Pith. sign in

REVIEW 5 major objections 7 minor 53 references

Neural networks for the prediction of peel force for skin adhesive interface using FEM simulation

T0 review · 5 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A neural network predicts peel force for skin adhesives from material and fracture parameters alone.

desk verdict A standard NN surrogate for FEM peel simulations, applied to skin-adhesive systems; the narrow claim checks out, but the paper overreaches to physical prediction without experimental validation. read the letter →

arxiv 2506.19855 v1 pith:DJ5K6FBR submitted 2025-06-09 physics.med-ph cs.LG

Ashish Masarkar , Rakesh Gupta , Naga Neehar Dingari , Beena Rai This is my paper
classification physics.med-phcs.LG
keywords peelforceneuralnetworksurrogateskin-adhesiveinterfacefiniteelementmethodcohesivezonemodel90-degreetestvisco-hyperelasticadhesivemachinelearningbiomechanics
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

Peel force is the practical measure of how strongly a medical adhesive sticks to skin, and it is normally obtained by running time-consuming experiments or finite-element simulations—each FEM run in this paper takes 20–40 minutes. The paper argues that this cost is unnecessary once a surrogate model is trained: a feedforward neural network learns the mapping from the adhesive's visco-hyperelastic material parameters and the interface's cohesive-zone parameters to the minimum peel force $F_{\min}$ of a 90° peel test. On a held-out set of FEM-generated cases the network reaches a mean squared error of $3.65\times10^{-7}$ and an $R^2$ of 0.94, so the authors claim it reproduces simulation accuracy while cutting the time per evaluation to minutes. If true, this makes high-throughput screening of adhesive formulations and transdermal-patch designs feasible. The paper's own claim is the surrogate's accuracy against FEM; no physical peel experiments are used anywhere in the study.

What carries the argument

The load-bearing mechanism is the FEM-to-neural-network surrogate pipeline. A bilinear triangular cohesive-zone model, with fracture energy $G = \tfrac{1}{2}\sigma_{\max}\delta_{\max}$, governs interfacial debonding; the adhesive is a first-order Ogden hyperelastic material with Prony-series viscoelasticity. Systematic variation of these parameters in ANSYS generates a dataset of force–displacement curves, from which the minimum peel force $F_{\min}$ is extracted. The neural network—one hidden layer of 32 neurons, ReLU activations, L2 regularization, dropout, early stopping, and Adam optimization on z-scored inputs—then learns the mapping from the input parameter vector to $F_{\min}$. This two-stage design is what lets the network replace repeated FEM runs.

What would settle it

Run a 90° peel test on ex-vivo or artificial skin with an adhesive whose Ogden, Prony, and cohesive-zone parameters have been independently characterized, and compare the measured minimum peel force with the network's prediction for the same parameter values. If the FEM model carries systematic bias relative to physical peeling, the surrogate will inherit it, so a large disagreement would falsify the paper's practical claim even though the network matches its own FEM data. Within the simulation world, a direct reproducibility check is to hold out fresh parameter combinations, run new ANSYS simulations, and verify that the reported test-set MSE of $3.65\times10^{-7}$ and $R^2=0.94$ are actually achieved on that unseen data.

Watch

Extended reading notes

Core claim

The paper's central claim is that the minimum peel force for adhesive detachment from skin is a smooth, learnable function of the adhesive's constitutive parameters and the interface's fracture properties. Specifically, the authors construct a dataset by running a two-dimensional plane-strain finite-element model of a 90° peel test, varying the first-order Ogden shear modulus $\mu$ and coefficient $\alpha$, a five-term Prony series of relaxation moduli and times, the cohesive-zone maximum normal contact stress $\sigma_{\max}$, and the critical fracture energy $G$. They then train a single-hidden-layer network with 32 ReLU neurons to predict $F_{\min}$ from these inputs. The final model reports MSE $3.65\times10^{-7}$, MAE $4.43\times10^{-4}$, and $R^2=0.94$ on the test set, with averaged 5-fold cross-validation $R^2=0.90$. The authors state this is the first such FEM-plus-machine-learning framework for skin-adhesive peel-force prediction, and that it reduces computational cost from tens of minutes per simulation to near-instant prediction after a one-minute training.

Load-bearing premise

That the finite-element model of the peel test correctly represents real peeling of adhesives from skin; the neural network is trained and tested only on FEM-generated values, and no experimental peel-force measurements appear anywhere in the paper to check this assumption.

Editorial extensions

If this is right

  • Adhesive formulations can be screened in seconds rather than 20–40 minutes per candidate, making high-throughput exploration of the design space practical.
  • The trained network maps the full input space, so designers can invert it (for example, searching for parameter combinations that yield a target peel force) to guide adhesive formulation.
  • The same surrogate pipeline can be extended to other peel geometries, such as 180° or T-peel, and to outputs beyond $F_{\min}$, such as peak force or total work of adhesion, provided FEM data is generated for those cases.
  • Because the model is trained on FEM output, any improvement in FEM fidelity—finer meshing, rate-dependent cohesive zones, or skin anisotropy—can be absorbed by retraining on the improved dataset, preserving the fast-inference advantage.
  • The claim implies that the minimum peel force is a smooth, low-dimensional function of the constitutive and cohesive-zone parameters, which is why a single 32-neuron hidden layer suffices.

Reading between the lines

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

  • If the FEM model is later validated against physical peel experiments, the neural network inherits that validity; conversely, without such validation the reported $R^2$ of 0.94 certifies agreement with ANSYS, not with real skin, which is a separate question the paper does not address.
  • A sensitivity analysis of the trained network would likely show that the cohesive-zone parameters $\sigma_{\max}$ and $G$ dominate $F_{\min}$, which could direct experimental effort toward measuring those two quantities accurately rather than the full Prony series.
  • The same architecture could be trained on multi-fidelity data—coarse FEM plus a modest set of physical experiments—to correct systematic bias, a natural next step the paper leaves implicit.
  • The main text reports 18 input features while the listed feature table sums to fewer, and the dataset size is not stated; a public data table would make the reported MSE and $R^2$ independently checkable, and on a very small dataset the cross-validation scores could be optimistic.
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

5 major / 7 minor

Summary. The manuscript proposes a feedforward neural network to predict the minimum peel force Fmin of a 90-degree peel test at a skin-adhesive interface, using a dataset generated from finite element simulations in ANSYS. The inputs are adhesive Ogden and Prony series parameters, the incompressibility parameter, and cohesive zone parameters (σmax and G); the target is the FEM-computed Fmin. The authors report a final architecture with one hidden layer of 32 neurons, a test-set R² of 0.94 and MSE of 3.65×10⁻⁷, and a 5-fold cross-validation average R² of 0.90. They claim this is a reliable, computationally efficient alternative to repeated FEM simulations and experimental testing.

Significance. If the claims are confined to the statement that a neural network can accurately interpolate the output of a specific FEM model, the work is a standard but potentially useful surrogate-modeling exercise. The reported R² values are consistent with that narrow claim, and the paper includes a reasonably detailed description of the FEM setup and hyperparameter search. However, the paper's actual language repeatedly asserts prediction of physical skin-adhesive peel force and positions the model as an alternative to experimental testing. No experimental peel data on skin, skin surrogate, or any analogous substrate appear anywhere in the manuscript, and the FEM baseline itself is a simplified 2D plane-strain idealization with constant skin parameters, fixed adhesive thickness, and a bilinear CZM. Consequently, the physical predictive claim is not supported. The manuscript also omits several details that are load-bearing for reproducibility and for interpreting the reported accuracy, including the dataset size and an operational definition of Fmin.

major comments (5)
  1. [Data Preparation and Preprocessing / Cross Validation (Tables 2 and 3)] The dataset size is never reported. The paper repeatedly calls the dataset 'limited' and 'small' but gives no number of FEM simulations, no number of training/test samples, and no per-fold sample counts. Without N, the reported 5-fold CV average R² of 0.90 and the fold-level R² spread (Table 3: fold 1 R² = 0.67, folds 4 and 5 R² = 0.99) cannot be interpreted, and the 90/10 random split cannot be assessed. This is a load-bearing omission for the central performance claim.
  2. [Data Preparation and Preprocessing / Model Development] There is a direct inconsistency in the input dimension. Data Preparation states that the FEM simulation data 'consisted of 18 features' and that the three skin Ogden parameters were excluded, leaving the features listed in Table 1. Table 1 lists 15 features: µadhesive, αadhesive, d, g1–g5, t1–t5, σmax, and G. However, Model Development states 'The input layer had 18 features' for the final architecture. Either the input layer description is wrong or Table 1 is incomplete; this must be corrected and the actual feature count confirmed, as it is essential for reproducing the network.
  3. [Finite Element Simulation / Peel Test] The target variable Fmin is not operationally defined. The text says that 'The minimum peel force (Fmin) required for the debonding of the adhesive and substrate was calculated' from each F-d curve, but no equation, algorithm, or criterion is given. In a typical peel test the force-displacement curve may have an initial peak, a transient region, and a steady-state plateau; it is unclear whether Fmin is the global minimum, a local minimum at debonding onset, the minimum of the steady-state region, or some other statistic. Because Fmin is the quantity the neural network is trained to predict, an ambiguous target definition makes the reported MSE and R² uninterpretable as measures of a well-defined quantity.
  4. [Finite Element Simulation / Conclusion] The physical validity of the predicted peel force is unsubstantiated. The neural network is trained and evaluated exclusively on FEM-generated labels, and the FEM model is a simplified 2D plane-strain model with a homogeneous hyperelastic skin (constant Ogden parameters), fixed adhesive thickness, and a bilinear CZM with parameters drawn from literature-based ranges. No experimental peel measurement on skin or a skin surrogate is presented, and the cited experimental FEM validations (Refs. 18–19) concern pressure-sensitive adhesives on other substrates. The abstract and conclusion nevertheless describe the model as predicting 'peel force for skin adhesive interface' and as a 'reliable' alternative to experimental testing. These claims go beyond what the data support. The manuscript should be reframed as a surrogate for the specific FEM model, and all statements about physical skin-adhesive peel force should be qualified accordingly.
  5. [Machine Learning Implementation / SI Table S6] The hyperparameter values are inconsistent between the main text and the supporting information. The main text states that during hyperparameter tuning 'Regularization (L2 penalty) [was] Fixed at 0.01' and 'Dropout rate [was] Set as 20%', whereas SI Table S6 for the final model lists L2 regularization λ = 0.05 and dropout rate 0.4. The main text also describes early stopping with patience 10 during architecture selection but the final model and SI Table S6 use patience 20 and a learning-rate scheduler. These discrepancies must be reconciled so that the reported model configuration is unambiguous.
minor comments (7)
  1. [Title and Abstract] There are typographical errors, including '900 peel test' and '900-peeling' instead of '90° peel test' in the Abstract, Methods, and Figure 2 caption; the degree symbol is also missing in several places.
  2. [Optimizing Architecture (Results)] The text says the single-hidden-layer 32-neuron architecture is 'as shown in Figure 2', but Figure 2 is the peel test geometry and CZM schematic; the MSE/MAE/R² comparison plots are in Figure 3. The figure reference should be corrected.
  3. [Table 2 / Model Performance] The text states that the training MAE is 2.53×10⁻⁴, but Table 2 reports 2.79×10⁻⁴. These values should be checked and made consistent.
  4. [Supporting Information] The SI table numbering is inconsistent: the main text refers to 'supplementary table S1' for parameter combinations, and the SI text refers to 'Table S2', but the actual tables in the SI are labeled Table S4 and Table S5. The table numbering should be harmonized.
  5. [Data Preparation and Preprocessing / Model Performance] The reported MSE and MAE values are not accompanied by units or an explicit statement of whether they were computed on the standardized target or after inverse-transforming to the original scale. Since the target was z-score normalized, the numerical values 3.65×10⁻⁷ and 4.43×10⁻⁴ are not interpretable without this information.
  6. [Data Preparation and Preprocessing] The incompressibility parameter d is defined as D = 1/(2µ) in SI Table S5, making d perfectly collinear with the input µadhesive. Including both as independent features is redundant; the paper should state whether d was used as a separate input or explain the modeling choice.
  7. [Cross Validation / Table 3] The fold-level R² values vary substantially (0.67 to 0.99). Reporting the average R² without a standard deviation or per-fold discussion of the low-performing fold overstates the stability of the model; a brief comment on fold 1 would be useful.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the neural network is an explicit surrogate trained on FEM-generated Fmin labels, and its reported accuracy is fidelity to held-out FEM outputs, not a derivation that reduces to its own inputs.

full rationale

The paper's derivation chain is a standard supervised-learning surrogate construction: finite-element simulations produce force-displacement curves, the minimum peel force Fmin is extracted from those curves as the target label, and a neural network is trained on material and cohesive-zone parameters to reproduce that label. The reported MSE, MAE, and R2 therefore measure how well the network interpolates the FEM-generated dataset, and the paper explicitly frames the goal as 'mimicking the computationally expensive FEM simulations with a machine learning model' rather than as an analytical derivation. No equation in the paper defines Fmin in terms of the network inputs, no fitted parameter is renamed as a prediction, and no load-bearing result is justified solely by a self-citation. The main caveat, correctly noted in the reader's take, is that no experimental peel-force measurements appear anywhere, so physical skin-adhesive peel force is never directly validated; however, that is an external-validity or scope limitation, not circularity. The paper itself acknowledges that 'a direct comparison with traditional methods is not feasible' and lists future work such as larger datasets and other peel angles, which is consistent with an honest surrogate-model study. Accordingly, the circularity score is 0.

Assumptions & free parameters 3 free parameters · 8 assumptions · 0 invented entities

The central claim rests on the fidelity of the FEM model and the representativeness of the chosen material and CZM parameter ranges; the NN itself adds no new physics and its hyperparameters are tuned to the data.

free parameters (3)
  • Neural network hyperparameters = 1 hidden layer, 32 neurons, L2=0.05, dropout=0.4, batch size=8, learning rate=0.001, max epochs=2000, early stopping…
    Selected through grid search on 5-fold validation; directly affect reported MSE and R2 and thus the performance claim.
  • Random train/test split seed = 42
    Controls the single 90/10 split; one specific split may not be representative of model performance.
  • FEM mesh element size = 0.5 mm
    Chosen via mesh convergence study; affects the accuracy of FEM Fmin values used as training targets.
assumptions (8)
  • domain assumption 2D plane strain idealization of the peel test
    SI section FEM Simulation states a simplified plane-strain condition; ignores out-of-plane and edge effects.
  • domain assumption First-order Ogden model for adhesive and skin
    Eq 3 with N=1 for both materials; a higher-order model could change F-d response.
  • domain assumption Five-term Prony series for adhesive viscoelasticity
    Eq 1 with N=5; the relaxation spectrum is limited to this form.
  • domain assumption Bilinear triangular cohesive zone model with sigma_max and G
    Eq 4; alternative traction-separation laws are not tested.
  • domain assumption Skin substrate parameters are constant across all simulations
    Data Preparation excludes skin Ogden parameters because they are kept constant; ignores biological variability.
  • domain assumption sigma_max range 1-10 kPa and G range 23-32 J/m2 are representative of skin-adhesive interfaces
    Ranges taken from literature (refs 50-52); no direct measurement for the modeled system.
  • domain assumption FEM solutions are numerically converged and physically accurate
    Mesh convergence to 0.5 mm; implicit solver; no comparison to experimental peel data for this geometry.
  • domain assumption Minimum peel force Fmin is a well-defined scalar extractable from the F-d curve
    Text describes Fmin as occurring at onset of debonding or transition, but the exact extraction algorithm is not specified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neural networks for the prediction of peel force for skin adhesive interface using FEM simulation." pith.science (2026). https://pith.science/paper/DJ5K6FBR

@misc{pith2026250619855,
  author       = {Pith},
  title        = {Pith review of: Neural networks for the prediction of peel force for skin adhesive interface using FEM simulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DJ5K6FBR}},
  note         = {Machine review of arXiv:2506.19855}
}
read the original abstract

Studying the peeling behaviour of adhesives on skin is vital for advancing biomedical applications such as medical adhesives and transdermal patches. Traditional methods like experimental testing and finite element method (FEM), though considered gold standards, are resource-intensive, computationally expensive and time-consuming, particularly when analysing a wide material parameter space. In this study, we present a neural network-based approach to predict the minimum peel force (F_min) required for adhesive detachment from skin tissue, limiting the need for repeated FEM simulations and significantly reducing the computational cost. Leveraging a dataset generated from FEM simulations of 90 degree peel test with varying adhesive and fracture mechanics parameters, our neural network model achieved high accuracy, validated through rigorous 5-fold cross-validation. The final architecture was able to predict a wide variety of skin-adhesive peeling behaviour, exhibiting a mean squared error (MSE) of 3.66*10^-7 and a R^2 score of 0.94 on test set, demonstrating robust performance. This work introduces a reliable, computationally efficient method for predicting adhesive behaviour, significantly reducing simulation time while maintaining accuracy. This integration of machine learning with high-fidelity biomechanical simulations enables efficient design and optimization of skin-adhesive systems, providing a scalable framework for future research in computational dermato-mechanics and bio-adhesive material design.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages

  1. [1]

    M., Colson, Y

    Fitzgerald, D. M., Colson, Y . L., & Grinstaff, M. W. (2023). Synthetic pressure sensitive adhesives for biomedical applications. Progress in Polymer Science, 142, 101692

    work page 2023

  2. [2]

    O., Li, J., Freedman, B

    Blacklow, S. O., Li, J., Freedman, B. R., Zeidi, M., Chen, C., & Mooney, D. J. (2019). Bioinspired mechanically active adhesive dressings to accelerate wound closure. Science advances, 5(7), eaaw3963

    work page 2019

  3. [3]

    M., Amjadi, M., Yunusa, M., & Sitti, M

    Drotlef, D. M., Amjadi, M., Yunusa, M., & Sitti, M. (2017). Bioinspired composite microfibers for skin adhesion and signal amplificati on of wearable sensors. Advanced materials, 29(28), 1701353

    work page 2017

  4. [4]

    and Zhao, X.,

    Deng, J., Yuk, H., Wu, J., Varela, C.E., Chen, X., Roche, E.T., Guo, C.F. and Zhao, X.,

    work page

  5. [5]

    E., Nabzdyk, C

    Yuk, H., Varela, C. E., Nabzdyk, C. S., Mao, X., Padera, R. F., Roche, E. T., & Zhao, X. (2019). Dry double -sided tape for adhesion of wet tissues and devices. Nature, 575(7781), 169-174

    work page 2019

  6. [6]

    P., & Yasuda, K

    Wada, S., Kitamura, N., Nonoyama, T., Kiyama, R., Kurokawa, T., Gong, J. P., & Yasuda, K. (2016). Hydroxyapatite-coated double network hydrogel directly bondable to the bone: biological and biomechanical evaluations of the bonding property in an osteochondral defect. Acta Biomaterialia, 44, 125-134

    work page 2016

  7. [7]

    and Ma, Y ., Multifunctional Adhesive Hydrogels: From Design to Biomedical Applications

    Tang, S., Feng, K., Yang, R., Cheng, Y ., Chen, M., Zhang, H., Shi, N., Wei, Z., Ren, H. and Ma, Y ., Multifunctional Adhesive Hydrogels: From Design to Biomedical Applications. Advanced Healthcare Materials, p.2403734

    work page

  8. [8]

    , Bailey, A., Spiewak, B., Smith, G

    Turnbull, M.J., Grigsby, I., Unertl, K., Sokol, K., Nordby, T., Liu, C. , Bailey, A., Spiewak, B., Smith, G. and McNulty, A.K., 2024. Comparison of Medical Tape Performance Using Skin Response Quantitative Measurements on Healthy V olunteers. Cureus, 16(3)

    work page 2024

Show all 53 references
  1. [9]

    Rivlin, R. S. (1997). The effective work of adhesion. In Collected Papers of RS Rivlin: Volume I and II (pp. 2611-2614). New York, NY: Springer New York

    1997

  2. [10]

    Kendall, K. (1971). The adhesion and surface energy of elastic solids. Journal of Physics D: Applied Physics, 4(8), 1186

    1971

  3. [11]

    W., & Cordill, M

    Gerberich, W. W., & Cordill, M. J. (2006). Physics of adhesion. Reports on Progress in Physics, 69(7), 2157

    2006

  4. [12]

    Majumder, A., Sharma, A., & Ghatak, A. (2010). A bioinspired wet/dry microfluidic adhesive for aqueous environments. Langmuir, 26(1), 521-525

    2010

  5. [13]

    Patil, S., Mangal, R., Malasi, A., & Sharma, A. (2012). Biomimetic wet adhesion of viscoelastic liquid films anchored on micropatterned elastic substrates. Langmuir, 28(41), 14784-14791

    2012

  6. [14]

    Moore, D. R. (2008). An introduction to the special issue on peel testing. International Journal of Adhesion and Adhesives, 4(28), 153-157

    2008

  7. [15]

    Barquins, M., & Ciccotti, M. (1997). On the kinetics of peeling of an adhesive tape under a constant imposed load. International journal of adhesion and adhesives, 17(1), 65-68

    1997

  8. [16]

    S., Tian, Y ., Zhao, B., Rosenberg , K., Zeng, H., McGuiggan, P.,

    Pesika, N. S., Tian, Y ., Zhao, B., Rosenberg , K., Zeng, H., McGuiggan, P., ... & Israelachvili, J. N. (2007). Peel-zone model of tape peeling based on the gecko adhesive system. The Journal of Adhesion, 83(4), 383-401

    2007

  9. [17]

    Yang, Z., Zhu, Z., Yao, C., Xia, Y ., Chen, K., & Jiang, H. (2022). A rate -dependent cohesive zone model for adhesive damage considering fibrillation. International Journal of Mechanical Sciences, 234, 107697

    2022

  10. [18]

    K., Charalambides, M

    Mohammed, I. K., Charalambides, M. N., & Kinloch, A. J. (2015). Modelling the interfacial peeling of pressure-sensitive adhesives. Journal of Non -Newtonian Fluid Mechanics, 222, 141-150

    2015

  11. [19]

    K., Charalambides, M

    Mohammed, I. K., Charalambides, M. N., & Kinloch, A. J. (2016). Modeling the effect of rate and geometry on peeling and tack of pressure -sensitive adhesives. Journal of Non-Newtonian Fluid Mechanics, 233, 85-94

    2016

  12. [20]

    J., Grellmann, W., Geißler, G., & Kaliske, M

    Nase, M., Langer, B., Baumann, H. J., Grellmann, W., Geißler, G., & Kaliske, M. (2009). Evaluation and simulation of the peel behavior of polyethylene/polybutene ‐1 peel systems. Journal of applied polymer science, 111(1), 363-370

    2009

  13. [21]

    Yin, H., Peng, Z., & Chen, S. (2023). The peeling behavior of a heterogeneous elastic film on a rigid substrate. International Journal of Solids and Structures, 285, 112529

    2023

  14. [22]

    Garg, R., & Datla, N. V . (2019). Peeling of heterogeneous thin films: Effect of bending stiffness, adhesion energy, and level of heterogeneity. The Journal of Adhesion, 95(3), 169-186

    2019

  15. [23]

    B., Chen, S

    Yin, H. B., Chen, S. H., Liang, L. H., Peng, Z. L., & Wei, Y . G. (2018). Quantitative prediction of the whole peeling process of an elastic film on a rigid substrate. Journal of Applied Mechanics, 85(9), 091004

    2018

  16. [24]

    Özdemir, İ. (2017). A numerical solution framework for simultaneous peeling of thin elastic strips from a rigid substrate. Acta Mechanica, 228, 1735-1747

    2017

  17. [25]

    Adjeloua, A., Boualem, N., Meddah, H., & Belarbi, A. (2021). Numerical simulation of peel test for ductile thin film along ceramic substrate: Elasto-plastic analysis. Journal of Mechanical Engineering and Sciences, 15(1), 7762-7770

    2021

  18. [26]

    Peng, Z., & Peng, Z. (2015). Peeling Behavior of a Bio -inspired Nanofilm with Finite Length on a Rigid Substrate. Bio-inspired Studies on Adhesion of a Thin Film on a Rigid Substrate, 27-43

    2015

  19. [27]

    Mbithi, F., & Worsley, P. R. (2023). Adhesives for medical applic ation-Peel strength testing and evaluation of biophysical skin response. Journal of the Mechanical Behavior of Biomedical Materials, 148, 106168

    2023

  20. [28]

    Yazdi, S. J. M., & Baqersad, J. (2022). Mechanical modeling and characterization of human skin: A review. Journal of biomechanics, 130, 110864

    2022

  21. [29]

    Ghosh, A., Banerjee, S., Kaity, S., & W Wong, T. (2015). Current pharmaceutical design on adhesive based transdermal drug delivery systems. Current pharmaceutical design, 21(20), 2771-2783

    2015

  22. [30]

    (2016) Skin vs

    Kris,G. (2016) Skin vs. Stainless Steel Adhesion testing for medical design

    2016

  23. [31]

    Yang, Z., Ma, Z., Liu, S., & Li, J. (2021). Tissue adhesion with tough hydrogels: Experiments and modeling. Mechanics of Materials, 157, 103800

    2021

  24. [32]

    G., Geiger, B

    Hoffer, J. G., Geiger, B. C., Ofner, P., & Kern, R. (2021). Mesh-free surrogate models for structural mechanic FEM simulation: A comparative study of approaches. Applied Sciences, 11(20), 9411

    2021

  25. [33]

    A., & Gautam, S

    Gouravaraju, S., Narayan, J., Sauer, R. A., & Gautam, S. S. (2023). A Bayesian regularization-backpropagation neural network model for peeling computations. The Journal of Adhesion, 99(1), 92-115

    2023

  26. [34]

    X., Chen, C

    Gu, G. X., Chen, C. T., Richmond, D. J., & Buehler, M. J. (2018). Bioinspired hierarchical composite design using machine learning: simulation, additive manufacturing, and experiment. Materials Horizons, 5(5), 939-945

    2018

  27. [35]

    Oishi, A., & Yagawa, G. (2020). A surface-to-surface contact search method enhanced by deep learning. Computational Mechanics, 65(4), 1125-1147

    2020

  28. [36]

    X., & Ryu, S

    Kim, Y ., Yang, C., Kim, Y ., Gu, G. X., & Ryu, S. (2020). Designing an adhesive pillar shape with deep learning -based optimization. ACS applied materials & interfaces, 12(21), 24458-24465

    2020

  29. [37]

    Akhavan-Safar, A., Marques, E. A. S., Carbas, R. J. C., & da Silva, L. F. (2023). Predicting adhesive bond perfor mance. In Advances in Structural Adhesive Bonding (pp. 493-523). Woodhead Publishing

    2023

  30. [38]

    Pettersson, S., Engqvist, J., Hall, S., Toft, N., & Hallberg, H. (2019). Peel testing of a packaging material laminate studied by in -situ X-ray tomography and cohesive zo ne modeling. International Journal of Adhesion and Adhesives, 95, 102428

    2019

  31. [39]

    D., Case, S

    Bartlett, M. D., Case, S. W., Kinloch, A. J., & Dillard, D. A. (2023). Peel tests for quantifying adhesion and toughness: A review. Progress in Materials Science , 137, 101086

    2023

  32. [40]

    Y ., & Jeong, H

    Park, S. Y ., & Jeong, H. Y . (2022). Determination of cohesive zone model parameters for tape delamination based on tests and finite element simulations. Journal of Mechanical Science and Technology, 36(11), 5657-5666

    2022

  33. [41]

    Ziari, H., Aliha, M. R. M., Moniri, A. , & Saghafi, Y . (2020). Crack resistance of hot mix asphalt containing different percentages of reclaimed asphalt pavement and glass fiber. Construction and Building Materials, 230, 117015

    2020

  34. [42]

    Ghoreishy, M. H. R. (2012). Determination of the parameters of the Prony series in hyper-viscoelastic material models using the finite element method. Materials & Design, 35, 791-797

    2012

  35. [43]

    Dill, E. H. (2006). Continuum mechanics: elasticity, plasticity, viscoelasticity . CRC press

    2006

  36. [44]

    Joodaki, H., & Panzer, M. B. (2018). Skin mechanical properties and modeling: A review. Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in Medicine, 232(4), 323-343

    2018

  37. [45]

    Y ., Ruina, A., Long, R., & Jagota, A

    Hui, C. Y ., Ruina, A., Long, R., & Jagota, A. (2011). Cohesive zone models and fracture. The Journal of Adhesion, 87(1), 1-52

    2011

  38. [46]

    Song, Z., & Komvopoulos, K. (2013). Delamination of an elastic film from an elastic– plastic substrate during adhesive contact loading and unloading. International Journal of Solids and Structures, 50(16-17), 2549-2560

    2013

  39. [47]

    Qi, Y ., Caillard, J., & Long, R. (2018). Fracture toughness of soft materials with rate- independent hysteresis. Journal of the Mechanics and Physics of Solids, 118, 341-364

    2018

  40. [48]

    Zhang, T., Lin, S., Yuk, H., & Zhao, X. (2015). Predicting fracture energies and crack- tip fields of soft tough materials. Extreme Mechanics Letters, 4, 1-8

    2015

  41. [49]

    Ni, X., Chen, C., & Li, J. (2020). Interfacial fatigue fracture of tissue adhesive hydrogels. Extreme Mechanics Letters, 34, 100601

    2020

  42. [50]

    , Parada, G

    Yuk, H., Zhang, T., Lin, S. , Parada, G. A., & Zhao, X. (2016). Tough bonding of hydrogels to diverse non-porous surfaces. Nature materials, 15(2), 190-196

    2016

  43. [51]

    A., & Zhao, X

    Zhang, T., Yuk, H., Lin, S., Parada, G. A., & Zhao, X. (2017). Tough and tunable adhesion of hydrogels: experiments and models. Acta Mechanica Sinica, 33, 543-554

    2017

  44. [52]

    Load‐Sharing

    Chen, J., Wang, D., Wang, L. H., Liu, W., Chiu, A., Shariati, K., ... & Ma, M. (2020). An adhesive hydrogel with “Load‐Sharing” effect as tissue bandages for drug and cell delivery. Advanced Materials, 32(43), 2001628. Supporting Information Neural networks for the prediction ...

    2020

  45. [2021]

    Nature materials , 20(2), pp.229-236

    Electrical bioadhesive interface for bioelectronics. Nature materials , 20(2), pp.229-236

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.