, finds interface designs whose simulated tensile curves closely match the target. The framework deliberately expands the design space as the search proceeds, so targets that lie outside the initial parameter bounds remain reachable. In the second case study, two top solutions reproduce the same target curve through different mechanisms—one normal-dominated fracture, one shear-dominated fracture. The practical payoff is that designers can choose among several manufacturable interface recipes that meet the same mechanical specification.

What carries the argument

The load-bearing machinery is the pairing of a Gaussian-process surrogate with an expected-improvement acquisition function inside a design-space expansion loop. The objective is a symmetric pointwise curve-difference metric $\Delta_{\mathrm{TS}}$, defined as the average minimum distance from points on the target curve to the simulated curve plus the average minimum distance in the reverse direction; this replaces the worst-case Hausdorff distance, which the paper argues can miss subtle variations. At regular 'stall' intervals the parameter bounds expand by a factor $\sqrt[d]{\gamma/N}$ with $\gamma = N+1$, so early expansions are large and later ones taper, then contract to the minimal box enclosing the old space and the newly sampled designs. The interface itself is a bilinear traction-separation cohesive law with five independent parameters, simulated by cohesive elements in a two-dimensional Voronoi-grain finite-element model.

What would settle it

For Design 1 and Design 2 from Table 2, rerun the same finite-element simulation on several new Voronoi grain patterns (rather than the single geometry used for optimization) and measure the spread of the resulting stress-strain curves; if the spread is comparable to the ~0.01 curve-difference values that define a match, the geometry-insensitivity premise holds, whereas a spread an order of magnitude larger would mean the recovered parameters are artifacts of one grain arrangement.

Watch

Extended reading notes

Core claim

The central claim is that a Bayesian optimization (BO) framework equipped with a controlled, expandable design space solves the inverse interface-design problem for a two-dimensional Voronoi nacre-mimetic composite: given a target stress-strain curve, it recovers the five parameters of the bilinear traction-separation law—the normal and shear interface strengths ($\sigma_n^o$, $\sigma_s^o$), the normal and shear damage-initiation separations ($\delta_n^d$, $\delta_s^d$), and the damage-evolution separation ($\delta_e$)—that reproduce the curve in finite-element simulation. Validation against a ground-truth target outside the initial design space shows that space expansion recovers parameters close to the truth, whereas BO without expansion saturates at the boundary and misses the target. A second target, a purely quadratic curve not generated by FEM, yields top solutions that separate into two clusters differing by two orders of magnitude in shear strength; these designs give nearly indistinguishable stress-strain responses while failing by different mechanisms, one normal-dominated and one shear-dominated. The paper thereby claims that inverse design of nonlinear mechanics can be formulated as a tractable optimization problem and that non-uniqueness in interface design is not a nuisance but a resource.

Load-bearing premise

The load-bearing premise is that the two-dimensional Voronoi finite-element model with a bilinear traction-separation interface law faithfully represents the physical material; if the macroscopic response of the real composite is sensitive to grain arrangement or to interlayer shear that the 2D model cannot capture, the recovered interface parameters will not transfer to a fabricated specimen.

Editorial extensions

If this is right

  • If the framework is correct, any specified nonlinear stress-strain curve that lies within the expressive range of the traction-separation law can be treated as a design target, not only extremal objectives like maximum strength or toughness.
  • Because the design space expands and contracts during the search, a user does not need to know the feasible bounds of interface parameters in advance.
  • The two recovered solution clusters show that multiple manufacturable interface designs can meet the same mechanical specification, so cost, processability, or preferred failure mode can be used as secondary selection criteria.
  • The same acquisition-and-expansion loop can be coupled to other forward solvers, such as discrete element models, and to other inverse design objectives beyond interface parameters.

Reading between the lines

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

  • A natural extension the paper does not explore is to mask the curve-difference metric so that only a portion of the stress-strain response (e.g., the post-peak softening tail) is targeted, which would let the same loop design interfaces for energy absorption without dictating the elastic regime.
  • The sharp drop in objective improvement after roughly 45 iterations suggests that a budget-allocation strategy—stopping early and spending the saved FEM evaluations on diverse candidates—could preserve solution diversity at a fraction of the computational cost; the paper notes the plateau but keeps the full 150-iteration budget.
  • The two design clusters imply a broader design rule for nacre-like composites: low-shear-strength and high-normal-strength interfaces spread damage broadly, while the converse concentrates damage into a few normal cracks; this trade-off could inform selections for damage tolerance versus predictable failure, though the paper does not make that application.
  • Because the framework treats the simulator as ground truth, its success on the FEM-generated target does not guarantee success on experimental data; a closed-loop version that replaces the simulator with physical measurements would be the decisive next step.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. This paper addresses the inverse design of interface parameters in a two-dimensional, Voronoi-tessellated nacre-inspired composite. The interface is modeled with a five-parameter bilinear traction-separation law (σ_n^o, σ_s^o, δ_n^d, δ_s^d, δ_e), and the forward response is computed by finite element analysis. The authors propose a Bayesian optimization (BO) framework with an expandable design space, using Expected Improvement as the acquisition function and a two-way pointwise curve-distance metric between the target and simulated stress-strain curves. They validate the method on a target generated from known parameters lying outside the initial design range, showing that the expansion recovers parameters close to ground truth, whereas a fixed design space pins at the boundary. They then demonstrate for a quadratic target curve that BO identifies two distinct interface designs with nearly identical stress-strain responses but different failure mechanisms (normal-dominated vs. shear-dominated fracture). The paper includes a geometry-sensitivity study in the Supporting Information and makes code and data available on GitHub.

Significance. If the claims hold, the paper makes a useful contribution by demonstrating that BO with a controlled, expandable design space can solve a class of inverse problems for nonlinear mechanical responses, and by showing that multiple physically distinct interface designs can produce nearly the same macroscopic response. The validation design (known ground truth outside the initial range) is clean, and the comparison with the no-expansion baseline clearly illustrates the benefit of expansion. The GitHub availability of code and data is a strength. However, as detailed in the major comments, the robustness of the central claims is currently limited by the narrow scope of the geometry-sensitivity study and the absence of repeated runs, and the non-uniqueness claim would be stronger if demonstrated for an attainable target.

major comments (4)
  1. [Model Material and Interface Law (main text) and Tables S1, 1, 2] The geometry-insensitivity claim, used in the main text to justify the use of a single Voronoi geometry for all BO runs, is established only for three moderate interface parameter sets (Table S1: strengths 125–200 MPa, separations 2–4 nm, δ_e = 15 nm). The optimized designs reported in Tables 1 and 2 lie far outside this regime: the validation solution has δ_e ≈ 70 nm and δ_s^d ≈ 10–12 nm, Design 1 has δ_e = 65.38 nm and δ_s^d = 0.99 nm, and Design 2 has σ_s^o = 1.00 MPa, δ_s^d = 6.78 nm, and δ_e = 80.02 nm. Because damage localization and crack paths are known to be sensitive to interface strength and softening behavior in this range, the statement that the response is insensitive to grain distribution at the current grain density has not been demonstrated for the parameter values that the optimization actually returns. Since every BO iteration uses geometry 4, the reported optimal parameters and the two-group failure-mode split could be artifacts of a single Voronoi realization. I request a geometry-sensitivity study that samples interface parameters from the optimized regions (including small σ_s^o and large δ_e), or an explicit statement that the optimized designs are only established for geometry 4.
  2. [Results: Validation and Non-unique designs; Figs. 3 and 4] The central quantitative results are based on a single BO run per case, with one initial dataset of 50 curves and one geometry. The BO procedure is stochastic: the initial dataset is randomly generated, Gaussian-process hyperparameters are fitted, and the acquisition function is maximized numerically. The reported plateau of the objective after 45 iterations (Fig. 3C) and the appearance of two distinct groups among the top ten designs (Fig. 4F and Table S2) could depend on the particular run. Repeating the optimization with several random seeds or initial datasets would establish the robustness of these conclusions; without such repetitions, the claims of efficiency and of a natural split into two failure-mode groups are not fully supported.
  3. [Methods: Algorithm 1] The expansion scheme depends on two free parameters, the stall number S and the scaling parameter γ, and the refinement step in lines 11–13 contracts the search space to the bounding box of the previous space and the newly acquired designs. The paper adopts γ = N+1 and does not report the value of S or any sensitivity study with respect to these choices. The success of the expansion in the validation case may be sensitive to these parameters, particularly because an over-aggressive refinement could trap the search near the initial boundary, while an over-aggressive expansion could make the GP extrapolation unreliable. Please report the chosen S and include a brief sensitivity analysis (e.g., S ∈ {2, 5, 10} and a slower/faster γ) for the validation case.
  4. [Results: Non-unique Interface Designs; Fig. 4 and Table S2] The target curve in Case 2 is a quadratic function that is not known to be exactly representable by the five-parameter bilinear traction-separation model. The best objective values are 0.010–0.0176 (in the un-normalized units of Eq. (4)), and the paper does not establish whether the residual reflects optimization error or model mismatch. Consequently, the two distinct parameter groups may simply be two local minima for an unattainable target rather than genuinely non-unique solutions for an attainable target. To support the non-uniqueness claim, I recommend demonstrating with a target generated from a known parameter set (e.g., another ground-truth design) that BO, when run with different initialization or design-space expansion, recovers two or more distinct parameter sets that reproduce the target within the same tolerance. Alternatively, the paper should explicitly frame the result as approximate matching with non-unique local minima.
minor comments (7)
  1. [Results (text near Fig. 2F)] Typo: 'feture' should be 'feature'.
  2. [Methods, Eq. (4)] Please specify the discretization counts N_T and N_S and the curve-interpolation procedure; the current definition leaves the metric's scale and normalization ambiguous.
  3. [Fig. 4 caption] Missing space in 'BO.(A)'; also, the caption does not identify which solution is Design 1 and which is Design 2.
  4. [Supporting Information, Fig. S1] The text selects 'geometry 4' but does not specify which of the five patterns in Fig. S1 corresponds to geometry 4; please label or reference it.
  5. [References] Reference [29] is incomplete: it lacks a journal or venue; please provide full bibliographic details.
  6. [Methods, Algorithm 1] The stall logic in Algorithm 1 is not immediately transparent: line 4 expands when t%S==1 and line 11 updates on t%S==0; a short clarification in the text or a pseudocode comment would help.
  7. [Results: Computational budget] The paper does not report the number of FEM simulations needed to generate the 50-curve initial dataset or the per-simulation computational cost; stating these would help readers assess the practical budget for larger problems.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the BO inverse-design benchmark, the external quadratic target, and the failure-mode analysis are self-contained.

full rationale

The paper's derivation chain is a standard inverse-design loop: (i) a bilinear traction-separation interface law and a 2D Voronoi FEM model are defined; (ii) Bayesian optimization with Gaussian-process regression and expected improvement minimizes a curve-difference metric ΔTS between target and FEM-computed stress-strain curves; (iii) in Case 1 the target curve is generated by the same FEM with a hidden parameter set outside the initial design space, and the expanded-space BO solution is compared to that hidden ground truth; (iv) in Case 2 the target is an externally specified quadratic function with no known interface parameters, and the reported optimal designs are FEM outputs assessed by the same ΔTS metric. None of these steps defines the target in terms of the fitted parameters or vice versa, and the final stress-strain curves and damage/failure mechanisms are emergent FEM results rather than quantities used as fitting labels. The geometry-sensitivity study in the Supporting Information is an independent scan over five Voronoi patterns and three interface property sets, used only to justify the single-geometry choice; it does not smuggle in the optimization outcome, although its tested parameter range is narrower than the optimized designs (a correctness/external-validity concern, not circularity). The BO algorithm itself explicitly builds on standard external methods for Gaussian processes, expected improvement, and expandable search spaces from references [34], [35], [43], and [44]; no load-bearing conclusion rests on a self-citation or an author-imported uniqueness theorem. The only author-associated reference ([41]) is mentioned as a future extension and plays no role in the derivation. Thus there is no step that reduces, by the paper's own equations or by self-citation, to its own inputs, and no fitted parameter is renamed as a prediction.

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

The central claim rests on the FEM cohesive-zone model as ground truth and on several unreported algorithmic choices. The paper introduces no new physical entities. The validation benchmark uses the same model as the evaluator, which strengthens internal consistency but not external grounding.

free parameters (5)
  • Initial dataset size = 50
    The BO cycle starts from 50 random FEM curves; the reported results depend on this sample size, which is chosen without stated justification.
  • Stall number S = not reported (inferred as about 15 from 10 expansions within 150 iterations)
    Controls how often the design space expands and is central to the expansion behavior, yet it is never stated explicitly in the text.
  • Expansion scaling policy = gamma = N + 1
    The growth of the design space is set by this hand-chosen schedule; other schedules would change the search trajectory and the discovered non-unique designs.
  • GP kernel and hyperparameters = not reported
    The Gaussian process surrogate requires a kernel, length scales, and a noise model; none are specified, yet they determine the acquisition function and therefore the proposed designs.
  • Curve discretization N_T, N_S = not reported
    The curve difference metric in Eq. (4) depends on the number of sample points after uniform interpolation, which is never given.
assumptions (5)
  • domain assumption The bi-linear traction-separation cohesive law (Eqs. 1-3) adequately represents the mechanical behavior of the soft interfaces.
    Invoked in 'Model Material and Interface Law'; if real interfaces behave differently, for example with rate dependence or nonlinear unloading, the identified parameters lose physical meaning.
  • domain assumption A single 2D Voronoi geometry (geometry 4) is representative of the composite response at the optimal parameter sets.
    The sensitivity study (Supplementary Note 1) tests only five patterns and three material sets; all BO optimization results then use geometry 4.
  • domain assumption The chosen Voronoi-based 2D model captures the essential deformation mechanisms of nacre and bone extrafibrillar matrix.
    Stated in the Results section; the Discussion concedes that the 2D model cannot capture interlayer shear and staggered layering present in real nacre.
  • domain assumption The curve difference metric Delta_TS is a suitable objective for matching nonlinear stress-strain curves.
    Eq. (4) is introduced without a convergence study or a comparison to alternative metrics beyond the Hausdorff distance.
  • ad hoc to paper A physical parameter combination exists that can reproduce the arbitrary quadratic target curve in Case 2.
    The second target is not generated from FEM; the paper never verifies that any interface parameter set can produce this curve exactly, only that BO finds low-difference approximations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Inverse Design of Nonlinear Mechanics of Bio-inspired Materials Through Interface Engineering and Bayesian Optimization." pith.science (2026). https://pith.science/paper/3CEVBLMO

@misc{pith2026241214071,
  author       = {Pith},
  title        = {Pith review of: Inverse Design of Nonlinear Mechanics of Bio-inspired Materials Through Interface Engineering and Bayesian Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3CEVBLMO}},
  note         = {Machine review of arXiv:2412.14071}
}
read the original abstract

In many biological materials such as nacre and bone, the material structure consists of hard grains and soft interfaces, with the interfaces playing a significant role in the material's mechanical behavior. This type of structures has been utilized in the design of various bio-inspired composite materials. Such applications often require the materials to exhibit a specified nonlinear stress-strain relationship. A key challenge lies in identifying appropriate interface properties from an infinite search space to achieve a given target stress-strain curve. This study introduces a Bayesian optimization (BO) framework specifically tailored for the inverse design of interfaces in bio-inspired composites. As a notable advantage, this method is capable of expanding the design space, allowing the discovery of optimal solutions even when the target curve deviates significantly from the initial dataset. Furthermore, our results show that BO can identify distinct interface designs that produce similar target stress-strain responses, yet differ in their deformation and failure mechanisms. These findings highlight the potential of the proposed BO framework to address a wide range of inverse design challenges in nonlinear mechanics problems.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Data-efficient inverse design of spinodoid metamaterials

    cs.CE 2025-05 conditional novelty 6.0 of 10

    A permutation-equivariant neural network with built-in elasticity constraints predicts the stiffness tensor of spinodoid metamaterials from 75 training samples and solves multi-objective inverse design tasks.

Reference graph

Works this paper leans on

44 extracted references · 31 canonical work pages · cited by 1 Pith paper

  1. [1]

    Barthelat, H

    F. Barthelat, H. Tang, P. D. Zavattieri, C. M. Li, H. D. Espinosa, On the mechanics of mother- of-pearl: A key feature in the material hierarchical structure. Journal of the Mechanics and Physics of Solids 55, 306–337 (2007), doi:10.1016/J.JMPS.2006.07.007

    work page doi:10.1016/j.jmps.2006.07.007 2007

  2. [2]

    J.-Y. Rho, L. Kuhn-Spearing, P. Zioupos, Mechanical properties and the hierarchical structure of bone. Medical engineering & physics 20 (2), 92–102 (1998)

    work page 1998

  3. [3]

    Browning, C

    A. Browning, C. Ortiz, M. C. Boyce, Mechanics of composite elasmoid fish scale assemblies and their bioinspired analogues. Journal of the mechanical behavior of biomedical materials 19, 75–86 (2013)

    work page 2013

  4. [4]

    Huang, et al., Multiscale toughening mechanisms in biological materials and bioinspired designs

    W. Huang, et al., Multiscale toughening mechanisms in biological materials and bioinspired designs. Advanced Materials 31 (43), 1901561 (2019)

    work page 2019

  5. [5]

    A. K. Dastjerdi, R. Rabiei, F. Barthelat, The weak interfaces within tough natural composites: Experiments on three types of nacre. Journal of the Mechanical Behavior of Biomedical Materials 19, 50–60 (2013), doi:10.1016/J.JMBBM.2012.09.004

    work page doi:10.1016/j.jmbbm.2012.09.004 2013

  6. [6]

    Barthelat, Z

    F. Barthelat, Z. Yin, M. J. Buehler, Structure and mechanics of interfaces in biological materials. Nature Reviews Materials 2016 1:41, 1–16 (2016), doi:10.1038/natrevmats.2016.7

    work page doi:10.1038/natrevmats.2016.7 2016

  7. [7]

    Madhav, B

    D. Madhav, B. Buffel, P. Moldenaers, F. Desplentere, V. Vandeginste, A review of nacre- inspired materials: Chemistry, strengthening-deformation mechanism, synthesis, and appli- cations. Progress in Materials Science 139, 101168 (2023), doi:10.1016/J.PMATSCI.2023. 101168

    doi:10.1016/j.pmatsci.2023 2023

  8. [8]

    Yang, et al

    Y. Yang, et al. , Electrically assisted 3D printing of nacre-inspired structures with self- sensing capability. Science Advances 5 (2019), doi:10.1126/SCIADV.AAU9490/SUPPL FILE/AAU9490 SM.PDF

    work page doi:10.1126/sciadv.aau9490/suppl 2019

Show all 44 references
  1. [9]

    Peng, et al., Inverse Nacre-like Epoxy-Graphene Layered Nanocomposites with Integration of High Toughness and Self-Monitoring.Matter 2, 220–232 (2020), doi:10.1016/j.matt.2019

    J. Peng, et al., Inverse Nacre-like Epoxy-Graphene Layered Nanocomposites with Integration of High Toughness and Self-Monitoring.Matter 2, 220–232 (2020), doi:10.1016/j.matt.2019. 08.013. 20

    2020 doi

  2. [10]

    X. Hu, Z. Xu, C. Gao, Multifunctional, supramolecular, continuous artificial nacre fibres. Scientific Reports 2012 2:1 2, 1–8 (2012), doi:10.1038/srep00767

    2012 doi

  3. [11]

    Hu, et al., A novel wet-spinning method of manufacturing continuous bio-inspired com- posites based on graphene oxide and sodium alginate

    X. Hu, et al., A novel wet-spinning method of manufacturing continuous bio-inspired com- posites based on graphene oxide and sodium alginate. Nano Research 9, 735–744 (2016), doi:10.1007/S12274-015-0952-2/METRICS

    2016 doi

  4. [12]

    Y. X. Yan, et al. , Micrometer-Thick Graphene Oxide–Layered Double Hydroxide Nacre- Inspired Coatings and Their Properties. Small 12, 745–755 (2016), doi:10.1002/SMLL. 201502061

    2016 doi

  5. [13]

    Pan, et al., Double-Layer Nacre-Inspired Polyimide-Mica Nanocomposite Films with Excellent Mechanical Stability for LEO Environmental Conditions

    X.-F. Pan, et al., Double-Layer Nacre-Inspired Polyimide-Mica Nanocomposite Films with Excellent Mechanical Stability for LEO Environmental Conditions. Advanced Materials 34, 2105299 (2022), doi:10.1002/ADMA.202105299

    2022 doi

  6. [14]

    Z. Qin, et al., Nacrelike-structured multilayered polyelectrolyte/calcium carbonate nanocom- posite membrane via Ca-incorporated layer-by-layer-assembly and CO2-induced biomineral- ization. Journal of Membrane Science 498, 180–191 (2016), doi:10.1016/J.MEMSCI.2015. 09.016

    2016 doi

  7. [15]

    Y. Ni, Z. Song, H. Jiang, S. H. Yu, L. He, Optimization design of strong and tough nacreous nanocomposites through tuning characteristic lengths. Journal of the Mechanics and Physics of Solids 81, 41–57 (2015), doi:10.1016/j.jmps.2015.04.013

    2015 doi

  8. [16]

    L. C. Zhou, Y. B. Zhu, Z. Z. He, X. Jin, H. A. Wu, Multi-parameter structural optimization to reconcile mechanical conflicts in nacre-like composites.Composite Structures 259 (2021), doi:10.1016/j.compstruct.2020.113225

    2021

  9. [17]

    L. C. Zhou, Z. Z. He, Z. T. Zhang, Y. B. Zhu, H. A. Wu, Maximum utilization of nacre- mimetic composites by architecture manipulation and interface modification towards critical damage state. Composites Science and Technology 233 (2023), doi:10.1016/j.compscitech. 2022.109893. 21

    2023

  10. [18]

    K. Radi, D. Jauffres, S. Deville, C. L. Martin, Strength and toughness trade-off optimization of nacre-like ceramic composites. Composites Part B: Engineering 183, 107699 (2020), doi: 10.1016/J.COMPOSITESB.2019.107699

    2020

  11. [19]

    N. S. Al-Maskari, D. A. McAdams, J. N. Reddy, Modeling of a biological material nacre: Multi-objective optimization model. Mechanics of Advanced Materials and Structures 28, 430–439 (2021), doi:10.1080/15376494.2019.1568649

    2021

  12. [20]

    K. Park, C. Song, J. Park, S. Ryu, Multi-objective Bayesian optimization for the design of nacre-inspired composites: optimizing and understanding biomimetics through AI.Materials Horizons 10, 4329–4343 (2023), doi:10.1039/D3MH00137G

    2023 doi

  13. [21]

    Xu, et al., Nacre-inspired tunable strain sensor with synergistic interfacial interaction for sign language interpretation.Nano Energy90, 106606 (2021), doi:10.1016/J.NANOEN.2021

    W. Xu, et al., Nacre-inspired tunable strain sensor with synergistic interfacial interaction for sign language interpretation.Nano Energy90, 106606 (2021), doi:10.1016/J.NANOEN.2021. 106606

    2021 doi

  14. [22]

    Wang, et al., Flexible multiresponse-actuated nacre-like MXene nanocomposite for wear- able human-machine interfacing

    L. Wang, et al., Flexible multiresponse-actuated nacre-like MXene nanocomposite for wear- able human-machine interfacing. Matter 5 (10), 3417–3431 (2022)

    2022

  15. [23]

    Shahriari, K

    B. Shahriari, K. Swersky, Z. Wang, R. P. Adams, N. D. Freitas, Taking the human out of the loop: A review of Bayesian optimization. Proceedings of the IEEE 104, 148–175 (2016), doi:10.1109/JPROC.2015.2494218

    2016

  16. [24]

    Kuszczak, F

    I. Kuszczak, F. I. Azam, M. A. Bessa, P. J. Tan, F. Bosi, Bayesian optimisation of hexagonal honeycomb metamaterial.Extreme Mechanics Letters64, 102078 (2023), doi:10.1016/J.EML. 2023.102078

    2023

  17. [25]

    T. Xue, T. J. Wallin, Y. Menguc, S. Adriaenssens, M. Chiaramonte, Machine learning gen- erative models for automatic design of multi-material 3D printed composite solids. Extreme Mechanics Letters 41, 100992 (2020), doi:10.1016/J.EML.2020.100992

    2020

  18. [26]

    Ma, et al., Probabilistic Representation and Inverse Design of Metamaterials Based on a Deep Generative Model with Semi-Supervised Learning Strategy

    W. Ma, et al., Probabilistic Representation and Inverse Design of Metamaterials Based on a Deep Generative Model with Semi-Supervised Learning Strategy. Advanced Materials 31, 1901111 (2019), doi:10.1002/ADMA.201901111. 22

    2019 doi

  19. [27]

    Noh, et al., Inverse Design of Solid-State Materials via a Continuous Representation.Matter 1, 1370–1384 (2019), doi:10.1016/j.matt.2019.08.017

    J. Noh, et al., Inverse Design of Solid-State Materials via a Continuous Representation.Matter 1, 1370–1384 (2019), doi:10.1016/j.matt.2019.08.017

    2019 doi

  20. [28]

    J. H. Bastek, D. M. Kochmann, Inverse design of nonlinear mechanical metamaterials via video denoising diffusion models. Nature Machine Intelligence 5, 1466–1475 (2023), doi: 10.1038/s42256-023-00762-x

    2023 doi

  21. [29]

    Jin, et al

    H. Jin, et al. , Mechanical Characterization and Inverse Design of Stochastic Architected Metamaterials Using Neural Operators (2023)

    2023

  22. [30]

    Rousseau, et al., Sheet nacre growth mechanism: a Voronoi model

    M. Rousseau, et al., Sheet nacre growth mechanism: a Voronoi model. Journal of Structural Biology 149, 149–157 (2005), doi:10.1016/J.JSB.2004.09.005

    2005 doi

  23. [31]

    Y. Han, et al., Removal of glycosaminoglycans affects the in situ mechanical behavior of extrafibrillar matrix in bone.journal of the mechanical behavior of biomedical materials123, 104766 (2021)

    2021

  24. [32]

    P. P. Camanho, C. G. Davila, Mixed-Mode Decohesion Finite Elements for the Simulation of Delamination in Composite Materials (2002), http://www.sti.nasa.gov

    2002

  25. [33]

    D. P. Huttenlocher, G. A. Klanderman, W. J. Rucklidge, Comparing Images Using the Haus- dorff Distance. IEEE Transactions on Pattern Analysis and Machine Intelligence15, 850–863 (1993), doi:10.1109/34.232073

    1993 doi

  26. [34]

    Schulz, M

    E. Schulz, M. Speekenbrink, A. Krause, A tutorial on Gaussian process regression: Modelling, exploring, and exploiting functions. Journal of Mathematical Psychology 85, 1–16 (2018), doi:10.1016/j.jmp.2018.03.001

    2018 doi

  27. [35]

    D. R. Jones, M. Schonlau, W. J. Welch, Efficient global optimization of expensive black-box functions. Journal of Global optimization 13, 455–492 (1998)

    1998

  28. [36]

    Zhang, S

    N. Zhang, S. Yang, L. Xiong, Y. Hong, Y. Chen, Nanoscale toughening mechanism of nacre tablet. Journal of the Mechanical Behavior of Biomedical Materials 53, 200–209 (2016), doi:10.1016/J.JMBBM.2015.08.020. 23

    2016 doi

  29. [37]

    Otsuki, M

    N. Otsuki, M. Furukawa, T. Sugio, Adhesive analysis using molecular dynamics simulations. Technical News 98, ThreeBond(2021)

    2021

  30. [38]

    J. Shi, M. J. Quevillon, P. H. A. Valenc ¸a, J. K. Whitmer, Predicting Adhesive Free Energies of Polymer-Surface Interactions with Machine Learning. ACS Applied Materials and Inter- faces 14, 37161–37169 (2022), doi:10.1021/ACSAMI.2C08891/ASSET/IMAGES/LARGE/ AM2C08891 0009.JPEG

    2022 doi

  31. [39]

    N. Abid, M. Mirkhalaf, F. Barthelat, Discrete-element modeling of nacre-like materials: Effects of random microstructures on strain localization and mechanical performance.Journal of the Mechanics and Physics of Solids 112, 385–402 (2018), doi:10.1016/J.JMPS.2017.11. 003

    2018 doi

  32. [40]

    A. N. Karuriya, F. Barthelat, Plastic deformations and strain hardening in fully dense granular crystals. Journal of the Mechanics and Physics of Solids 186, 105597 (2024), doi:10.1016/J. JMPS.2024.105597

    2024

  33. [41]

    L. Lin, X. Wang, X. Zeng, Computational modeling of interfacial behaviors in nanocomposite materials. International Journal of Solids and Structures115-116, 43–52 (2017), doi:10.1016/ j.ijsolstr.2017.02.029

    2017

  34. [42]

    K. Ko, S. Jin, S. E. Lee, J.-W. Hong, Impact resistance of nacre-like composites diversely patterned by 3D printing. Composite Structures 238, 111951 (2020)

    2020

  35. [43]

    Nguyen, S

    V. Nguyen, S. Gupta, S. Rane, C. Li, S. Venkatesh, Bayesian Optimization in Weakly Specified Search Space, in 2017 IEEE International Conference on Data Mining (ICDM) (2017), pp. 347–356, doi:10.1109/ICDM.2017.44

    2017 doi

  36. [44]

    Shahriari, A

    B. Shahriari, A. Bouchard-Cote, N. Freitas, Unbounded Bayesian Optimization via Regular- ization, in Proceedings of the 19th International Conference on Artificial Intelligence and Statistics, A. Gretton, C. C. Robert, Eds. (PMLR, Cadiz, Spain), vol. 51 of Proceedings of Machi...

    2016

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.