Pith. sign in

REVIEW 4 major objections 6 minor 76 references

Robust impact localisation on composite aerostructures using kernel design and Bayesian fusion under environmental and operational uncertainties

T0 review · 4 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read The paper claims that impact location on a composite panel can be inferred from sensor arrival-time differences without explicit temperature, mass, or angle compensation, provided the regression model compares the order of the…

desk verdict Useful engineering result with a real internal inconsistency about what sample standardization does to the RBF kernel; worth refereeing after the mass-robustness and generalization evidence are strengthened. read the letter →

arxiv 2501.18393 v1 pith:TCFS2UQS submitted 2025-01-30 stat.AP

Dong Xiao , Zahra Sharif-Khodaei , M. H. Aliabadi This is my paper · ORCID
classification stat.AP MSC 62G0862F1560G15
keywords structuralhealthmonitoringimpactlocalisationenvironmentalandoperationaluncertaintiestimedifferenceofarrivalGaussianprocessregressioncompositekerneldesignBayesianmodelaveragingsamplestandardisation
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

The paper claims that impact location on a composite aircraft panel can be recovered from arrival-time differences without compensating for temperature, impact mass, or impact angle, provided the model compares only the order of those differences rather than their absolute magnitudes. The proposed Gaussian process regression uses a kernel that multiplies a radial-basis term by a cosine-similarity term, so a training impact and a test impact count as similar when their time-difference vectors point in the same direction even if one is a scaled version of the other. On a fibre-reinforced plate heated from 24 °C to 70 °C, models trained on the cooler reference data locate impacts with mean error 6.1 mm for the composite kernel and 7.2 mm for Bayesian model averaging over kernels. The authors conclude that this removes the need for explicit temperature or mass compensation and makes the method suitable for structural health monitoring under real-world uncertainty.

What carries the argument

The load-bearing object is the composite kernel $k_{\mathrm{comp}}(x,x')=k_{\mathrm{rbf}}(x,x')\,k_{\mathrm{cos}}(x,x')$, where the radial-basis factor $\exp(-\|x-x'\|^2/2l_{\mathrm{rbf}}^2)$ measures distance between TDOA vectors and the cosine factor $l_{\mathrm{cos}}^2\,x^{T}x'/(|x||x'|)$ measures their angular alignment. Because TDOA vectors are non-negative, the product stays in $[0,1]$ and is high only when two impacts' sensor time-difference vectors are close in both magnitude and ordering. Sample standardisation $x\mapsto x/\|x\|$ is the preprocessing that preserves this ordering while removing the global scale, which is exactly the ambiguity that temperature and frequency changes introduce. Bayesian model averaging then weights each kernel by the product of its marginal likelihood and the inverse predictive variance at the query point, so the method can switch between distance-like and order-like behaviour automatically.

What would settle it

Measure the six TDOA values at every grid location at 24 °C and 70 °C and check whether the ratio between any two entries stays constant; if any location shows a changed ordering or non-uniform rescaling, as the paper's own Fig. 6 hints, then the cosine kernel's justification fails and the claimed mean error should be recomputed on a continuous temperature ramp rather than a single hot case.

Watch

Extended reading notes

Core claim

The central claim is that the order of time differences of arrival between sensors is invariant under the uncertainties studied: temperature changes and shifts in the dominant frequency each rescale all TDOA values by one positive factor, because low-frequency flexural group velocity is proportional to the square root of frequency. On that basis the authors construct a Gaussian process whose composite kernel is the product of an RBF kernel (sensitive to the magnitude of the TDOA vector) and a cosine kernel (sensitive only to its direction), and they fuse the COS, RBF, and COMP kernels with Bayesian model averaging. The experimental claim is that this combination locates impacts across the plate under a 46 °C temperature rise, a change of impact height, a 45 degree angle change, and a 5.5 kg drop tower impact, using only 24 °C reference impacts as training data, with no explicit compensation. The reported numbers are a mean error of 6.1 mm for the COMP kernel and 7.2 mm for BMA on 70 °C impacts, and the paper states the framework 'achieves accurate localisation without requiring compensation strategies for variations in temperature or impact mass.'

Load-bearing premise

The argument assumes that changing temperature, impact mass, or dominant frequency multiplies every sensor's arrival-time difference by the same positive number, so the ordering of the time differences never changes; the paper itself reports that this is only approximate for temperature.

Editorial extensions

If this is right

  • A model trained once at 24 °C can localise impacts at 70 °C with mean errors of about 6 to 7 mm when six sensors are used, so thermal compensation may not be needed in practice.
  • Sample standardisation of the TDOA input beats feature standardisation for every kernel tested, because it keeps the order of arrival differences intact while removing the scale that the uncertainties alter.
  • Fewer reference impacts still give reasonable interpolation: reducing the training grid from 35 to 15 to 9 points raises the mean temperature-case error from 7.2 mm to 10.9 mm to 12.9 mm, while extrapolation with 9 inner points gives 14.2 mm.
  • The method also transfers across boundary conditions and a 55-fold change in impact mass: a 5.5 kg drop-tower impact is localised from 100 g guided-drop training data with 16.1 mm error using the composite kernel and 19.7 mm using Bayesian model averaging under sample standardisation.

Reading between the lines

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

  • If the single-scaling premise holds for a wider class of changes, such as uniform stiffness reduction, thickness change, or global heating, the same cosine component should transfer the model across those conditions without retraining; a direct test would heat the plate over several temperatures and check that mean error stays near the 6 to 7 mm range.
  • The paper's own temperature measurements in Section 6.2 show the scaling is not uniform at every location, with roughly one-third of impacts deviating from the expected monotonic TDOA increase, so a natural stress test is a continuous temperature ramp rather than a single 70 °C condition.
  • Because the cosine kernel discards magnitude information, the method is insensitive to global scale but also blind to information carried by absolute arrival-time sizes; local stiffness changes or sensor faults that rescale only one channel would need a distance-based component, a case the paper does not examine.
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 / 6 minor

Summary. The paper proposes a Gaussian process regression framework for impact localisation on composite plates using TDOA features. The framework introduces a composite kernel (product of RBF and cosine kernels), a multitask output kernel, Bayesian model averaging over kernels, and sample standardisation of input TDOA vectors. The central premise is that the order of TDOA values across sensors is preserved under variations in temperature, impact mass, and extraction frequency, so that a model trained on 24°C reference impacts can localise impacts under other conditions without explicit compensation. Validation is carried out on a 290×200 mm composite plate with 35 reference locations and additional impacts varying height, angle, temperature (70°C), and impact mass. Reported mean localisation errors under a 46°C temperature shift are 6.1 mm for the COMP kernel and 7.2 mm for BMA with six sensors and 35 references.

Significance. If the order-invariance premise holds, the proposed method is practically attractive for structural health monitoring: it avoids explicit temperature or mass compensation, provides probabilistic predictions, and is validated on a broader experimental grid than many comparable studies, including sensor-count sweeps, reference-density ablations, and interpolation/extrapolation comparisons. The BMA weights are computed from training-data marginal likelihood and test-input predictive variance only, not from test outputs, so the validation is not circular. The paper also usefully demonstrates that sample standardisation outperforms feature standardisation for TDOA inputs and explains the ratio-scale issue behind that result. The main caveat is that the implemented sample standardisation discards input amplitude information, so the method's actual signal is TDOA ordering; the evidence for order preservation under mass and temperature is incomplete and partially contradicted by the paper's own Section 6.2. These issues are fixable with additional analysis or more carefully scoped claims, rather than being fatal to the empirical approach.

major comments (4)
  1. [§4.2 and §7.2, Eq. (17)] With sample standardisation, every input TDOA vector is normalised to unit length: Xss(i,:) = X(i,:)/||X(i,:)||. For unit-norm vectors, the RBF kernel in Eq. (8) becomes exp(-||x-x'||^2/(2 l_rbf^2)) = exp(-(1 - cosθ)/l_rbf^2), a strictly monotone function of the cosine similarity, and the COMP kernel k_comp = k_rbf * k_cos is likewise a function of the angle alone. Therefore the statement in §4.2 item 4 that the RBF kernel 'relies on the magnitudes of TDOA values' is not realised in the experiments, which all use sample standardisation. The paper's interpretation that COMP combines distance/amplitude similarity with order similarity is thus unsupported by the implemented preprocessing; the authors should either run a comparison without sample standardisation or reframe the claims and the interpretation of Sections 7.1, 7.3, and 7.4 as being about pure order-based similarity.
  2. [§6.2 and Eq. (4)] The central theoretical premise, Eq. (4), is that temperature and frequency variations scale all TDOA values by one positive factor, preserving their order. The paper's own data in §6.2 show that the temperature effect is not uniform: only about two-thirds of the 35 impact locations show the expected increase in TDOA at 70°C, and the cosine similarity to REF falls below 0.9 at some central locations with four sensors. Since the implemented preprocessing makes TDOA order the only source of signal, the paper needs a direct analysis of order preservation, such as the fraction of pairwise sensor-order inversions between REF and each of TEM, HEI, ANG, and DT, together with how localisation error varies with those inversions. As written, the order-invariance premise is partially contradicted rather than established.
  3. [§6.1 and §7.4, Table 1 and Fig. 11] The mass-robustness claim is supported by a single drop-tower impact at location D and a single hammer impact at the same location. The errors reported in §7.4 (16.1 mm for COMP and 19.7 mm for BMA with SSI) are point estimates with no repeatability information, so they cannot establish generalisation over impact mass. The authors should either add multi-location, repeated drop-tower and hammer tests, or explicitly scope the mass claim as a single-case demonstration rather than a robust generalisation result.
  4. [§8 and §7.5/§7.6] The conclusion states that 'with four passive sensors and nine reference impacts' the method achieves a mean localisation error of approximately 12.9 mm under temperature variations. However, the 12.9 mm figure is the Int-9RI result reported in §7.5, which appears to use the full six-sensor configuration; Section 7.1 shows that reducing from six to four sensors with 35 reference impacts already increases the BMA mean error to 16.2 mm. No four-sensor/nine-reference experiment is described anywhere in Section 7, so the headline accuracy quoted in the conclusion is not supported by the reported experiments. This should be corrected or the missing experiment should be added.
minor comments (6)
  1. [§5.2 and Table 2] The text says the reference drop height is 1 cm and the HEI height is 2 cm, while Table 2 lists 2.5 cm and 5 cm respectively; please reconcile the text with the table.
  2. [§6.1 and Fig. 5] The text describes TDOA extraction at 0.5 kHz and 1 kHz, but the Figure 5 caption labels panels (b) and (c) as 1 kHz and 2 kHz; also, §6.1 states a threshold of 2.5% of the maximum amplitude, while §6.2 states 0.25%.
  3. [Fig. 11 caption] The caption for Figure 11(b) says 'sample standardisation (FSI)', but the context and the comparison with Figure 11(a) indicate that (b) should be 'feature standardisation (FSI)'.
  4. [§1] The roadmap paragraph contains a duplicated phrase: 'Building on this foundation, Building on this foundation, Section 4 introduces the GPR model'.
  5. [§7.3, Fig. 10 caption] The caption uses the abbreviation SDI without defining it; elsewhere the paper uses SSI for sample-standardised inputs, so the caption should use the consistent abbreviation.
  6. [Table 1 heading] The heading contains the typo 'T able 1'; also, the table formatting for the caption is inconsistent with the rest of the paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the order-invariance premise is physically motivated and independently inspected, and kernel/BMA predictions never use test-location labels.

full rationale

The derivation chain is not circular in the prohibited sense. The central premise, TDOA order invariance under temperature and mass variations, is introduced as a physics-based assumption via Eq. (4), grounded in the standard flexural-wave relation vg ∝ √ω (Eq. 3, citing Abrate [53] as well as the authors' prior [33]), and is then separately checked against measured TEM/HEI/ANG TDOA data in Sections 6.1 and 6.2 using cosine similarity. The kernel and BMA weights are fitted only to REF training data (marginal likelihood) and test-input predictive variance (Eqs. 12-14), never to TEM/HEI/ANG location labels, so there is no fitted-input-called-prediction reduction. The COMP kernel does formally combine RBF and COS terms by Eq. (9); although sample standardisation in Section 7.1 makes the RBF term a monotone function of cosine similarity for unit inputs, which is a legitimate correctness concern about whether the RBF component adds true distance information, it is not a circular claim that the prediction is equivalent to an input by construction. BMA fusion is also not circular: the COMP kernel receives higher weight because of training-data marginal likelihood and test-input predictive variance, not because test outputs are used. The TEM data were produced within the same research group ([26,44]), but this is data provenance, not a self-citation that carries the argument; the localisation results are independent experimental evaluations. No uniqueness theorem is imported from the authors, and no ansatz is smuggled solely through self-citation, since the scaling relation is standard flexural-wave dispersion also supported by an external reference. Score 0 is therefore appropriate.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The method's central claims rest on a small set of fitted hyperparameters and on three domain assumptions about TDOA scaling and order invariance. The paper introduces no new physical entities; the 'singularity point' is a geometric property of the sensor array, not a postulated object.

free parameters (4)
  • GPR kernel lengthscales (l_rbf, l_cos) = Optimized with Adam on training data (Section 4.2)
    Kernel hyperparameters control the GP posterior and are learned from the reference dataset; the paper does not report their final values.
  • Task covariance matrix parameters (ktasks) = Learned during training
    The low-rank inter-task covariance in Eq. (10) is fitted to training data; no values are reported.
  • Observation noise variance (sigma_n^2) = Learned during training
    Standard GPR noise hyperparameter optimized on training data.
  • NSET TDOA extraction threshold = 2.5 percent (Sec. 6.1) or 0.25 percent (Sec. 6.2) of peak amplitude
    The envelope threshold determines extracted TDOA values; the two sections state different values, which affects results.
assumptions (5)
  • domain assumption Flexural wave group velocity is proportional to the square root of frequency at low frequencies, vg proportional to sqrt(omega) (Eq. 3).
    Used in Section 3.2 to derive the scaling property in Eq. (4); an idealized single-mode result from classical laminate theory.
  • domain assumption Temperature changes uniformly scale material properties and hence all TDOA values by a common factor.
    Assumed in Section 3.2; the paper reports deviations in Section 6.2, where only about two-thirds of locations follow the expected trend.
  • domain assumption TDOA order is preserved under large impact mass variations.
    Empirically shown only at one location (D) in Section 6.1, yet used to justify generalization of the method.
  • ad hoc to paper The composite kernel k_comp = k_rbf * k_cos is an appropriate similarity measure for TDOA vectors under uncertainty.
    Introduced as a design choice in Section 4.2 based on qualitative reasoning, not derived from the wave equation.
  • standard math Standard GPR inference equations (Eq. 7) and Bayesian model averaging (Eqs. 11-15) are valid.
    Standard results from Gaussian process theory and BMA, cited as [60,66].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Robust impact localisation on composite aerostructures using kernel design and Bayesian fusion under environmental and operational uncertainties." pith.science (2026). https://pith.science/paper/TCFS2UQS

@misc{pith2026250118393,
  author       = {Pith},
  title        = {Pith review of: Robust impact localisation on composite aerostructures using kernel design and Bayesian fusion under environmental and operational uncertainties},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TCFS2UQS}},
  note         = {Machine review of arXiv:2501.18393}
}
read the original abstract

Impact localisation on composite aircraft structures remains a significant challenge due to operational and environmental uncertainties, such as variations in temperature, impact mass, and energy levels. This study proposes a novel Gaussian Process Regression framework that leverages the order invariance of time difference of arrival (TDOA) inputs to achieve probabilistic impact localisation under such uncertainties. A composite kernel function, combining radial basis function and cosine similarity kernels, is designed based on wave propagation dynamics to enhance adaptability to diverse conditions. Additionally, a task covariance kernel is introduced to enable multitask learning, facilitating the joint prediction of spatial coordinates while capturing interdependencies between outputs. To further improve robustness and accuracy, Bayesian model averaging is employed to dynamically fuse kernel predictions, assigning adaptive weights that account for varying conditions. Extensive experimental validation on a composite plate, including scenarios with large-mass drop tower impacts and small-mass guided drop mass impacts, demonstrates the proposed method's robustness and generalisability. Notably, the framework achieves accurate localisation without requiring compensation strategies for variations in temperature or impact mass, highlighting its suitability for real-world applications. The study also highlights the critical role of sample standardisation for preprocessing TDOA inputs, demonstrating its superiority over feature standardisation by preserving TDOA order invariance and enhancing model compatibility. These advancements establish the proposed method as a reliable and effective solution for structural health monitoring in complex and uncertain operational environments.

Figures

Figures reproduced from arXiv: 2501.18393 by the authors.

Figure 1
Figure 1. Passive sensing for impact monitoring. Passive sensors capture impact responses, forming the basis for reconstructing the impact event. Accurate impact identification under environmental and operational uncertainties is essential to ensure the reliability and effectiveness of practical applications. Ensuring the reliability of the sensing network requires rigorous calibration and testing in both laboratory and opera… view at source ↗
Figure 2
Figure 2. Experimental setup: (a) drop tower impact testing, (b) small hammer impact testing. The primary uncertainties investigated were impact mass and impactor type, given their significant influence on the spectral characteristics of the sensor signals. The composite plate was supported along its two longitudinal edges and clamped at four rect￾angular corners using fixtures beneath the drop tower, as depicted in [PITH_FU… view at source ↗
Figure 3
Figure 3. Layout of the composite plate used for impact testing. Six PZT sensors, numbered S1 to S6, are installed on the surface of the plate, forming a rectangular arrangement. Drop mass or hammer impacts are applied at the centre of the rectangle, while 35 uniformly distributed locations, marked as solid squares, indicate the positions for guided drop mass impacts. These will be further detailed in the following subsection… view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Experimental setup for guided drop mass impact testing, considering uncertainties in impact height, angle and temperature. as detailed in [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 5
Figure 5. Figure 5: (a) compares the power spectral density (PSD) of normalised impact sensor signals recorded by sensors 1 and 3 for DT and HA impacts. The results are consistent with the dynamics of the impacts. Large-mass DT impacts are dominated by low-frequency components, with PSD v…
Figure 6
Figure 6. Figure 6: TDOA values extracted using the NSET method for four cases of drop mass impact testing: (a)-(b) Maximum TDOA differences between REF impacts and the HEI, ANG, and TEM cases, with 4 and 5 sensors, respectively; (c)-(d) Cosine similarity of TDOA between REF impacts and t…
Figure 7
Figure 7. Figure 7: illustrates the localisation results for TEM impacts based on REF impacts, leveraging these kernels and BMA fusion with six passive sensors. In this figure, red hollow circles represent sensor locations along the X and Y axes, black solid squares denote actual impact l…
Figure 8
Figure 8. Figure 8: Empirical CDF of localisation error for TEM impacts based on REF impacts, under variations in sensors: (a) 6 sensors (S1-S6), (b) 5 sensors (S1-S5), (c) 4 sensors (S1-S4). the empirical CDFs of localisation errors for TEM impacts using five and four sensors, respective…
Figure 9
Figure 9. Figure 9: Empirical CDF of localisation error for TEM impacts based on REF impacts, comparing different data preprocessing approaches with: (a) composite (COMP) kernel, (b) Bayesian Model Averaging (BMA) fusion. The results indicate that for both the COMP kernel and BMA fusion, …
Figure 10
Figure 10. Figure 10: Localisation of ANG and HEI impacts based on REF impacts: (a) localisation illustration of ANG impacts using BMA fusion with SDI, (b) localisation illustration of HEI impacts using BMA fusion with SDI, (c) empirical CDF of localisation errors for ANG impacts, comparin…
Figure 11
Figure 11. Figure 11: presents the localisation of DT impacts based on REF impacts using various kernel designs and BMA fusion, comparing the performance of sample standardisation (SSI) and feature standardi￾sation (FSI). Due to the substantially larger TDOA magnitudes of the DT impacts co…
Figure 12
Figure 12. Figure 12: Interpolated localisation of TEM impacts using different sizes of reference impact (RI) sets with kernel design and BMA fusion: (a) illustration of the three RI sets, (b) impact localisation with 15 RI using BMA fusion, (c) impact localisation with 9 RI using BMA fusi…
Figure 13
Figure 13. Figure 13: Interpolated localisation of TEM impacts with 9 RI using three different kernels: (a) COS kernel, (b) RBF kernel, and (c) COMP kernel. The COS kernel performs well in localising impacts in regions with larger TDOA values, while the RBF kernel excels in regions with sm…
Figure 14
Figure 14. Figure 14: Extrapolated localisation of TEM impacts with the 9 innermost reference impacts (RI), referred to as ’Ext-9RI’: (a) illustration of the RI set for extrapolation, denoted as ’Ext-9RI’, (b) localisation results using the COS kernel, (c) localisation results using the RB…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

76 extracted references · 58 canonical work pages

  1. [1]

    Damage tolerant composite fuselage sizing, characterization of accidental damage threat

    Faivre Vincent and Morteau Emilie. Damage tolerant composite fuselage sizing, characterization of accidental damage threat. Airbus Tech Mag (F AST), 48:10–16, 2011

    work page 2011

  2. [2]

    G. a. O. Davies and R. Olsson. Impact on composite structures. The Aeronautical Journal , 108(1089): 541–563, November 2004. ISSN 0001-9240, 2059-6464. doi: 10.1017/S0001924000000385

    work page doi:10.1017/s0001924000000385 2004

  3. [3]

    Dvorak and Norman Laws

    George J. Dvorak and Norman Laws. Analysis of Progressive Matrix Cracking In Composite Laminates II. First Ply Failure. Journal of Composite Materials , 21(4):309–329, April 1987. ISSN 0021-9983. doi: 10.1177/002199838702100402. Publisher: SAGE Publications Ltd STM

    work page doi:10.1177/002199838702100402 1987

  4. [4]

    Donadon, and Brian G

    Robin Olsson, Mauricio V. Donadon, and Brian G. Falzon. Delamination threshold load for dynamic impact on plates. International Journal of Solids and Structures , 43(10):3124–3141, May 2006. ISSN 0020-7683. doi: 10.1016/j.ijsolstr.2005.05.005

    work page doi:10.1016/j.ijsolstr.2005.05.005 2006

  5. [5]

    S. Z. H. Shah, S. Karuppanan, P. S. M. Megat-Yusoff, and Z. Sajid. Impact resistance and damage tolerance of fiber reinforced composites: A review. Composite Structures, 217:100–121, June 2019. ISSN 0263-8223. doi: 10.1016/j.compstruct.2019.03.021

    work page doi:10.1016/j.compstruct.2019.03.021 2019

  6. [6]

    Impact identification of stiffened composite panels: I

    Robert Seydel and Fu-Kuo Chang. Impact identification of stiffened composite panels: I. Sys- tem development. Smart Materials and Structures , 10(2):354–369, 2001. ISSN 0964-1726. doi: 10.1088/0964-1726/10/2/323

    work page doi:10.1088/0964-1726/10/2/323 2001

  7. [7]

    Impact localization and force reconstruction for composite plates based on virtual time reversal processing with time-domain spectral finite element method

    Zexing Yu, Chao Xu, Jiaying Sun, and Fei Du. Impact localization and force reconstruction for composite plates based on virtual time reversal processing with time-domain spectral finite element method. Structural Health Monitoring , 22(6):4149–4170, November 2023. ISSN 1475-9217. doi: 10.1177/14759217231164297

    work page doi:10.1177/14759217231164297 2023

  8. [8]

    Quantification, localization, and reconstruction of impact force on interval composite structures

    Yaru Liu and Lei Wang. Quantification, localization, and reconstruction of impact force on interval composite structures. International Journal of Mechanical Sciences , 239:107873, February 2023. ISSN 0020-7403. doi: 10.1016/j.ijmecsci.2022.107873

    work page arXiv 2023

Show all 76 references
  1. [9]

    Impact Identification Based on Surrogate- assisted Efficient Global Optimisation

    Dong Xiao, Zahra Sharif Khodaei, and M H Ferri Aliabadi. Impact Identification Based on Surrogate- assisted Efficient Global Optimisation. Procedia Structural Integrity, 52:667–678, January 2024. ISSN 2452-3216. doi: 10.1016/j.prostr.2023.12.067

    2024 doi

  2. [10]

    Accelerated strategies for impact identi- fication of composite laminate using sparse sensor networks and model-based virtual time rever- sal

    Zexing Yu, Jiaying Sun, Maoling Yue, and Chao Xu. Accelerated strategies for impact identi- fication of composite laminate using sparse sensor networks and model-based virtual time rever- sal. Structural Health Monitoring , page 14759217241290260, November 2024. ISSN 1475-9217...

    2024 doi

  3. [11]

    Impulse identification technique by estimating specific modal ponderations from vibration measurements

    Dimitri Goutaudier, Didier Gendre, V´ eronique Kehr-Candille, and Roger Ohayon. Impulse identification technique by estimating specific modal ponderations from vibration measurements. Journal of Sound and Vibration, 474:115263, May 2020. ISSN 0022-460X. doi: 10.1016/j.jsv.2020...

    2020

  4. [12]

    Single-sensor approach for impact localization and force reconstruction by using discriminating vibration modes

    Dimitri Goutaudier, Didier Gendre, V´ eronique Kehr-Candille, and Roger Ohayon. Single-sensor approach for impact localization and force reconstruction by using discriminating vibration modes. Mechanical Systems and Signal Processing , 138:106534, April 2020. ISSN 0888-3270. d...

    2020 doi

  5. [13]

    An efficient impact force identification methodology via a single sensor utilizing the concept of generalized transmissibility

    Li Zhang, Mingyao Liu, Liu Hong, Zechao Wang, Zude Zhou, and Wenlin Liao. An efficient impact force identification methodology via a single sensor utilizing the concept of generalized transmissibility. Mechanical Systems and Signal Processing , 211:111222, April 2024. ISSN 088...

    2024

  6. [14]

    Assessing impact force localization by using a particle swarm optimization algorithm

    Abdelali El-Bakari, Abdellatif Khamlichi, Eric Jacquelin, and Rachid Dkiouak. Assessing impact force localization by using a particle swarm optimization algorithm. Journal of Sound and Vibration , 333(6): 1554–1561, March 2014. ISSN 0022-460X. doi: 10.1016/j.jsv.2013.11.032

    2014 doi

  7. [15]

    Impact damage monitoring of FRP pressure vessels based on impact force identification

    Satoshi Atobe, Sunao Sugimoto, Ning Hu, and Hisao Fukunaga. Impact damage monitoring of FRP pressure vessels based on impact force identification. Advanced Composite Materials , 23(5-6):491–505, September 2014. ISSN 0924-3046. doi: 10.1080/09243046.2014.915112. Publisher: Tayl...

    2014

  8. [16]

    Impact detection in anisotropic materials using a time reversal approach

    F Ciampa and M Meo. Impact detection in anisotropic materials using a time reversal approach. Struc- tural Health Monitoring, 11(1):43–49, 2012. ISSN 1475-9217. doi: 10.1177/1475921710395815. Publisher: SAGE Publications

    2012 doi

  9. [17]

    A hierarchical method for the impact force reconstruction in composite structures

    Mario Emanuele De Simone, Francesco Ciampa, and Michele Meo. A hierarchical method for the impact force reconstruction in composite structures. Smart Materials and Structures , 28(8):085022, 2019. ISSN 0964-1726. doi: 10.1088/1361-665X/aae11c

    2019 doi

  10. [18]

    Impact source identification in finite isotropic plates using a time-reversal method: experimental study

    Chunlin Chen, Yulong Li, and Fuh-Gwo Yuan. Impact source identification in finite isotropic plates using a time-reversal method: experimental study. Smart Materials and Structures , 21(10):105025, August 2012. ISSN 0964-1726. doi: 10.1088/0964-1726/21/10/105025

    2012 doi

  11. [19]

    A novel strategy for response and force reconstruction under impact excitation

    Jie Liu and Bing Li. A novel strategy for response and force reconstruction under impact excitation. Journal of Mechanical Science and Technology , 32(8):3581–3596, August 2018. ISSN 1976-3824. doi: 10.1007/s12206-018-0709-4

    2018 doi

  12. [20]

    Material parameter optimization of flax/epoxy composite laminates under low-velocity impact

    Valentina Giammaria, Giulia Del Bianco, Elena Raponi, Dario Fiumarella, Raffaele Ciardiello, Simon- etta Boria, Fabian Duddeck, and Giovanni Belingardi. Material parameter optimization of flax/epoxy composite laminates under low-velocity impact. Composite Structures, 321:11730...

    2023

  13. [21]

    A FEM-based direct method for identification of Young’s modulus and boundary conditions in three-dimensional linear elasticity from local observation

    Tian Xu, Zhen Wang, Yingda Hu, Shilun Du, Ao Du, Zhenyang Yu, and Yong Lei. A FEM-based direct method for identification of Young’s modulus and boundary conditions in three-dimensional linear elasticity from local observation. International Journal of Mechanical Sciences , 237...

  14. [22]

    Review of finite element model updating methods for structural applications

    Suzana Ereiz, Ivan Duvnjak, and Javier Fernando Jim´ enez-Alonso. Review of finite element model updating methods for structural applications. Structures, 41:684–723, July 2022. ISSN 2352-0124. doi: 10.1016/j.istruc.2022.05.041

    2022 doi

  15. [23]

    John Wiley & Sons, June 2021

    Barna Szab´ o and Ivo Babuˇ ska.Finite Element Analysis: Method, Verification and Validation . John Wiley & Sons, June 2021. ISBN 978-1-119-42642-4

    2021

  16. [24]

    Sharif-Khodaei, M

    Z. Sharif-Khodaei, M. Ghajari, and M. H. Aliabadi. Determination of impact location on composite stiffened panels. Smart Materials and Structures , 21(10):105026, 2012. ISSN 0964-1726. doi: 10.1088/ 0964-1726/21/10/105026. 27

    2012

  17. [25]

    Ghajari, Z

    M. Ghajari, Z. Sharif-Khodaei, M. H. Aliabadi, and A. Apicella. Identification of impact force for smart composite stiffened panels. Smart Materials and Structures , 22(8):085014, 2013. ISSN 0964-1726. doi: 10.1088/0964-1726/22/8/085014

    2013 doi

  18. [26]

    Uncertainty quantification for impact location and force estimation in composite structures

    Aldyandra Hami Seno and MH Ferri Aliabadi. Uncertainty quantification for impact location and force estimation in composite structures. Structural Health Monitoring , page 147592172110202, June 2021. ISSN 1475-9217, 1741-3168. doi: 10.1177/14759217211020255

    2021 doi

  19. [27]

    M. R. Jones, T. J. Rogers, K. Worden, and E. J. Cross. A Bayesian methodology for localising acoustic emission sources in complex structures. Mechanical Systems and Signal Processing, 163:108143, January

  20. [28]

    Detection and time-of-arrival estimation of impact-induced waves in composite laminates

    Lukas Grasboeck, Alexander Humer, and Ayech Benjeddou. Detection and time-of-arrival estimation of impact-induced waves in composite laminates. Structural Health Monitoring , page 14759217241260848, July 2024. ISSN 1475-9217. doi: 10.1177/14759217241260848. Publisher: SAGE Pub...

    2024 doi

  21. [29]

    Hybrid DoA- TDoA method for impact localization on thin-walled structures using sensor clusters

    Xu Zeng, Deshuang Deng, Hongjuan Yang, Zhengyan Yang, Lei Yang, and Zhanjun Wu. Hybrid DoA- TDoA method for impact localization on thin-walled structures using sensor clusters. IEEE Sensors Journal, pages 1–1, 2025. ISSN 1558-1748. doi: 10.1109/JSEN.2024.3515461. Conference Na...

    2025

  22. [30]

    Qing, and Bin Liu

    Kaige Zhu, Xinlin P. Qing, and Bin Liu. A two-step impact localization method for composite structures with a parameterized laminate model. Composite Structures, 192:500–506, May 2018. ISSN 0263-8223. doi: 10.1016/j.compstruct.2018.03.052

    2018 doi

  23. [31]

    Qing, Bin Liu, and Tiren He

    Kaige Zhu, Xinlin P. Qing, Bin Liu, and Tiren He. A passive localization method for stiffened composite structures with a parameterized laminate model. Journal of Sound and Vibration , 489:115683, December

  24. [32]

    M. Meo, G. Zumpano, M. Piggott, and G. Marengo. Impact identification on a sandwich plate from wave propagation responses. Composite Structures, 71(3):302–306, December 2005. ISSN 0263-8223. doi: 10.1016/j.compstruct.2005.09.028

    2005 doi

  25. [33]

    Dong Xiao, Francisco de S´ a Rodrigues, Zahra Sharif-Khodaei, and M. H. Aliabadi. A general probabilistic framework for impact localisation based on flexural wave propagation. Mechanical Systems and Signal Processing, 226:112320, March 2025. ISSN 0888-3270. doi: 10.1016/j.ymss...

    2025

  26. [34]

    Dehghan Niri, A

    E. Dehghan Niri, A. Farhidzadeh, and S. Salamone. Nonlinear Kalman Filtering for acoustic emission source localization in anisotropic panels. Ultrasonics, 54(2):486–501, February 2014. ISSN 0041-624X. doi: 10.1016/j.ultras.2013.07.016

    2014 doi

  27. [35]

    Dehghan Niri and S

    E. Dehghan Niri and S. Salamone. A probabilistic framework for acoustic emission source localization in plate-like structures. Smart Materials and Structures , 21(3):035009, 2012. ISSN 0964-1726. doi: 10.1088/0964-1726/21/3/035009

    2012 doi

  28. [36]

    Multi-frequency probabilistic imaging fusion for impact localization on aircraft composite structures

    Deshuang Deng, Xu Zeng, Zhengyan Yang, Yu Yang, Sheng Zhang, Shuyi Ma, Hao Xu, Lei Yang, and Zhanjun Wu. Multi-frequency probabilistic imaging fusion for impact localization on aircraft composite structures. Structural Health Monitoring , page 14759217241233181, March 2024. IS...

    2024 doi

  29. [37]

    A Convolutional Neural Network for Impact Detection and Characterization of Complex Composite Structures

    Iuliana Tabian, Hailing Fu, and Zahra Sharif Khodaei. A Convolutional Neural Network for Impact Detection and Characterization of Complex Composite Structures. Sensors, 19(22):4933, January 2019. ISSN 1424-8220. doi: 10.3390/s19224933. 28

    2019 doi

  30. [38]

    Impact monitoring of large size complex metal structures based on sparse sensor array and transfer learning

    Bowen Zhao, Yiliang Zhang, Qijian Liu, and Xinlin Qing. Impact monitoring of large size complex metal structures based on sparse sensor array and transfer learning. Ultrasonics, 140:107305, May 2024. ISSN 0041-624X. doi: 10.1016/j.ultras.2024.107305

    2024

  31. [39]

    Impact force localization and reconstruction via gated temporal convolutional network

    Rui Zhou, Baijie Qiao, Junjiang Liu, Wei Cheng, and Xuefeng Chen. Impact force localization and reconstruction via gated temporal convolutional network. Aerospace Science and Technology, 144:108819, January 2024. ISSN 1270-9638. doi: 10.1016/j.ast.2023.108819

    2024

  32. [40]

    A hybrid FCN-BiGRU with trans- fer learning for low-velocity impact identification on aircraft structure

    Chenhui Huang, Weilin Liao, Hu Sun, Yishou Wang, and Xinlin Qing. A hybrid FCN-BiGRU with trans- fer learning for low-velocity impact identification on aircraft structure. Smart Materials and Structures , 32(5):055012, April 2023. ISSN 0964-1726. doi: 10.1088/1361-665X/acc623....

    2023 doi

  33. [41]

    Data-physics hybrid-driven deep learning method for impact force identification

    Jiaming Zhou, Yinshan Cai, Longlei Dong, Bo Zhang, and Zhike Peng. Data-physics hybrid-driven deep learning method for impact force identification. Mechanical Systems and Signal Processing , 211:111238, April 2024. ISSN 0888-3270. doi: 10.1016/j.ymssp.2024.111238

    2024

  34. [42]

    Spatial-temporal graph convolutional networks (STGCN) based method for localizing acoustic emission sources in composite panels

    Zhimin Zhao and Nian-Zhong Chen. Spatial-temporal graph convolutional networks (STGCN) based method for localizing acoustic emission sources in composite panels. Composite Structures, 323:117496, November 2023. ISSN 0263-8223. doi: 10.1016/j.compstruct.2023.117496

    2023

  35. [43]

    Impact force reconstruction and localization using Distance-assisted Graph Neural Network

    Chun Huang, Chongcong Tao, Hongli Ji, and Jinhao Qiu. Impact force reconstruction and localization using Distance-assisted Graph Neural Network. Mechanical Systems and Signal Processing , 200:110606, October 2023. ISSN 0888-3270. doi: 10.1016/j.ymssp.2023.110606

    2023

  36. [44]

    Aldyandra Hami Seno, Zahra Sharif Khodaei, and M. H. Ferri Aliabadi. Passive sensing method for im- pact localisation in composite plates under simulated environmental and operational conditions. Mechan- ical Systems and Signal Processing, 129:20–36, 2019. ISSN 0888-3270. doi:...

    2019 doi

  37. [45]

    Aldyandra Hami Seno and M. H. Ferri Aliabadi. Impact Localisation in Composite Plates of Different Stiffness Impactors under Simulated Environmental and Operational Conditions. Sensors, 19(17):3659, January 2019. ISSN 1424-8220. doi: 10.3390/s19173659

    2019 doi

  38. [46]

    Aldyandra Hami Seno and M. H. Ferri Aliabadi. Multifidelity data augmentation for data driven passive impact location and force estimation in composite structures under simulated environmental and oper- ational conditions. Mechanical Systems and Signal Processing , 195:110288,...

    2023

  39. [47]

    Nur M. M. Kalimullah, Amit Shelke, and Anowarul Habib. A probabilistic framework for source localiza- tion in anisotropic composite using transfer learning based multi-fidelity physics informed neural network (mfPINN). Mechanical Systems and Signal Processing , 197:110360, Aug...

    2023

  40. [48]

    Dong Xiao, Zahra Sharif-Khodaei, and M. H. Aliabadi. Hybrid physics-based and data-driven impact localisation for composite laminates. International Journal of Mechanical Sciences , 274:109222, July

  41. [49]

    Structural Health Monitoring for Advanced Composite Struc- tures, volume 08 of Computational and Experimental Methods in Structures

    M H Ferri Aliabadi and Z Sharif Khodaei. Structural Health Monitoring for Advanced Composite Struc- tures, volume 08 of Computational and Experimental Methods in Structures . World Scientific (Europe), February 2018. ISBN 978-1-78634-392-5 978-1-78634-393-2. doi: 10.1142/q0114

    2018 doi

  42. [50]

    F. C. Moon. A critical survey of wave propagation and impact in composite materials. Technical Report CR-121226, NASA, May 1973

    1973

  43. [51]

    Impact Induced Stress Waves in an Anisotropic Plate

    Byoung Sung Kim and Francis Moon. Impact Induced Stress Waves in an Anisotropic Plate. AIAA Journal, 17(10):1126–1133, October 1979. ISSN 0001-1452, 1533-385X. doi: 10.2514/3.61287. 29

    1979 doi

  44. [52]

    Hailing Fu, Zahra Sharif Khodaei, and M. H. Ferri Aliabadi. An Event-Triggered Energy-Efficient Wireless Structural Health Monitoring System for Impact Detection in Composite Airframes. IEEE Internet of Things Journal , 6(1):1183–1192, 2019. ISSN 2327-4662. doi: 10.1109/JIOT.2...

    2019

  45. [53]

    Impact on Composite Structures

    Serge Abrate. Impact on Composite Structures . Cambridge University Press, Cambridge, 1998. ISBN 978-0-521-47389-7. doi: 10.1017/CBO9780511574504

    1998 doi

  46. [54]

    F. C. Moon. Theoretical analysis of impact in composite plates. Technical Report AMS-1099, NASA, January 1973

    1973

  47. [55]

    I. M. Daniel, T. Liber, and R. H. LaBedz. Wave propagation in transversely impacted composite lami- nates. Experimental Mechanics, 19(1):9, 1979. ISSN 0014-4851. doi: 10.1007/BF02327764

    1979 doi

  48. [56]

    T. M. Tan and C. T. Sun. Wave propagation in graphite/epoxy laminates due to impact. Technical Report NASA-CR-168057, NASA, December 1982

    1982

  49. [57]

    Acoustic source localization

    Tribikram Kundu. Acoustic source localization. Ultrasonics, 54(1):25–38, January 2014. ISSN 0041-624X. doi: 10.1016/j.ultras.2013.06.009

    2014 doi

  50. [58]

    Dong Xiao, Zahra Sharif-Khodaei, and M. H. Aliabadi. Impact force identification for composite struc- tures using adaptive wavelet-regularised deconvolution. Mechanical Systems and Signal Processing , 220: 111608, November 2024. ISSN 0888-3270. doi: 10.1016/j.ymssp.2024.111608

    2024

  51. [59]

    R. Olsson. Mass criterion for wave controlled impact response of composite plates. Composites Part A: Applied Science and Manufacturing , 31(8):879–887, August 2000. ISSN 1359-835X. doi: 10.1016/ S1359-835X(00)00020-8

  52. [60]

    A tutorial on Gaussian process regression: Modelling, exploring, and exploiting functions

    Eric Schulz, Maarten Speekenbrink, and Andreas Krause. A tutorial on Gaussian process regression: Modelling, exploring, and exploiting functions. Journal of Mathematical Psychology , 85:1–16, August

  53. [61]

    Alexander I. J. Forrester, Andrs Sbester, and Andy J. Keane. Engineering Design via Surrogate Mod- elling. John Wiley & Sons, Ltd, Chichester, UK, July 2008. ISBN 978-0-470-77080-1 978-0-470-06068-1. doi: 10.1002/9780470770801

    2008 doi

  54. [62]

    Gaussian Process Surrogate Model with Composite Kernel Learning for Engineering Design

    Pramudita Satria Palar, Lavi Rizki Zuhal, and Koji Shimoyama. Gaussian Process Surrogate Model with Composite Kernel Learning for Engineering Design. AIAA Journal , 58(4):1864–1880, 2020. ISSN 0001-1452. doi: 10.2514/1.J058807. Publisher: American Institute of Aeronautics and ...

    2020 doi

  55. [63]

    JGPR: a computationally efficient multi- target Gaussian process regression algorithm

    Mohammad Nabati, Seyed Ali Ghorashi, and Reza Shahbazian. JGPR: a computationally efficient multi- target Gaussian process regression algorithm. Machine Learning, 111(6):1987–2010, June 2022. ISSN 1573-0565. doi: 10.1007/s10994-022-06170-3

    1987 doi

  56. [64]

    GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration

    Jacob Gardner, Geoff Pleiss, Kilian Q Weinberger, David Bindel, and Andrew G Wilson. GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration. In Advances in Neural Information Processing Systems, volume 31, pages 1–11. Curran Associates, Inc., 2018

    2018

  57. [65]

    PyTorch: An Imperative Style, High-Performance Deep Learning Library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu ...

    2019

  58. [66]

    Bayesian Model Selection and Model Averaging

    Larry Wasserman. Bayesian Model Selection and Model Averaging. Journal of Mathematical Psychology, 44(1):92–107, March 2000. ISSN 0022-2496. doi: 10.1006/jmps.1999.1278

  59. [67]

    Multiple Delaminations and their Severity in Circular Axisym- metric Plates Subjected to Transverse Loading

    Hiroshi Suemasu and Osamu Majima. Multiple Delaminations and their Severity in Circular Axisym- metric Plates Subjected to Transverse Loading. Journal of Composite Materials , 30(4):441–453, March

  60. [68]

    Analytical prediction of large mass impact damage in composite laminates

    Robin Olsson. Analytical prediction of large mass impact damage in composite laminates. Composites Part A: Applied Science and Manufacturing , 32(9):1207–1215, September 2001. ISSN 1359-835X. doi: 10.1016/S1359-835X(01)00073-2

    2001 doi

  61. [69]

    Thiene, M

    M. Thiene, M. Ghajari, U. Galvanetto, and M. H. Aliabadi. Effects of the transfer function evaluation on the impact force reconstruction with application to composite panels. Composite Structures, 114:1–9, August 2014. ISSN 0263-8223. doi: 10.1016/j.compstruct.2014.03.055

    2014 doi

  62. [70]

    Dong Xiao, Zahra Sharif-Khodaei, and M. H. Aliabadi. Time of Arrival Extraction for Impact Lo- calisation on Composite Structures. e-Journal of Nondestructive Testing , 29(07), July 2024. doi: 10.58286/29601. 31

    2024 doi

  63. [1996]

    doi: 10.1177/002199839603000402

    ISSN 0021-9983. doi: 10.1177/002199839603000402

    doi

  64. [2018]

    doi: 10.1016/j.jmp.2018.03.001

    ISSN 0022-2496. doi: 10.1016/j.jmp.2018.03.001

    2018 doi

  65. [2020]

    doi: 10.1016/j.jsv.2020.115683

    ISSN 0022-460X. doi: 10.1016/j.jsv.2020.115683

    2020

  66. [2022]

    doi: 10.1016/j.ymssp.2021.108143

    ISSN 0888-3270. doi: 10.1016/j.ymssp.2021.108143

    2021

  67. [2023]

    doi: 10.1016/j.ijmecsci.2022.107797

    ISSN 0020-7403. doi: 10.1016/j.ijmecsci.2022.107797

    2022

  68. [2024]

    doi: 10.1016/j.ijmecsci.2024.109222

    ISSN 0020-7403. doi: 10.1016/j.ijmecsci.2024.109222

    2024

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.