Pith. sign in

REVIEW 2 major objections 5 minor 107 references

Automated Data Engineering and Feature Selection for the Case Study of Warpage Detection in Fused Deposition Modeling

T0 review · 2 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Pith's one-line read A multi-armed-bandit AutoML loop that ranks features with SHAP raises FDM warpage-detection test AUC from 0.925 to 0.973 and lifts mean reward by more than fifty percent.

desk verdict A useful AutoML-for-manufacturing idea is undermined by a reward function that cannot produce the reported numbers and by test-set selection of the feature count. read the letter →

arxiv 2607.18515 v1 pith:GW5YXWRR submitted 2026-07-20 cs.LG cs.AI

classification cs.LGcs.AI
keywords automateddataprocessingmulti-armedbanditSHAPfeatureselectionfuseddepositionmodelingwarpagedetectionAutoMLexplainableAIsensor-basedprocessmonitoring
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 is trying to establish that an automated data-processing (ADP) framework, driven by a reinforcement-learning-style multi-armed bandit and guided by SHAP feature rankings, can find a model-feature combination that beats training on all available features for warpage detection in fused deposition modeling. On 217 sensor datasets from FDM printing, the framework converges to a Random Forest trained on the top-20 SHAP-ranked features, reporting a held-out test AUC of 0.9731 versus 0.9248 for the same model with all features, and a mean reward increase from 0.3347 to 0.5148. The authors interpret this as evidence that explainable-AI-based feature selection, integrated into an automated model-selection loop, yields more accurate and more stable predictions while also reducing the input dimensionality. If correct, the practical payoff is that manufacturers can build high-performing defect detectors with less manual feature engineering and less domain-specific ML expertise.

What carries the argument

The engine is the ADP framework, a multi-armed bandit whose states index 24 model-feature extraction configurations (8 feature-engineering scenarios times 3 classifiers: Random Forest, XGBoost, and a multi-layer perceptron). Actions select a configuration, and rewards come from a sigmoid-shaped function of accuracy and F1-score centered at 0.9, which updates Q-values through epsilon-greedy and UCB exploration. SHAP (Shapley additive explanations) supplies a per-state feature ranking, and Algorithm 1 averages SHAP importance over 10 shuffled episodes with an exponential moving average, then picks k from {5,10,15,20,25,All} by retraining on the top-k features. That k-selection step is what car

What would settle it

Rerun the framework with a nested protocol: within each training fold, choose k for each state on a validation fold, then evaluate the selected configuration once on a truly untouched test set. Compare the Random Forest trained on the top-20 SHAP features against the same model trained on all features. If the AUC margin shrinks to roughly zero or reverses, the reported 0.9731 value is an artifact of k-selection on the test set rather than a genuine advantage of SHAP feature ranking.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a reinforcement-learning-inspired AutoML pipeline can converge to a single best model-feature pair for warpage detection: a Random Forest trained on the top-20 SHAP-ranked features (State 4, feature-extraction scenario 5). With that configuration, the test-set AUC improves from 0.9248 to 0.9731, and the mean reward for the optimal state rises from 0.3347 to 0.5148. The framework converges within 1000 iterations, with State 4 selected as the optimal policy in more than 600 of them. The authors present this as evidence that XAI-guided feature reduction integrated into the bandit's action space improves both discrimination and stability, and that the

Load-bearing premise

The load-bearing premise is that the held-out test set, which is scored for the headline AUC, was never used to pick the feature-count k; Algorithm 1 computes k-selection accuracy on D_test, so the claim rests on that choice being benign.

Editorial extensions

If this is right

  • The optimal policy is a Random Forest on the top-20 SHAP features, reaching test AUC 0.9731 versus 0.9248 for the all-feature baseline.
  • Mean reward for the chosen state rises from 0.3347 to 0.5148, a more than fifty percent improvement, with reduced skewness and a more balanced reward distribution.
  • SHAP-selected features outperform PCA and mutual information: PCA yields negative rewards for most k, while MI only approaches SHAP's reward at k values near the full feature set, where the two methods share 80-100% of features.
  • The framework converges within 1000 iterations, selecting State 4 in more than 600 of them, indicating stable convergence to a single policy.
  • The search procedure has a lower computational complexity than exhaustive global search, O(S x A x E) versus O(S x S x E), making the approach scalable to larger AutoML action spaces.

Reading between the lines

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

  • A strict nested-validation re-run, where the top-k choice is made on validation folds rather than the test set, would reveal how much of the reported AUC gain is genuinely attributable to SHAP-guided feature selection versus selection-on-test; until that is done, the size of the true advantage over full-feature training remains an open question.
  • The same ADP+SHAP loop could plausibly transfer to other small-data manufacturing monitoring tasks such as tool wear, dimensional error, or defect detection, where tabular sensor features dominate; the transfer depends on whether SHAP's model-aware ranking consistently identifies interacting features.
  • A direct test of the paper's core comparison is to fix one model and vary k from 10 to 20, measuring whether SHAP-selected features maintain their edge over MI when the two methods still disagree substantially in their rankings.
  • The paper itself notes that tabular Q-learning will not scale if the search space grows to include image or point-cloud modalities, which implies the practical next step is a deep-RL decision unit rather than the current bandit.
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

2 major / 5 minor

Summary. This paper presents an Automated Data Processing (ADP) framework for FDM warpage detection. A multi-armed bandit selects among 24 model-feature configurations (3 classifiers × 8 feature schemes). The framework first trains models on all features, uses SHAP to rank features, and then selects top-k subsets. Rewards based on accuracy and F1 update Q-values. The authors claim that the SHAP-based version improves test AUC from 0.9248 to 0.9731 and increases mean reward by more than fifty percent, and they compare SHAP with PCA and MI. The paper includes a detailed experimental setup and dataset description.

Significance. A reliable AutoML feature-selection framework for manufacturing would be valuable, and the authors have collected a substantial dataset and compared three feature-selection strategies. However, the quantitative claims are invalid as presented: the reward values cannot be produced by Eq. (6), and the evaluation protocol leaks test information into configuration selection. These issues affect the central claims of convergence, reward improvement, and AUC improvement. The strengths of the experimental design do not outweigh these fundamental validity problems.

major comments (2)
  1. [Section 3.3, Eq. (6); Table 7; Fig. 4(a); Table 8] The reward is R = 0.5[sigmoid(acc−0.9)−0.5] + 0.5[sigmoid(F1−0.9)−0.5]. For any acc,F1∈[0,1], the maximum sigmoid argument is 0.1, so each term is at most 1/(1+e^{-0.1})−0.5 ≈ 0.0250, and thus R ≤ 0.025. Since Eq. (1) updates Q as a convex combination of rewards, every Q-value is also ≤ 0.025. Table 7 reports mean rewards of 0.3347 and 0.5148; Fig. 4(a) reports Q = 0.5469; Table 8 reports Q = 0.5369. These values are 13–20× above the maximum allowed by the paper's own equations. The statement in §3.3 that rewards lie approximately in [−0.5,0.5] is also inconsistent with Eq. (6). The reward-based half of the central claim, including the >50% mean-reward improvement, is therefore unsupported.
  2. [Algorithm 1; §4.2; Fig. 6] In the second loop of Algorithm 1, for each state the model is retrained on D_train for k ∈ {5,10,15,20,25,All}, and K_opt is chosen as the k with the highest accuracy computed on D_test. The same D_test is then used to report the test-set AUC in Fig. 6 and to select the optimal state. §4.2 states that test data are 'completely unseen during model training, validation, decision-state evaluation, and reward computation,' which is directly contradicted by Algorithm 1. The comparison with the full-feature baseline is also unfair because the baseline receives no equivalent k-selection on test data. This invalidates the headline AUC improvement as an unbiased estimate.
minor comments (5)
  1. [Section 3.1 vs §3.3] The learning rate α is set to 0.2 in Section 3.1 but described as α = 0.1 in Section 3.3. Since Eq. (1) and the reported Q-values depend on α, this inconsistency prevents exact reproduction.
  2. [Section 3.2.3] The text says 'Equation 2 describes how the SHAP criterion determines...' but the SHAP formula is Eq. (5), not Eq. (2). The cross-reference is incorrect.
  3. [Fig. 6] The subfigure labels are confusing: panels (a)–(e) appear under both the all-features and SHAP-selected-features strategies, and the text refers to 'Fig. 6(b)' for the SHAP configuration. Please relabel panels uniquely or refer to the strategy in the caption.
  4. [Appendix] The code link is given as 'Project Code' with no URL or repository identifier. As written, the appendix does not provide the promised code.
  5. [Abstract/§4.1] The abstract and Section 4.1 also report average Q-values of 0.1895 and 0.3101. These are likewise above the 0.025 bound from Eq. (6) and should be corrected or reconciled with the stated reward function.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline test-AUC gain is partly constructed: Algorithm 1 picks K_opt by accuracy on D_test and the same held-out test set then produces the reported AUC; the paper's own reward function also makes the reported reward/Q values unattainable.

  1. fitted input called prediction [Algorithm 1 (Section 3.3) with Section 4.2 'Evaluation of ADP results' and Fig. 6]
    "for k ∈ {5,10,15,20,25,All} do Select top-k features F_k from F Iavg[s_i] Shuffle D[s_i], split into Dtrain, Dtest with F_k Train new model M_i^k on Dtrain A_k ← accuracy of M_i^k on Dtest if A_k > A_max then A_max ← A_k K_opt[s_i] ← k end if end for ... All AUC values reported in this study are calculated solely on the held-out test data that remain completely unseen during model training, validation, decision-state evaluation, and reward computation."

    K_opt is selected by maximizing A_k computed on D_test. The final best configuration (State 4, top-20 SHAP features) is then scored on the held-out test set to produce the headline AUC increase from 0.9248 to 0.9731. The reported improvement is therefore not an unbiased prediction of a fixed pipeline: it is the best of six feature-count configurations chosen on the test set. The full-feature baseline receives no such test-set selection, so the comparison is partly by construction.

full rationale

The main circularity is the test-set-dependent selection of K_opt in Algorithm 1 combined with the claim in Section 4.2 that all AUC values come from a test set 'completely unseen' during decision-state evaluation. Since K_opt is chosen by A_k on D_test, the reported AUC gain is partly a fitted quantity; this is a genuine fitted-input-called-prediction circularity affecting the paper's headline result. I found no additional circular reductions: the SHAP feature rankings, the Q-update rule, and the comparison against PCA/MI are described as empirical evaluations rather than results derived from their own definitions. Two further validity problems should be flagged, though they are not circularities. First, the reward/Q numbers are internally inconsistent with Eq. 6: because accuracy and F1 are at most 1, each sigmoid term is at most sigmoid(0.1)-0.5 ≈ 0.025, so R (w1=w2=0.5) cannot exceed 0.025, yet Table 7 reports mean rewards 0.3347/0.5148 and Fig. 4(a) reports Q=0.5469; the inserted clarification claiming a [−0.5,0.5] range does not repair this. This makes the reward-based half of the central claim unverifiable but does not by itself raise the circularity score. Second, the computational-complexity advantage O(S×A×E) vs O(S×S×E) is imported from the authors' prior work by self-citation rather than derived in this paper; it is ancillary, so I treat it as a minor self-citation rather than load-bearing circularity. Overall, one central 'prediction' reduces in part to test-set selection, so the circularity score is 6 rather than 0-2; the result still has independent empirical content (the SHAP-ranked features and model training are non-trivially executed), so it is not a full 8-10 construction.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

All bandit and reward constants are hand-set (α, ε, c, sigmoid center 0.9, weights 0.5/0.5), η is unspecified, and the per-state feature count k is fitted on the held-out test set. Ground-truth warpage labels rest entirely on visual inspection. The ADP framework is imported from the authors' prior work, and the only first-principles content is the standard Shapley definition (Eqs. 4–5). No new physical entities are postulated.

free parameters (7)
  • Learning rate α (Q-update) = 0.2 (§3.1) vs 0.1 (§3.3) — inconsistent
    Hand-set to stabilize Q-updates; the text reports two different values.
  • Exploration rate ε = 0.1
    Hand-set in §3.1 to balance exploration and exploitation in action selection.
  • UCB exploration constant c = 0.5
    Hand-set in §3.1 to control the exploration bonus in Eq. 2.
  • Reward sigmoid center = 0.9 (accuracy/F1 threshold)
    §3.3: chosen so accuracies/F1 below 0.9 receive negative reward; a hand-set target level defining the reward scale.
  • Reward weights w1, w2 = 0.5, 0.5
    §3.3: equal weighting of accuracy and F1 terms in Eq. 6.
  • SHAP EMA update rate η = not specified
    Algorithm 1 takes η as input but never assigns it a value anywhere in the text.
  • Top-k feature count K_opt = k ∈ {5,10,15,20,25,All}, chosen per state by D_test accuracy
    Algorithm 1 fits k to maximize held-out test accuracy; the same test set then scores the final model. Selection on the test set.
assumptions (5)
  • domain assumption Visual inspection yields reliable warpage ground truth
    §2: 'Warpage labels were assigned based on visual inspection'; no metrology, inter-rater check, or uncertainty is reported.
  • domain assumption The 217 prints are independent, identically distributed samples for classification
    §2–§4: each print is one labeled sample; an 80/20 grouped split yields ~43 test samples, and within-configuration replicates are treated as independent without clustering.
  • ad hoc to paper The reward function (Eq. 6) with center 0.9 and sigmoid scaling provides a valid optimization signal
    §3.3: the sigmoid transform, its center, the −0.5 centering, and equal weights are introduced for this paper and are not derived or validated against any external criterion.
  • domain assumption SHAP values estimated on the trained model over the test set faithfully rank feature importance for generalization
    Algorithm 1 applies SHAP to D_test and assumes top-k by SHAP is the best subset — an assumption the paper's own results only partially support ('negative effect on the rest of the ML models').
  • standard math Standard ML training assumptions (5-fold CV, standardization on training folds, loss minimization)
    §3.2.1 and §4: standardization parameters are computed from training folds; five-fold cross-validation is used on the 80% pool.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Automated Data Engineering and Feature Selection for the Case Study of Warpage Detection in Fused Deposition Modeling." pith.science (2026). https://pith.science/paper/GW5YXWRR

@misc{pith2026260718515,
  author       = {Pith},
  title        = {Pith review of: Automated Data Engineering and Feature Selection for the Case Study of Warpage Detection in Fused Deposition Modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GW5YXWRR}},
  note         = {Machine review of arXiv:2607.18515}
}
abstract

This study contributes toward development of an Automated Data Processing (ADP) framework designed to evaluate and reinforce optimal machine learning model-feature combinations for predictive tasks in fused deposition modeling (FDM) process datasets. The methodology is centered around a reinforcement learning-inspired policy updating mechanism, where multiple machine learning models are trained on both full feature sets and feature subsets selected through Shapley-based Explainable AI (SHAP XAI) across 217 datasets. At each episode, the framework assesses the predictive accuracy and F1-scores of each model-feature pair, computes a scalar reward, and updates $Q$ values to guide future model selection. SHAP XAI feature importance was employed to generate reduced yet informative feature subsets to enable the framework to explore performance with dimensionality. The policy was shown to evolve over multiple episodes, with reward distributions used to visualize performance stability. Overall, results indicate that leveraging the ADP framework through XAI algorithms successfully converges toward optimal model-feature configurations with improved accuracy and stability. Specifically, the proposed framework improves the test-set AUC from 0.9248 to 0.9731 and increases the mean reward value by more than fifty percent compared with the baseline full-feature configuration.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

107 extracted references · 4 linked inside Pith

  1. [1]

    Campbell, S. L. and Gear, C. W. The index of general nonlinear D A E S. Numer. M ath. 1995

  2. [2]

    Slifka, M. K. and Whitton, J. L. Clinical implications of dysregulated cytokine production. J. M ol. M ed. 2000. doi:10.1007/s001090000086

  3. [3]

    Quasimonotonicity, regularity and duality for nonlinear systems of partial differential equations

    Hamburger, C. Quasimonotonicity, regularity and duality for nonlinear systems of partial differential equations. Ann. Mat. Pura. Appl. 1995

  4. [4]

    Geddes, K. O. and Czapor, S. R. and Labahn, G. Algorithms for C omputer A lgebra. 1992

  5. [5]

    Software engineering---from auxiliary to key technologies

    Broy, M. Software engineering---from auxiliary to key technologies. Software Pioneers. 1992

  6. [6]

    Applied Sciences , volume=

    Investigation of FDM-based 3D printing for optimized tooling in automotive and electronics sheet metal cutting , author=. Applied Sciences , volume=. 2025 , publisher=

  7. [7]

    A Review , author=

    Defects in 3D printing and strategies to enhance quality of FFF additive manufacturing. A Review , author=. ChemRxiv , year=

  8. [8]

    Journal of Composite Materials , pages=

    A multi-objective investigation of warpage and mechanical performance in FDM-printed thermoplastic composites , author=. Journal of Composite Materials , pages=. 2025 , publisher=

Show all 107 references
  1. [9]

    Case Studies in Construction Materials , volume=

    A hybrid strategy of AutoML and SHAP for automated and explainable concrete strength prediction , author=. Case Studies in Construction Materials , volume=. 2023 , publisher=

  2. [10]

    The International Journal of Advanced Manufacturing Technology , volume=

    Automated data processing for efficient development of multimodal machine learning models in tool wear detection , author=. The International Journal of Advanced Manufacturing Technology , volume=. 2025 , publisher=

  3. [11]

    SoftwareX , volume=

    auto-xFS: An explanation-based feature selection tool for more meaningful and trustworthy machine learning models , author=. SoftwareX , volume=. 2025 , publisher=

  4. [12]

    Conductive P olymers. 1981

  5. [13]

    Smith, S. E. Neuromuscular blocking drugs in man. Neuromuscular junction. H andbook of experimental pharmacology. 1976

  6. [14]

    Chung, S. T. and Morris, R. L. Isolation and characterization of plasmid deoxyribonucleic acid from Streptomyces fradiae. 1978

  7. [15]

    and AghaKouchak, A

    Hao, Z. and AghaKouchak, A. and Nakhjiri, N. and Farahmand, A. Global integrated drought monitoring and prediction system (GIDMaPS) data sets. 2014

  8. [16]

    Babichev, S. A. and Ries, J. and Lvovsky, A. I. Quantum scissors: teleportation of single-mode optical states by means of a nonlocal single photon. 2002

  9. [17]

    and Buchalla, G

    Beneke, M. and Buchalla, G. and Dunietz, I. Mixing induced CP asymmetries in inclusive B decays. Phys. L ett. 1997. arXiv:0707.3168

  10. [18]

    deep SIP : deep learning of S upernova I a P arameters

    Stahl, B. deep SIP : deep learning of S upernova I a P arameters. 2020. ascl:2006.023

  11. [19]

    Abbott, T. M. C. and others. Dark Energy Survey Year 1 Results: Constraints on Extended Cosmological Models from Galaxy Clustering and Weak Lensing. Phys. Rev. D. 2019. doi:10.1103/PhysRevD.99.123505. arXiv:1810.02499

  12. [20]

    Proceedings of the machining innovations conference (MIC) , year=

    Using AutoML to optimize shape error prediction in milling processes , author=. Proceedings of the machining innovations conference (MIC) , year=

  13. [21]

    Journal of manufacturing systems , volume=

    Data-driven smart manufacturing , author=. Journal of manufacturing systems , volume=. 2018 , publisher=

  14. [22]

    2020 25th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA) , volume=

    Towards real-time process monitoring and machine learning for manufacturing composite structures , author=. 2020 25th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA) , volume=. 2020 , organization=

  15. [23]

    Sustainability , volume=

    Machine learning in predictive maintenance towards sustainable smart manufacturing in industry 4.0 , author=. Sustainability , volume=. 2020 , publisher=

  16. [24]

    Engineered Science , volume=

    Role and applications of artificial intelligence and machine learning in manufacturing engineering: a review , author=. Engineered Science , volume=. 2024 , publisher=

  17. [25]

    Journal of Intelligent Manufacturing , volume=

    Machine-learning based process monitoring for automated composites manufacturing , author=. Journal of Intelligent Manufacturing , volume=. 2025 , publisher=

  18. [26]

    Applied Composite Materials , volume=

    Quantitative damage monitoring of filament wound composites by using machine learning-based techniques , author=. Applied Composite Materials , volume=. 2024 , publisher=

  19. [27]

    2018 , publisher=

    Reinforcement Learning, second edition: An Introduction , author=. 2018 , publisher=

  20. [28]

    Knowledge-Based Systems , volume=

    AutoML: A survey of the state-of-the-art , author=. Knowledge-Based Systems , volume=. 2021 , publisher=

  21. [29]

    IEEE Transactions on Image Processing , volume=

    Recognition from web data: A progressive filtering approach , author=. IEEE Transactions on Image Processing , volume=. 2018 , publisher=

  22. [30]

    ACM Sigkdd Explorations Newsletter , volume=

    Learning from imbalanced data sets with boosting and data generation: the databoost-im approach , author=. ACM Sigkdd Explorations Newsletter , volume=. 2004 , publisher=

  23. [31]

    2016 international conference on digital image computing: techniques and applications (DICTA) , pages=

    Understanding data augmentation for classification: when to warp? , author=. 2016 international conference on digital image computing: techniques and applications (DICTA) , pages=. 2016 , organization=

  24. [32]

    arXiv preprint arXiv:2111.14260 , year=

    A practical guide on explainable ai techniques applied on biomedical use case applications , author=. arXiv preprint arXiv:2111.14260 , year=

  25. [33]

    Journal of the Franklin Institute , volume=

    The jensen-shannon divergence , author=. Journal of the Franklin Institute , volume=. 1997 , publisher=

  26. [34]

    Proceedings , volume=

    On the estimation of mutual information , author=. Proceedings , volume=. 2020 , organization=

  27. [35]

    Wiley interdisciplinary reviews: computational statistics , volume=

    Principal component analysis , author=. Wiley interdisciplinary reviews: computational statistics , volume=. 2010 , publisher=

  28. [36]

    International Manufacturing Science and Engineering Conference , volume=

    Integrated Data Processing and Model Selection in Machine Learning Framework Development to Predict Dimensional Errors in Wire Arc Additive Manufacturing (WAAM) , author=. International Manufacturing Science and Engineering Conference , volume=. 2024 , organization=

  29. [37]

    Procedia Computer Science , volume=

    Benchmarking AutoML-supported lead time prediction , author=. Procedia Computer Science , volume=. 2022 , publisher=

  30. [38]

    IFIP international conference on artificial intelligence applications and innovations , pages=

    Production time prediction for contract manufacturing industries using automated machine learning , author=. IFIP international conference on artificial intelligence applications and innovations , pages=. 2022 , organization=

  31. [39]

    Processes , volume=

    Improved Fault Classification for Predictive Maintenance in Industrial IoT Based on AutoML: A Case Study of Ball-Bearing Faults , author=. Processes , volume=. 2023 , publisher=

  32. [40]

    2020 , month =

    PyCaret: An open source, low-code machine learning library in Python , author =. 2020 , month =

  33. [41]

    Journal of Machine Learning Research , volume=

    Autokeras: An automl library for deep learning , author=. Journal of Machine Learning Research , volume=

  34. [42]

    arXiv:2007.04074 [cs.LG] , year =

    Auto-Sklearn 2.0: Hands-free AutoML via Meta-Learning , author =. arXiv:2007.04074 [cs.LG] , year =

  35. [43]

    NeuroImage , volume=

    Assessing the effectiveness of spatial PCA on SVM-based decoding of EEG data , author=. NeuroImage , volume=. 2024 , publisher=

  36. [44]

    Workshop on New Challenges for Feature Selection in Data Mining and Knowledge Discovery , year=

    A comparison of classiffication accuracy achieved with wrappers, filters and PCA , author=. Workshop on New Challenges for Feature Selection in Data Mining and Knowledge Discovery , year=

  37. [45]

    Processes , volume=

    Research on Valve Life Prediction Based on PCA-PSO-LSSVM , author=. Processes , volume=. 2023 , publisher=

  38. [46]

    Additive Manufacturing , volume=

    In-process monitoring of part warpage in fused filament fabrication through the analysis of the repulsive force acting on the extruder , author=. Additive Manufacturing , volume=. 2022 , publisher=

  39. [47]

    Polymers 2021, 13, 3849 , author=

    Optimization of the Warpage of Fused Deposition Modeling Parts Using Finite Element Method. Polymers 2021, 13, 3849 , author=. 2021 , publisher=

  40. [48]

    Journal of Intelligent Manufacturing , volume=

    Warpage detection in 3D printing of polymer parts: a deep learning approach , author=. Journal of Intelligent Manufacturing , volume=. 2025 , publisher=

  41. [49]

    Mechanical Systems and Signal Processing , volume=

    Mutual information based anomaly detection of monitoring data with attention mechanism and residual learning , author=. Mechanical Systems and Signal Processing , volume=. 2023 , publisher=

  42. [50]

    Engineering with Computers , volume=

    Optimal sensor placement for digital twin based on mutual information and correlation with multi-fidelity data , author=. Engineering with Computers , volume=. 2024 , publisher=

  43. [51]

    Electronic Markets , volume=

    Designing a feature selection method based on explainable artificial intelligence , author=. Electronic Markets , volume=. 2022 , publisher=

  44. [52]

    Structural Control and Health Monitoring , volume=

    Explainable AI-Driven Optimal Feature Selection for the Identification of Structural Damage , author=. Structural Control and Health Monitoring , volume=. 2025 , publisher=

  45. [53]

    International Conference on Neural Information Processing , pages=

    Explainable AI in Feature Selection: Improving Classification Performance on Imbalanced Datasets , author=. International Conference on Neural Information Processing , pages=. 2024 , organization=

  46. [54]

    Processes , volume=

    Solutions of feature and Hyperparameter model selection in the intelligent manufacturing , author=. Processes , volume=. 2022 , publisher=

  47. [55]

    Journal of Manufacturing Processes , volume=

    Data-driven smart manufacturing: Tool wear monitoring with audio signals and machine learning , author=. Journal of Manufacturing Processes , volume=. 2019 , publisher=

  48. [56]

    Journal of manufacturing systems , volume=

    Deep learning for smart manufacturing: Methods and applications , author=. Journal of manufacturing systems , volume=. 2018 , publisher=

  49. [57]

    Computers & Chemical Engineering , volume=

    Feature engineering in big data analytics for IoT-enabled smart manufacturing--Comparison between deep learning and statistical learning , author=. Computers & Chemical Engineering , volume=. 2020 , publisher=

  50. [58]

    IEEE/CAA Journal of Automatica Sinica , volume=

    AI-based modeling and data-driven evaluation for smart manufacturing processes , author=. IEEE/CAA Journal of Automatica Sinica , volume=. 2020 , publisher=

  51. [59]

    2008 American Control Conference , pages=

    Principal component based k-nearest-neighbor rule for semiconductor process fault detection , author=. 2008 American Control Conference , pages=. 2008 , organization=

  52. [60]

    2006 IEEE Odyssey-The Speaker and Language Recognition Workshop , pages=

    Feature selection based on genetic algorithms for speaker recognition , author=. 2006 IEEE Odyssey-The Speaker and Language Recognition Workshop , pages=. 2006 , organization=

  53. [61]

    Multi-Modal Explainable Artificial Intelligence for neural network-based tool wear detection in machining , journal =

    Saleh Valizadeh Sotubadi and Shyam Sasi Pallissery and Vinh Nguyen , keywords =. Multi-Modal Explainable Artificial Intelligence for neural network-based tool wear detection in machining , journal =. 2025 , issn =

  54. [62]

    Wiley interdisciplinary reviews: computational statistics 2(4):433--459

    Abdi H, Williams LJ (2010) Principal component analysis. Wiley interdisciplinary reviews: computational statistics 2(4):433--459

  55. [63]

    ://www.pycaret.org, pyCaret version 1.0

    Ali M (2020) PyCaret: An open source, low-code machine learning library in Python. ://www.pycaret.org, pyCaret version 1.0

  56. [64]

    Procedia Computer Science 200:482--494

    Bender J, Trat M, Ovtcharova J (2022) Benchmarking automl-supported lead time prediction. Procedia Computer Science 200:482--494

  57. [65]

    arXiv preprint arXiv:211114260

    Bennetot A, Donadello I, Qadi AE, et al (2021) A practical guide on explainable ai techniques applied on biomedical use case applications. arXiv preprint arXiv:211114260

  58. [66]

    Journal of Intelligent Manufacturing 36(5):3129--3141

    Bhandarkar VV, Kumar A, Tandon P (2025) Warpage detection in 3d printing of polymer parts: a deep learning approach. Journal of Intelligent Manufacturing 36(5):3129--3141

  59. [67]

    Engineered Science 29:1088

    Bunian S, Al-Ebrahim MA, Nour AA (2024) Role and applications of artificial intelligence and machine learning in manufacturing engineering: a review. Engineered Science 29:1088

  60. [68]

    In: Proceedings, MDPI, p 31

    Carrara N, Ernst J (2020) On the estimation of mutual information. In: Proceedings, MDPI, p 31

  61. [69]

    Sustainability 12(19):8211

    C nar ZM, Abdussalam Nuhu A, Zeeshan Q, et al (2020) Machine learning in predictive maintenance towards sustainable smart manufacturing in industry 4.0. Sustainability 12(19):8211

  62. [70]

    In: Proceedings of the machining innovations conference (MIC)

    Denkena B, Dittrich MA, Lindauer M, et al (2020) Using automl to optimize shape error prediction in milling processes. In: Proceedings of the machining innovations conference (MIC)

  63. [71]

    a review

    Erokhin K, Naumov S, Ananikov V (2023) Defects in 3d printing and strategies to enhance quality of fff additive manufacturing. a review. ChemRxiv

  64. [72]

    arXiv:200704074 [csLG]

    Feurer M, Eggensperger K, Falkner S, et al (2020) Auto-sklearn 2.0: Hands-free automl via meta-learning. arXiv:200704074 [csLG]

  65. [73]

    IEEE/CAA Journal of Automatica Sinica 7(4):1026--1037

    Ghahramani M, Qiao Y, Zhou MC, et al (2020) Ai-based modeling and data-driven evaluation for smart manufacturing processes. IEEE/CAA Journal of Automatica Sinica 7(4):1026--1037

  66. [74]

    ACM Sigkdd Explorations Newsletter 6(1):30--39

    Guo H, Viktor HL (2004) Learning from imbalanced data sets with boosting and data generation: the databoost-im approach. ACM Sigkdd Explorations Newsletter 6(1):30--39

  67. [75]

    Processes 11(5):1507

    Hadi RH, Hady HN, Hasan AM, et al (2023) Improved fault classification for predictive maintenance in industrial iot based on automl: A case study of ball-bearing faults. Processes 11(5):1507

  68. [76]

    In: 2008 American Control Conference, IEEE, pp 1606--1611

    He QP, Wang J (2008) Principal component based k-nearest-neighbor rule for semiconductor process fault detection. In: 2008 American Control Conference, IEEE, pp 1606--1611

  69. [77]

    Knowledge-Based Systems 212:106622

    He X, Zhao K, Chu X (2021) Automl: A survey of the state-of-the-art. Knowledge-Based Systems 212:106622

  70. [78]

    In: Workshop on New Challenges for Feature Selection in Data Mining and Knowledge Discovery

    Janecek AG, Gansterer WN (2008) A comparison of classiffication accuracy achieved with wrappers, filters and pca. In: Workshop on New Challenges for Feature Selection in Data Mining and Knowledge Discovery

  71. [79]

    Journal of Machine Learning Research 24(6):1--6

    Jin H, Chollet F, Song Q, et al (2023) Autokeras: An automl library for deep learning. Journal of Machine Learning Research 24(6):1--6

  72. [80]

    Mechanical Systems and Signal Processing 182:109607

    Lei X, Xia Y, Wang A, et al (2023) Mutual information based anomaly detection of monitoring data with attention mechanism and residual learning. Mechanical Systems and Signal Processing 182:109607

  73. [81]

    Journal of Manufacturing Processes 48:66--76

    Li Z, Liu R, Wu D (2019) Data-driven smart manufacturing: Tool wear monitoring with audio signals and machine learning. Journal of Manufacturing Processes 48:66--76

  74. [82]

    Journal of the Franklin Institute 334(2):307--318

    Men \'e ndez ML, Pardo JA, Pardo L, et al (1997) The jensen-shannon divergence. Journal of the Franklin Institute 334(2):307--318

  75. [83]

    Additive Manufacturing 49:102505

    Moretti M, Senin N (2022) In-process monitoring of part warpage in fused filament fabrication through the analysis of the repulsive force acting on the extruder. Additive Manufacturing 49:102505

  76. [84]

    Journal of Intelligent Manufacturing 36(2):1095--1110

    Mujtaba A, Islam F, Kaeding P, et al (2025) Machine-learning based process monitoring for automated composites manufacturing. Journal of Intelligent Manufacturing 36(2):1095--1110

  77. [85]

    Computers & Chemical Engineering 141:106970

    Shah D, Wang J, He QP (2020) Feature engineering in big data analytics for iot-enabled smart manufacturing--comparison between deep learning and statistical learning. Computers & Chemical Engineering 141:106970

  78. [86]

    Processes 11(5):1396

    Shi M, Tan P, Qin L, et al (2023) Research on valve life prediction based on pca-pso-lssvm. Processes 11(5):1396

  79. [87]

    In: International Conference on Neural Information Processing, Springer, pp 303--318

    Siddique Ayon S, Ebrahim Hossain M, Ullah Miah MS, et al (2024) Explainable ai in feature selection: Improving classification performance on imbalanced datasets. In: International Conference on Neural Information Processing, Springer, pp 303--318

  80. [88]

    Engineering Applications of Artificial Intelligence 144:110141

    Sotubadi SV, Pallissery SS, Nguyen V (2025) Multi-modal explainable artificial intelligence for neural network-based tool wear detection in machining. Engineering Applications of Artificial Intelligence 144:110141

  81. [89]

    In: IFIP international conference on artificial intelligence applications and innovations, Springer, pp 262--273

    Sousa A, Ferreira L, Ribeiro R, et al (2022) Production time prediction for contract manufacturing industries using automated machine learning. In: IFIP international conference on artificial intelligence applications and innovations, Springer, pp 262--273

  82. [90]

    Case Studies in Construction Materials 19:e02405

    Sun B, Cui W, Liu G, et al (2023) A hybrid strategy of automl and shap for automated and explainable concrete strength prediction. Case Studies in Construction Materials 19:e02405

  83. [91]

    Adaptive Computation and Machine Learning series, MIT Press

    Sutton R, Barto A (2018) Reinforcement Learning, second edition: An Introduction. Adaptive Computation and Machine Learning series, MIT Press

  84. [92]

    polymers 2021, 13, 3849

    Syrlybayev D, Zharylkassyn B, Seisekulova A, et al (2021) Optimization of the warpage of fused deposition modeling parts using finite element method. polymers 2021, 13, 3849

  85. [93]

    Applied Sciences 15(1):442

    Szalai S, Sz \' v \'o s BF, Nemes V, et al (2025) Investigation of fdm-based 3d printing for optimized tooling in automotive and electronics sheet metal cutting. Applied Sciences 15(1):442

  86. [94]

    Journal of Composite Materials p 00219983251330717

    Tanabi H, Ndacyayisenga W, Ghaziani M (2025) A multi-objective investigation of warpage and mechanical performance in fdm-printed thermoplastic composites. Journal of Composite Materials p 00219983251330717

  87. [95]

    Journal of manufacturing systems 48:157--169

    Tao F, Qi Q, Liu A, et al (2018) Data-driven smart manufacturing. Journal of manufacturing systems 48:157--169

  88. [96]

    In: International Manufacturing Science and Engineering Conference, American Society of Mechanical Engineers, p V001T01A037

    Valizadeh Sotubadi S, Hendrickson N, Nguyen V (2024) Integrated data processing and model selection in machine learning framework development to predict dimensional errors in wire arc additive manufacturing (waam). In: International Manufacturing Science and Engineering Confer...

  89. [97]

    The International Journal of Advanced Manufacturing Technology 139(5):2505--2524

    Valizadeh Sotubadi S, Mahjourian N, Nguyen V (2025) Automated data processing for efficient development of multimodal machine learning models in tool wear detection. The International Journal of Advanced Manufacturing Technology 139(5):2505--2524

  90. [98]

    Processes 10(5):862

    Wang CY, Huang CY, Chiang YH (2022) Solutions of feature and hyperparameter model selection in the intelligent manufacturing. Processes 10(5):862

  91. [99]

    SoftwareX 31:102268

    Wang H, Aligon J, Zhou H, et al (2025 a ) auto-xfs: An explanation-based feature selection tool for more meaningful and trustworthy machine learning models. SoftwareX 31:102268

  92. [100]

    Journal of manufacturing systems 48:144--156

    Wang J, Ma Y, Zhang L, et al (2018) Deep learning for smart manufacturing: Methods and applications. Journal of manufacturing systems 48:144--156

  93. [101]

    Engineering with Computers 40(2):1289--1308

    Wang S, Lai X, He X, et al (2024) Optimal sensor placement for digital twin based on mutual information and correlation with multi-fidelity data. Engineering with Computers 40(2):1289--1308

  94. [102]

    Structural Control and Health Monitoring 2025(1):7253150

    Wang X, Wei Z, Wang Z, et al (2025 b ) Explainable ai-driven optimal feature selection for the identification of structural damage. Structural Control and Health Monitoring 2025(1):7253150

  95. [103]

    Wong SC, Gatt A, Stamatescu V, et al (2016) Understanding data augmentation for classification: when to warp? In: 2016 international conference on digital image computing: techniques and applications (DICTA), IEEE, pp 1--6

  96. [104]

    IEEE Transactions on Image Processing 27(11):5303--5315

    Yang J, Sun X, Lai YK, et al (2018) Recognition from web data: A progressive filtering approach. IEEE Transactions on Image Processing 27(11):5303--5315

  97. [105]

    Electronic Markets 32(4):2159--2184

    Zacharias J, von Zahn M, Chen J, et al (2022) Designing a feature selection method based on explainable artificial intelligence. Electronic Markets 32(4):2159--2184

  98. [106]

    In: 2006 IEEE Odyssey-The Speaker and Language Recognition Workshop, IEEE, pp 1--8

    Zamalloa M, Bordel G, Rodr \' guez LJ, et al (2006) Feature selection based on genetic algorithms for speaker recognition. In: 2006 IEEE Odyssey-The Speaker and Language Recognition Workshop, IEEE, pp 1--8

  99. [107]

    NeuroImage 293:120625

    Zhang G, Carrasco CD, Winsler K, et al (2024) Assessing the effectiveness of spatial pca on svm-based decoding of eeg data. NeuroImage 293:120625

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.