Pith. sign in

REVIEW 4 major objections 7 minor 52 references

This paper claims that an unsupervised dynamic variational autoencoder paired with a particle filter can learn a temporally coherent latent state of the weld pool that reflects weld penetration—measured by backside bead width—and that the s

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 21:35 UTC pith:EJDARE4Q

load-bearing objection A plausible DVAE+PF combination for weld monitoring, but the particle filter update is not the Bayes rule it claims and the validation is too thin. the 4 major comments →

arxiv 2607.16013 v1 pith:EJDARE4Q submitted 2026-07-17 eess.IV

Robust Monitoring of Arc Welding Processes: A Generalizable Framework with DVAE and Particle Filter

Yue Cao , Hai Lin , YuMing Zhang This is my paper
classification eess.IV
keywords arc weldingweld penetration monitoringdynamic variational autoencoderparticle filterlatent state estimationGTAWGMAWunsupervised learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper is trying to show that a single, unsupervised monitoring pipeline can learn a compact latent state of the weld pool that tracks weld penetration and transfers across different arc welding processes. The proposed pipeline pairs a dynamic variational autoencoder (DVAE), which jointly learns image encoding and LSTM-driven latent dynamics under control inputs, with a particle filter that fuses model predictions with noisy observations for real-time state estimation. The authors argue that the filter's state estimate encodes physically meaningful penetration information, demonstrated by calibrating it to backside bead width in GTAW, and that the same unretuned framework produces stable latent states on GMAW data. If true, this would reduce the need for large labeled datasets and process-specific model tuning in weld monitoring.

Core claim

The central discovery claimed is that the latent process state x̂_t, inferred by the DVAE–particle-filter framework, is a faithful, temporally coherent representation of the welding process rather than a frame-wise image code. In GTAW, a quadratic regression from the five-dimensional x̂_t to measured backside bead width gives an RMSE of 0.27 mm on a representative experiment, which the authors take as evidence that the unsupervised latent state carries penetration information. The same DVAE architecture, training hyperparameters, and PF settings are then applied to GMAW without customisation; the reconstructions preserve dominant pool semantics and the PF yields smooth, observation-corrected

What carries the argument

The central object is the latent process state x_t, a 5-dimensional Gaussian code produced by a CNN encoder–decoder and evolved by an LSTM transition model conditioned on control inputs. The DVAE is trained with an ELBO whose KL term aligns the encoder posterior with the LSTM transition prior, while the second half of each training sequence is predicted autoregressively to force long-horizon dynamics. Inference is done by a particle filter that deterministically updates each particle's LSTM hidden and cell states, samples the latent state from the transition prior, weights particles by the encoder posterior, resamples by effective sample size, and reports the weighted mean x̂_t.

Load-bearing premise

The load-bearing assumption is that the encoder's posterior distribution over latent states given a weld-pool image can stand in for the observation likelihood in the particle filter update; if it is miscalibrated or overconfident, the filter's state estimates will be biased even if they look smooth.

What would settle it

Run the PF on a held-out GTAW sequence with the same particles but weight them by the decoder likelihood pθ(z_t|x_t) instead of the encoder posterior qφ(x_t|z_t), and compare the resulting x̂_t and the bead-width regression RMSE; a significant difference would show the filter depends on the unjustified likelihood substitution.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • A small labeled sample suffices: because x̂_t is learned unsupervisedly, only a small set of bead-width measurements is needed to calibrate the latent state to penetration.
  • The framework yields a real-time process-state estimate with a principled uncertainty measure, which can be used to detect deviations from a desired state encoded from high-quality frames.
  • The learned LSTM transition model acts as a data-enabled surrogate of weld-pool dynamics, opening a route to model-based process control.
  • Because the DVAE encoder–decoder is modality-agnostic, the same DVAE–PF pipeline could be applied to electrical, acoustic, or multi-modal welding data.
  • Demonstrated transfer from GTAW to GMAW without process-specific tuning suggests the approach may scale to other arc welding variants.

Where Pith is reading between the lines

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

  • The PF update uses the encoder posterior qφ(x_t|z_t) as the observation likelihood; the paper never derives or justifies this substitution. If that posterior is overconfident or biased, the estimated state would be biased even when the trajectory appears smooth.
  • The GMAW experiment demonstrates stable latent-state estimation, but it does not report a ground-truth penetration measure for GMAW; claiming that the same latent state carries penetration information in GMAW would require a bead-width or weld-cross-section correlation there.
  • All experiments vary only welding speed, so the tested dynamics are a slice of the full operating envelope; extrapolation to current, voltage, or wire-feed variations is plausible but untested.
  • Because the training objective forces long-horizon autoregressive prediction, one could probe whether a shorter prediction horizon or observation-injected training changes the PF's advantage; the paper's design choice is motivated but not ablated.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes a generalizable framework for arc-welding monitoring that combines a Dynamic Variational Autoencoder (DVAE) with a particle filter (PF). The DVAE, consisting of a CNN encoder–decoder and an LSTM transition model, is trained unsupervisedly on weld-pool image sequences and control inputs. The PF then propagates the latent state and LSTM hidden/cell states to produce a temporally coherent state estimate x̂_t. The main claims are that x̂_t reflects weld penetration (validated against backside bead width in GTAW) and that the same architecture and hyperparameters generalize to GMAW without process-specific tuning. The manuscript includes qualitative reconstructions, state-estimation plots, and one quantitative GTAW bead-width regression result.

Significance. If the claims hold, the framework would be a practically valuable, label-efficient alternative to supervised welding-monitoring models, with a plausible path to real-time control. The paper has clear strengths: the DVAE jointly learns representation and dynamics rather than decoupling them; the same architecture and hyperparameters are applied to two different arc-welding processes; and the GTAW bead-width correlation, if robust across all held-out experiments, would demonstrate physical interpretability of the learned latent state. However, the current evidence is not fully commensurate with the strength of the claims: the quantitative validation rests on a single representative case, the GMAW validation is qualitative, no baselines are compared, and the PF update uses an unjustified likelihood substitution. These issues are load-bearing but appear fixable within the manuscript's scope.

major comments (4)
  1. [Appendix A, Table A.2 step 2; Eq. (3); Eq. (5)] The PF update weights particles by evaluating the DVAE encoder posterior q_φ(x_t|z_t) = N(μ_obs, Σ_obs) at each particle, i.e., w_t^(i) ∝ w_{t-1}^{(i)} q_φ(x_t^(i)|z_t). But the Bayesian recursion in Eq. (3) requires the observation likelihood p_θ_obs(z_t|x_t), and Eq. (5) identifies the decoder likelihood as the generative likelihood. Using the amortized posterior as a likelihood double-counts the transition prior, since particles were already sampled from that prior, and can bias the filter whenever q_φ is miscalibrated or overconfident. No derivation in Section 4 or Appendix A reconciles this discrepancy. Because x̂_t from this filter is the input to the bead-width regression in Eq. (6), the paper's main quantitative evidence inherits this issue.
  2. [Section 4, Eq. (6), Fig. 10] The central quantitative claim that the PF state accurately reflects backside bead width is supported by a single 'representative case' with RMSE = 0.27 mm. The text states that a quadratic regression was trained on three of eight GTAW experiments and validated on the remaining ones, but no aggregate RMSE, R², per-experiment table, or error bars are reported. A single favorable curve is insufficient to establish that x̂_t 'encodes physically meaningful process information' or that the calibration generalizes across the dataset. Please report all validation cases with summary statistics and, where possible, confidence intervals for the calibration.
  3. [Section 5, Figs. 13–14] The GMAW validation is entirely qualitative. Figure 14 shows that the PF estimate roughly tracks the encoder posterior and the transition prior, and Fig. 13 shows image reconstructions, but there is no ground-truth penetration or bead-width comparison, no error metric, and no held-out evaluation. Given that the abstract and conclusions claim that 'validation on GTAW and GMAW... demonstrates the framework's generalizability and robustness,' quantitative validation on GMAW is needed, or the claim should be explicitly weakened to a qualitative transfer demonstration.
  4. [Sections 4–5] No baselines are compared. The robustness advantages (noise suppression, drift mitigation, superiority over encoder-only and transition-prior-only estimates) are asserted from Fig. 9 but never quantified. Comparisons against the encoder-only estimate, the transition-prior-only prediction, and a supervised CNN-LSTM method such as the one in Ref. [4] would substantiate the claimed advantages and would also provide context for interpreting the reported RMSE of 0.27 mm. Without such comparisons, the 'robustness' component of the central claim is not established beyond a qualitative demonstration.
minor comments (7)
  1. [Eq. (3)] The posterior notation is inconsistent: the left-hand side conditions on z_{t-1:t-n}, but the right-hand side includes the current observation z_t. This should be p(x_t | z_{1:t}, u) or a similarly explicit filtering distribution.
  2. [Eqs. (2), (3), (5)] The observation likelihood is denoted p_θ_obs in Eqs. (2)–(3) but p_θ_dec in Eq. (5). Please unify the notation, especially because the correct likelihood choice is central to the PF update.
  3. [Section 5, first paragraph] The text says the GMAW experimental setup is 'identical as in Section 4.2,' but the GTAW setup is described in Section 3.2. The cross-reference is wrong.
  4. [Section 4, penultimate paragraph] The statement 'This contrasts with prior work [2], where deep networks were trained in a fully supervised manner' appears to cite the wrong reference. Reference [2] is a general review of advanced welding manufacturing; a more appropriate citation would be Ref. [3], [4], or the supervised-learning works in Section 1.2.1.
  5. [Section 2, first paragraph] The nucleation-and-growth statement cites Ref. [43], which is a review of Bayesian filters with machine learning. This seems to be a citation error; a metallurgy reference would be appropriate.
  6. [Fig. 6 and Appendix A] Minor presentation issues: the y-axis and the meaning of the 'ratio' in Fig. 6 are not labeled; Table A.2 has the header 'Symbol Meaning' but contains an algorithm; and the title/abstract inconsistently types 'DV AE' versus 'DVAE'.
  7. [Data and code availability] No data or code availability statement is provided. Given the unsupervised-training and filtering claims, making the trained model and a representative dataset available would materially improve reproducibility.

Circularity Check

0 steps flagged

No significant circularity: the bead-width calibration is external and the DVAE-PF pipeline does not reduce to its own inputs.

full rationale

The paper's central validation is not circular. The DVAE is trained unsupervisedly on weld-pool images and speed inputs via the ELBO (Eq. 5), while the backside bead width W_b is measured post-weld and enters only through the quadratic calibration in Eq. (6), fit on three GTAW experiments and evaluated on the remaining ones. Because the bead-width labels are not used in learning the latent representation or the particle filter, the reported RMSE is an external, label-based check rather than a quantity re-derived from training targets. The self-citations ([34] related latent-dynamics work; [50] teleoperation hardware) are contextual and not load-bearing. Two non-circular concerns should be flagged rather than scored as circularity: (i) Appendix A, Table A.2 step 2 updates particle weights by evaluating the encoder posterior q_phi(x_t|z_t)=N(mu_obs, Sigma_obs) at each particle, whereas Eq. (3) defines the observation likelihood as p_theta_obs(z_t|x_t); this substitution is an unsupported approximation and a correctness risk, but the PF output remains a nontrivial combination of the transition prior and the encoder density, and the bead-width regression still provides an external test. (ii) The DVAE training sequences appear to be drawn from all eight GTAW experiments, including some later used for bead-width validation, which weakens the independence of the representation-level validation; this is a data-split limitation, not a circular derivation. The agreement in Fig. 8 between the transition prior and encoder posterior is partly a training-objective check, but the PF/bead-width result does not reduce to that agreement.

Axiom & Free-Parameter Ledger

7 free parameters · 7 axioms · 0 invented entities

The central claim rests on standard variational inference and Bayesian filtering, plus domain assumptions about weld pool geometry and dynamics. The most fragile item is the ad hoc substitution of the encoder posterior for the observation likelihood in the particle filter, which is not justified. Free parameters are numerous and mostly hand-selected with no sensitivity analysis.

free parameters (7)
  • Latent state dimension = 5
    Hand-chosen in Section 3.1; the central latent state size is set to 5 without ablation.
  • LSTM hidden/cell dimension = 128
    Section 3.1: 'Both internal states are 128-dimensional vectors.' Hand-chosen.
  • Particle count N
    PF algorithm in Appendix A does not specify the number of particles; a critical free parameter for filter performance.
  • Resampling threshold τ = 0.5
    Appendix A: 'Resampling threshold (e.g., 0.5)' — chosen without sensitivity analysis.
  • Learning rates = 1e-4 (encoder-decoder), 1e-3 (LSTM)
    Section 3.3: two-optimizer scheme with Adam; values hand-chosen.
  • Training sequence length and split = L=100, first 50 init, last 50 autoregressive
    Section 3.2/3.3: sliding window step 5; training strategy is ad hoc.
  • Bead width regression coefficients β = not reported
    Section 4, Eq. 6: quadratic regression fitted on three GTAW experiments; coefficients are fitted to data.
axioms (7)
  • standard math Variational inference and reparameterization trick are valid for training the DVAE.
    Used implicitly in Section 3.1 and Eq. 5 (ELBO).
  • standard math Bayesian filtering recursion (Eq. 3) correctly represents the posterior of the latent state.
    Based on the state-space model in Eqs. 1-2.
  • domain assumption Weld pool geometry can be represented by a low-dimensional latent space (dim 5) that preserves penetration-relevant information.
    Section 2: justified via dimensionless analysis and limited variation of welding parameters; no direct evidence that dim 5 is sufficient.
  • domain assumption Topside weld pool appearance is correlated with backside bead width (penetration).
    Section 2, cited from prior work [4-6, 20-24]; used to validate the latent state against bead width.
  • domain assumption The welding process is slow and inertial enough that LSTM dynamics and particle filtering provide an accurate state representation.
    Section 1.4 and 4; stated as a design rationale, not empirically verified via dynamics timescales.
  • ad hoc to paper The encoder posterior q(x_t|z_t) can be used as the observation likelihood p(z_t|x_t) in the particle filter update.
    Appendix A, step 2: the weight update evaluates particles under N(μ_obs, Σ_obs), the encoder posterior, not under the decoder likelihood. This swap is not derived or discussed.
  • ad hoc to paper The training strategy of initializing the LSTM on the first half of a sequence and autoregressively predicting the second half yields a transition model that generalizes to new sequences.
    Section 3.1/3.2: this is a specialized training procedure; no theoretical guarantee, and its effect on performance is not ablated.

pith-pipeline@v1.3.0-alltime-deepseek · 15337 in / 15212 out tokens · 130264 ms · 2026-08-01T21:35:10.807558+00:00 · methodology

0 comments
read the original abstract

Arc welding processes are essential for continuous fabrication but prone to disturbances that impair weld quality, making real-time monitoring critical yet difficult due to complex visual patterns and nonlinear, time-varying dynamics. Deep learning shows promise but faces scalability limits because of its dependence on large labeled datasets and application-specific tuning. We explore whether a unified approach can characterize major arc welding processes across applications and improve scalability through consistent state monitoring. This paper introduces a robust and generalizable monitoring framework for arc welding. It combines unsupervised deep latent representation learning, which extracts compact features from weld pool images, with Bayesian filtering to handle persistent and fluctuating disturbances such as arc radiation and specular reflections. Specifically, a Dynamic Variational Autoencoder (DVAE), consisting of a CNN-based encoder-decoder and an LSTM-based transition model, jointly learns latent representations and their evolution under control inputs. For robust real-time inference, a specialized Particle Filter (PF) propagates the latent and LSTM hidden states, preserving process history while suppressing sensor noise. This design is well suited to welding's slow and inertial dynamics. Validation on GTAW and GMAW without process-specific tuning demonstrates the framework's generalizability and robustness.

Figures

Figures reproduced from arXiv: 2607.16013 by Hai Lin, Yue Cao, Yuming Zhang.

Figure 1
Figure 1. Figure 1: Fig.1. Structure of Dynamic Variational Autoencoder [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: ELBO Loss 3.2 GTAW Dataset The proposed probabilistic state estimation framework is first validated in Gas Tungsten Arc Welding (GTAW), a widely used arc welding technique valued for its stable arc, precise heat input control, and clean welds. The experimental system is illustrated in [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 5
Figure 5. Figure 5: Weld Pool Views. 3.3 DVAE Training Results The DVAE was trained for 100 epochs on 555 sequences, each consisting of 100 frames of weld pool images paired with welding speed inputs. In each iteration, four sequences were sampled as a batch (400 frames in total). A two-optimizer scheme was adopted: the CNN encoder–decoder was updated with Adam (learning rate 1×10⁻⁴) using 25 micro-batches of 16 images each, … view at source ↗
Figure 6
Figure 6. Figure 6: Training Curves of DVAE on GTAW Dataset. [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: presents the comparison between reconstructed weld pool images and the corresponding original images using the trained encoder-decoder. The reconstructions resemble the originals in overall size and boundary but appear more blurred, with subtle irregular details in the pool morphology are not preserved. This indicates that the learned latent representation successfully captures the essential geometry of th… view at source ↗
Figure 8
Figure 8. Figure 8: Validation of LSTM Transition Model on GTAW Experiment. [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

52 extracted references · 4 linked inside Pith

  1. [1]

    Norrish, Advanced welding processes, Elsevier, 2006

    J. Norrish, Advanced welding processes, Elsevier, 2006

    2006

  2. [2]

    Y . M. Zhang, Y .-P. Yang, W. Zhang, S.-J. Na, Advanced welding manufacturing: a brief analysis and review of challenges and solutions, Journal of manufacturing science and engineering 142 (11) (2020) 110816

    2020

  3. [3]

    R. Yu, Y . Cao, H. Chen, Q. Ye, Y . Zhang, Deep learning based real-time and in-situ monitoring of weld penetration: Where we are and what are needed revolutionary solutions?, Journal of Manufacturing Processes 93 (2023) 15–46

    2023

  4. [4]

    R. Yu, J. Kershaw, P. Wang, Y . Zhang, How to accurately monitor the weld penetration from dynamic weld pool serial images using cnn-lstm deep learning model?, IEEE Robotics and Automation Letters 7 (3) (2022) 6519–6525

    2022

  5. [5]

    Y . K. Liu, Y . M. Zhang, Model-based predictive control of weld penetration in gas tungsten arc welding, IEEE Transactions on Control Systems Technology 22 (3) (2014) 955 –966. doi:10.1109/TCST.2013.2266662

    arXiv 2014

  6. [6]

    X. Gao, Z. Liang, X. Zhang, L. Wang, X. Yang, Penetration state recognition based on stereo vision in gmaw process by deep learning, Journal of Manufacturing Processes 89 (2023) 349–361

    2023

  7. [7]

    Y . Xiao, G. Den Ouden, Weld pool oscillation during GTA welding of mild steel, Welding Journal -New York- 72 (1993) 428–s

    1993

  8. [8]

    C. Li, Y . Shi, Y . Gu, P. Yuan, Monitoring weld pool oscillation using reflected laser pattern in gas tungsten arc welding, Journal of materials processing technology 255 (2018) 876–885

    2018

  9. [9]

    Y . Cao, Z. Wang, S. Hu, W. Wang, Modeling of weld penetration control system in GMAW-P using narmax methods, Journal of Manufacturing Processes 65 (2021) 512–524

    2021

  10. [10]

    Zhang, B

    Y . Zhang, B. Chen, C. Tan, X. Song, H. Zhao, An end-to-end framework based on acoustic emission for welding penetration prediction, Journal of Manufacturing Processes 107 (2023) 411–421

    2023

  11. [11]

    Y . Wang, W. Lee, S. Jang, V . D. Truong, Y . Jeong, C. Won, J. Lee, J. Yoon, Prediction of internal welding penetration based on ir thermal image supported by machine vision and ann -model during automatic robot welding process, Journal of Advanced Joining Processes 9 (2024) 100199

    2024

  12. [12]

    Kotecki, D

    D. Kotecki, D. Cheever, D. Howden, et al., Mechanism of ripple formation during weld solidification, WELD J 51 (8) (1972) 368

    1972

  13. [13]

    Scheck, A

    M. Scheck, A. Richter, S. Beitler, T. Gehling, K. Treutler, V . Wesling, C. Rembe, C. Bohn, High geometric fidelity through closed -loop control of the weld pool size in gas metal arc welding based direct energy deposition, Additive Manufacturing 80 (2024) 103944

    2024

  14. [14]

    Richter, T

    A. Richter, T. Gehling, K. Treutler, V . Wesling, C. Rembe, Real-time measurement of temperature and volume of the weld pool in wire-arc additive manufacturing, Measurement: Sensors 17 (2021) 100060

    2021

  15. [15]

    Solis, A

    G. Solis, A. Myers, G. Q uirarte, B. M ondal, J. M alen, S. N arra, Two-color thermography of GMAW to enable real-time hardness prediction

  16. [16]

    Z. Wang, Y . Shi, Y . Cui, W. Yan, Three-dimensional weld pool monitoring and penetration state recognition for variable-gap keyhole tungsten inert gas welding based on stereo vision, Sensors (Basel, Switzerland) 24 (23) (2024) 7591

    2024

  17. [18]

    Y . Hong, J. Xu, S. Li, L. Zhu, B. Chang, D. Du, Vision-based penetration monitoring method for climbing helium arc welding using deep learning with multivariate time series forecasting, Journal of Manufacturing Processes 141 (2025) 1522–1534

    2025

  18. [19]

    F. Xu, Y . Xu, H. Zhang, S. Chen, Application of sensing technology in intelligent robotic arc welding: A review, Journal of Manufacturing Processes 79 (2022) 854–880

    2022

  19. [20]

    Zhang, B

    Z. Zhang, B. Li, W. Zhang, R. Lu, S. Wada, Y . Zhang, Real -time penetration state monitoring using convolutional neural network for laser welding of tailor rolled blanks, Journal of Manufacturing Systems 54 (2020) 348–360

    2020

  20. [21]

    S. Kang, K. Lee, M. Kang, Y . H. Jang, C. Kim, Weld -penetration-depth estimation using deep learning models and multisensor signals in al/cu laser overlap welding, Optics & Laser Technology 161 (2023) 109179

    2023

  21. [22]

    Z. Wang, H. Chen, Q. Zhong, S. Lin, J. Wu, M. Xu, Q. Zhang, Recognition of penetration state in GTAW based on vision transformer using weld pool image, The International Journal of Advanced Manufacturing Technology 119 (7) (2022) 5439–5452

    2022

  22. [23]

    Y.-H. Shi, Z. -S. Wang, X. -Y . Chen, Y .-X. Cui, T. Xu, J. -Y . Wang, Realtime k-tig welding penetration prediction on embedded system using a segmentation -LSMT model, Advances in Manufacturing 11 (3) (2023) 444–461

    2023

  23. [24]

    Y . Hong, X. He, J. Xu, R. Yuan, K. Lin, B. Chang, D. Du, Af-fttsnet: An end-to-end two-stream convolutional neural network for online quality monitoring of robotic welding, Journal of Manufacturing Systems 74 (2024) 422–434

    2024

  24. [25]

    Mucllari, R

    E. Mucllari, R. Yu, Y . Cao, Q. Ye, Y . Zhang, Do we need a new foundation to use deep learning to monitor weld penetration?, IEEE Robotics and Automation Letters 8 (6) (2023) 3669–3676

    2023

  25. [26]

    Kim, D.-H

    E.-S. Kim, D.-H. Lee, G.-J. Seo, D.-B. Kim, S.-J. Shin, Development of a cnn -based real-time monitoring algorithm for additively manufactured molybdenum, Sensors and Actuators A: Physical 352 (2023) 114205

    2023

  26. [27]

    M. M. Mahdi, M. A. G. Raju, K.-C. Lee, D. B. Kim, Federated learning enabled secure multi-modal anomaly detection for wire arc additive manufacturing, Machines 13 (11) (2025) 1063

    2025

  27. [28]

    Y . Cai, C. Li, H. Chen, J. Xiong, Towards automatic anomaly detection and diagnosis in positional arc - directed energy deposition based on deep learning, Journal of Manufacturing Systems 79 (2025) 1 –13

    2025

  28. [29]

    D. B. Kim, H. Chong, M. M. Mahdi, S. -J. Shin, Data-fused and concatenated-ensemble learning for in-situ anomaly detection in wire and arc -based direct energy deposition, Journal of Manufacturing Processes 112 (2024) 273–289

    2024

  29. [30]

    D. P. Kingma, M. Welling, Auto-encoding variational bayes, arXivpreprint arXiv:1312.6114 (2013)

    Pith/arXiv arXiv 2013

  30. [31]

    Van Den Oord, O

    A. Van Den Oord, O. Vinyals, et al., Neural discrete representation learning, Advances in neural information processing systems 30 (2017)

    2017

  31. [32]

    R. T. Chen, X. Li, R. B. Grosse, D. K. Duvenaud, Isolating sources of disentanglement in variational autoencoders, Advances in neural information processing systems 31 (2018)

    2018

  32. [33]

    Y . Hahn, R. Maack, H. Tercan, T. Meisen, M. Purrio, G. Buchholz, M. Angerhausen, Quality prediction in arc welding: Leveraging transformer models and discrete representations from vector quantised -vae, in: Proceedings of the 33rd ACM International Conference on Information and Knowledge Management, 2024, pp. 4522–4529

    2024

  33. [34]

    Synthesizing weld pool dynamics in GTAW with latent space modeling and adversarial refinement

    Yue Cao, Qiang Ye, YuMing Zhang. Synthesizing weld pool dynamics in GTAW with latent space modeling and adversarial refinement. Accepted for publication, Journal of Manufacturing Processes

  34. [35]

    R. Li, H. Ma, R. Wang, H. Song, X. Zhou, L. Wang, H. Zhang, K. Zeng, C. Xia, Application of unsupervised learning methods based on video data for real-time anomaly detection in wire arc additive manufacturing, Journal of Manufacturing Processes 143 (2025) 37–55

    2025

  35. [36]

    Girin, S

    L. Girin, S. Leglaive, X. Bie, J. Diard, T. Hueber, X. Alameda-Pineda, Dynamical variational autoencoders: A comprehensive review, arXiv preprint arXiv:2008.12595 (2020)

    Pith/arXiv arXiv 2008

  36. [37]

    Y . S. Lee, J. Chen, Developing semi-supervised latent dynamic variational autoencoders to enhance prediction performance of product quality, Chemical Engineering Science 265 (2023) 118192

    2023

  37. [38]

    N. Chen, M. Karl, P. Van Der Smagt, Dynamic movement primitives in latent space of time -dependent variational autoencoders, in: 2016 IEEE -RAS 16th international conference on humanoid robots (Humanoids), IEEE, 2016, pp. 629–636

    2016

  38. [39]

    X. Bie, W. Guo, S. Leglaive, L. Girin, F. Moreno -Noguer, X. Alameda-Pineda, Hit-dvae: Human motion generation via hierarchical transformer dynamical vae, arXiv preprint arXiv:2204.01565 (2022)

    Pith/arXiv arXiv 2022

  39. [40]

    X. Lin, X. Bie, S. Leglaive, L. Girin, X. Alameda-Pineda, Speech modeling with a hierarchical transformer dynamical vae, in: ICASSP 2023 -2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2023, pp. 1–5

    2023

  40. [41]

    Chen, et al., Bayesian filtering: From kalman filters to particle filters, and beyond, Statistics 182 (1) (2003) 1–69

    Z. Chen, et al., Bayesian filtering: From kalman filters to particle filters, and beyond, Statistics 182 (1) (2003) 1–69

    2003

  41. [42]

    Zheng, H

    T. Zheng, H. Lin, Pde -based dynamic control and estimation of soft robotic arms, in: 2022 IEEE 61st Conference on decision and control (CDC), IEEE, 2022, pp. 2702–2707

    2022

  42. [43]

    S. Kim, I. Petrunin, H. -S. Shin, A review of bayes filters with machine learning techniques and their applications, Information Fusion 114 (2025) 102707

    2025

  43. [44]

    D. Kim, M. Park, Y .-L. Park, Probabilistic modeling and bayesian filtering for improved state estimation for soft robots, IEEE Transactions on Robotics 37 (5) (2021) 1728–1741

    2021

  44. [45]

    Z. Wang, L. Zheng, S. Yan, R. J. van Sloun, N. Shlezinger, Y . C. Eldar, Eotnet: Deep memory aided bayesian filter for extended object tracking, arXiv preprint arXiv:2505.18684 (2025)

    Pith/arXiv arXiv 2025

  45. [46]

    Kovacevic, Z

    R. Kovacevic, Z. Cao, Y . Zhang, Role of welding parameters in deter- mining the geometrical appearance of weld pool (1996)

    1996

  46. [47]

    Arora, G

    A. Arora, G. Roy, T. DebRoy, Unusual wavy weld pool boundary from dimensional analysis, Scripta Materialia 60 (2) (2009) 68–71

    2009

  47. [48]

    Y . Deng, L. Shu, Y . Wang, D. Ma, Z. Luo, P. Jiang, A data-driven framework for molten pool morphology prediction in laser welding with dimensionless numbers discovery, Journal of Manufacturing Processes 151 (2025) 826–842

    2025

  48. [49]

    Y . Yu, X. Si, C. Hu, J. Zhang, A review of recurrent neural networks: Lstm cells and network architectures, Neural computation 31 (7) (2019) 1235–1270

    2019

  49. [50]

    Y . Cao, S. Guo, Y . Zhang, Robotizing GTAW through learning human response, Welding in the World (2025) 1–11

    2025

  50. [51]

    Mattera, J

    G. Mattera, J. Polden, J. Norrish, Monitoring the gas metal arc additive manufacturing process using unsupervised machine learning, Welding in the World 68 (11) (2024) 2853–2867

    2024

  51. [52]

    Purohit, T

    H. Purohit, T. Endo, M. Yamamoto, Y . Kawaguchi, Hierarchical conditional variational autoencoder based acoustic anomaly detection, in: 2022 30th European Signal Processing Conference (EUSIPCO), IEEE, 2022, pp. 274–278

    2022

  52. [53]

    Q. Luo, N. Huang, D. L. Bartles, A. M. Beese, Geometry -informed multimodal variational autoencoder for real-time prediction of properties for ti–6al–4v fabricated using pbf-lb, Journal of Intelligent Manufacturing (2025) 1–20

    2025