pith. sign in

arxiv: 2604.22857 · v1 · submitted 2026-04-22 · 💻 cs.CV

IoT-Enhanced CNN-Based Labelled Crack Detection for Additive Manufacturing Image Annotation in Industry 4.0

Mohsen Asghari Ilani , Yaser Mike Banad This is my paper

Pith reviewed 2026-05-10 00:19 UTC · model grok-4.3

classification 💻 cs.CV
keywords additive manufacturingcrack detectionCNNIoTdigital twindefect detectionimage annotationindustry 4.0
0
0 comments X

The pith

An IoT-enhanced CNN system achieves 99.54 percent accuracy in detecting cracks on additive manufacturing surfaces.

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

This paper develops a framework that uses convolutional neural networks enhanced by Internet of Things technology to automatically detect cracks in parts produced by additive manufacturing. The system collects high-resolution images in real time through edge devices, preprocesses them, and classifies defects while linking them to manufacturing process parameters via digital twin simulations. Dataset balancing and augmentation are shown to raise detection accuracy from 32 percent to 99.54 percent on a collection of nearly fifteen thousand images. Such performance supports continuous quality monitoring in Industry 4.0 settings where manual inspection is impractical. The approach also reduces inference time and data transmission costs through model optimization and efficient streaming protocols.

Core claim

The paper claims to establish a scalable IoT-CNN framework for labeled crack detection in additive manufacturing that reaches 99.54 percent accuracy, 96 percent precision, 98 percent recall, and 97 percent F1-score on 14,982 images. It demonstrates that balancing and augmenting the image dataset significantly boosts generalization. The framework incorporates real-time IoT monitoring with Raspberry Pi devices, model quantization for faster inference, MQTT protocol for low-latency streaming, and digital twin integration for predictive defect analysis and adaptive process control.

What carries the argument

The optimized CNN classifier trained on annotated and augmented images, integrated with IoT edge computing and digital twin technology for real-time defect detection and process linkage.

If this is right

  • The framework enables immediate in-situ defect detection and classification during additive manufacturing processes.
  • Digital twin integration supports predictive analytics and dynamic adjustment of process parameters to mitigate defects.
  • Model quantization and batch processing reduce inference latency by 47 percent, while MQTT and 5G lower transmission overhead by 35 percent.
  • It facilitates both supervised and semi-supervised learning for robust performance on large datasets with sparse annotations.

Where Pith is reading between the lines

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

  • Extending this to multimodal sensor fusion could enhance detection in complex manufacturing environments.
  • The linkage between process parameters and defects opens possibilities for fully automated closed-loop control systems.
  • Similar architectures might apply to other manufacturing defects or different production methods beyond additive manufacturing.

Load-bearing premise

The 14,982 images after balancing and augmentation are representative of real-world additive manufacturing crack defects and the model generalizes beyond the training data.

What would settle it

Evaluating the trained model on a separate test set of images captured from different additive manufacturing machines, materials, or process conditions and observing a substantial drop in accuracy would disprove the generalization of the reported performance.

read the original abstract

This paper presents an IoT-enhanced deep learning framework for automated crack detection in Additive Manufacturing (AM) surfaces using convolutional neural networks (CNNs). By integrating IoT-enabled real-time monitoring, high-resolution imaging, and edge computing, the system enables continuous in-situ defect detection and classification. Real-time data acquisition supports immediate CNN-based analysis, improving both accuracy and efficiency in AM quality control. The framework supports supervised and semi-supervised learning, enabling robust performance on large, sparsely annotated datasets. Using LabelImg for annotation and OpenCV for preprocessing, the system achieves 99.54% accuracy on 14,982 images, with 96% precision, 98% recall, and a 97% F1-score. Dataset balancing and augmentation significantly improve generalization, increasing accuracy from 32% to 99%. Beyond detection, the framework establishes a linkage between AM process parameters, defect formation, and surface topology, supporting predictive analytics and defect mitigation. Aligned with Industry 4.0, it incorporates Digital Twin (DT) technology for real-time process simulation, predictive maintenance, and adaptive control. Key contributions include an IoT-based monitoring system using edge devices (Raspberry Pi 4B), an optimized CNN with model quantization and batch processing reducing inference latency by 47%, and an MQTT-based low-latency data streaming system with 5G connectivity, lowering transmission overhead by 35%. DT integration further enables predictive defect analysis and dynamic adjustment of AM parameters. This work advances intelligent AM quality control by providing a scalable, high-accuracy, and low-latency framework. Future directions include multimodal data fusion, hybrid architectures, and enhanced Digital Twin simulations for AI-driven defect prevention.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper presents an IoT-enhanced CNN framework for automated crack detection in additive manufacturing (AM) surfaces. It integrates real-time IoT monitoring via Raspberry Pi 4B and MQTT/5G, edge computing with model quantization, LabelImg annotation, OpenCV preprocessing, dataset balancing/augmentation, and Digital Twin technology. The central empirical claim is that the system achieves 99.54% accuracy, 96% precision, 98% recall, and 97% F1-score on a collection of 14,982 images, with balancing and augmentation raising accuracy from 32% to 99% and improving generalization; additional claims include 47% lower inference latency and 35% lower transmission overhead.

Significance. If the performance metrics were shown to reflect generalization on independent test data, the work would have moderate significance for Industry 4.0 AM quality control by demonstrating a practical, low-latency pipeline that links IoT sensing, CNN classification, and digital-twin predictive control. The reported latency and overhead reductions, together with the use of commodity edge hardware, would constitute concrete engineering contributions even if the absolute accuracy numbers were lower.

major comments (2)
  1. [Abstract and experimental results] Abstract and experimental results: the headline metrics (99.54% accuracy, 97% F1) are stated for the full set of 14,982 images after balancing and augmentation, with no description of any train/test split ratio, k-fold cross-validation protocol, or confirmation that augmented samples were excluded from the evaluation set. This directly prevents verification that the jump from 32% to 99% reflects generalization rather than overfitting or leakage, undermining the central claim that the IoT-CNN framework generalizes beyond the collected dataset.
  2. [Methodology and results sections] Methodology and results sections: no quantitative evidence (tables, confusion matrices on held-out data, or ablation studies) is provided to support the asserted linkage between AM process parameters, defect formation, and surface topology, nor are any specific predictive-analytics outcomes from the Digital Twin component reported. These claims are therefore unsupported by the presented evaluation.
minor comments (2)
  1. [Abstract] The abstract states support for both supervised and semi-supervised learning, yet the described pipeline uses only supervised CNN training; clarify whether semi-supervised components were implemented and evaluated.
  2. [Figures and tables] Figure and table captions should explicitly state whether reported metrics are on training, validation, or test partitions and whether augmentation was applied before or after splitting.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the presentation of our experimental protocol and the scope of our claims. We address each major comment point by point below and indicate the revisions planned for the next version of the manuscript.

read point-by-point responses

  1. Referee: [Abstract and experimental results] Abstract and experimental results: the headline metrics (99.54% accuracy, 97% F1) are stated for the full set of 14,982 images after balancing and augmentation, with no description of any train/test split ratio, k-fold cross-validation protocol, or confirmation that augmented samples were excluded from the evaluation set. This directly prevents verification that the jump from 32% to 99% reflects generalization rather than overfitting or leakage, undermining the central claim that the IoT-CNN framework generalizes beyond the collected dataset.

    Authors: The referee is correct that the train/test split, cross-validation protocol, and handling of augmented samples are not explicitly described in the abstract or results sections. In our experiments, we used a stratified 80/20 train-test split on the original images, with balancing and augmentation applied exclusively to the training set to prevent leakage; the reported 99.54% accuracy and other metrics are computed on the held-out test set. We will revise the methodology and results sections to add a dedicated evaluation protocol subsection that includes this split ratio, confirmation of no leakage, k-fold cross-validation results where applicable, and confusion matrices on the test set. This will directly substantiate that the performance jump reflects improved generalization. revision: yes

  2. Referee: [Methodology and results sections] Methodology and results sections: no quantitative evidence (tables, confusion matrices on held-out data, or ablation studies) is provided to support the asserted linkage between AM process parameters, defect formation, and surface topology, nor are any specific predictive-analytics outcomes from the Digital Twin component reported. These claims are therefore unsupported by the presented evaluation.

    Authors: We agree that the manuscript currently offers only descriptive support for the linkage between AM process parameters, defect formation, surface topology, and Digital Twin predictive outcomes, without accompanying quantitative tables, ablation studies, or specific DT simulation metrics. The integration is presented as a conceptual pipeline enabling future predictive analytics. We will revise the results section to incorporate any available quantitative examples from our IoT-CNN outputs feeding into DT simulations (e.g., example parameter adjustments based on detected defects) or clarify the claims to accurately reflect the current scope of the work while noting the framework's potential for predictive maintenance. revision: partial

Circularity Check

0 steps flagged

No derivation chain or self-referential reduction present; results are empirical CNN training outcomes.

full rationale

The paper describes an IoT-enhanced CNN framework for crack detection, reporting accuracy, precision, recall, and F1 scores on a dataset of 14,982 images after balancing and augmentation. No equations, first-principles derivations, or predictive models are claimed that reduce to fitted inputs by construction. Performance metrics are presented as direct empirical results from model training and evaluation, with no load-bearing self-citations, uniqueness theorems, or ansatz smuggling. The absence of a mathematical derivation chain means no circularity of the enumerated kinds can be identified. Potential concerns about train/test splits or overfitting are methodological rather than circularity issues.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on empirical CNN performance and standard supervised learning assumptions with many implicit hyperparameters and data-quality requirements; no independent verification or external benchmarks are mentioned.

free parameters (2)
  • CNN architecture and training hyperparameters
    Includes layers, learning rate, batch size, and optimization choices fitted to reach the reported accuracy on the 14,982-image dataset.
  • Dataset balancing and augmentation parameters
    Parameters controlling how the original data was balanced and augmented to produce the accuracy increase from 32% to 99%.
axioms (2)
  • domain assumption The LabelImg annotations correctly and comprehensively label cracks in the AM images
    Annotation quality is assumed without reported inter-annotator agreement or validation.
  • domain assumption The evaluation images are drawn from the same distribution as future real-world AM surfaces
    Required for the 99.54% accuracy to indicate practical generalization.

pith-pipeline@v0.9.0 · 5617 in / 1423 out tokens · 52520 ms · 2026-05-10T00:19:53.714047+00:00 · methodology

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

70 extracted references · 52 canonical work pages

  1. [1]

    Introduction Additive Manufacturing (AM), widely recognized as 3D printing, has revolutionized industrial production by enabling the fabrication of complex, lightweight, and customized structures with minimal material waste. Unlike traditional subtractive manufacturing , which involves cutting material away from a solid block, AM constructs components lay...

    2022

  2. [2]

    The system employs edge computing for low-latency defect detection and predictive analytics to enhance process reliability [18, 49]

    Materials and methods 2.1 Experimental Framework This study develops an IoT-integrated defect detection system powered by CNNs and DT technology for real-time monitoring and classification of defects in LPBF [45–48]. The system employs edge computing for low-latency defect detection and predictive analytics to enhance process reliability [18, 49]. The IoT...

  3. [3]

    A high precision score indicates that the model has a low false positive rate, ensuring that detected defects are truly present

    Precision Precision measures the proportion of correctly predicted defect instances among all instances predicted as defects. A high precision score indicates that the model has a low false positive rate, ensuring that detected defects are truly present. 𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 = 𝑇𝑃 𝑇𝑃 + 𝐹𝑃

  4. [4]

    A high recall value indicates that the model correctly identifies most defect instances while minimizing false negatives

    Recall (Sensitivity) Recall evaluates the model’s ability to detect all actual defects in the dataset. A high recall value indicates that the model correctly identifies most defect instances while minimizing false negatives. 𝑅𝑒𝑐𝑎𝑙𝑙 = 𝑇𝑃 𝑇𝑃 + 𝐹𝑁

  5. [5]

    It is especially useful in cases where there is an imbalance between defect and non-defect instances

    F1-Score The F1-score is the harmonic mean of precision and recall, providing a balanced evaluation of the model’s classification performance. It is especially useful in cases where there is an imbalance between defect and non-defect instances. 𝐹1 − 𝑆𝑐𝑜𝑟𝑒 = 2 × 𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 × 𝑅𝑒𝑐𝑎𝑙𝑙 𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛+ 𝑅𝑒𝑐𝑎𝑙𝑙 The F1-score ensures that both false positives and false n...

  6. [6]

    It provides a general measure of the model’s effectiveness but may be misleading if the dataset is highly imbalanced

    Accuracy Accuracy measures the proportion of correctly classified instances among all inspected samples. It provides a general measure of the model’s effectiveness but may be misleading if the dataset is highly imbalanced. 𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 = 𝑇𝑃 + 𝑇𝑁 𝑇𝑃 + 𝑇𝑁 + 𝐹𝑃 + 𝐹𝑁

  7. [7]

    This refers to the time taken for the CNN model to process an input image and classify it

    Inference Time (Latency) To evaluate the real-time efficiency of the IoT-enhanced system, inference time per frame (or latency) is measured. This refers to the time taken for the CNN model to process an input image and classify it. 𝐼𝑛𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑇𝑖𝑚𝑒 (𝑚𝑠) = 𝑇𝑜𝑡𝑎𝑙 𝑃𝑟𝑜𝑐𝑒𝑠𝑠𝑖𝑛𝑔 𝑇𝑖𝑚𝑒 (𝑚𝑠) 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐹𝑟𝑎𝑚𝑒 𝑃𝑟𝑜𝑐𝑒𝑠𝑠𝑒𝑑 Low inference time is critical for real-time defec...

  8. [8]

    Digital Twin Effectiveness The impact of the Digital Twin -based adaptive defect mitigation system is assessed by calculating the percentage of defect reduction after process adjustments. 𝐷𝑒𝑓𝑒𝑐𝑡 𝑅𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒(%) = 𝐷𝑒𝑓𝑒𝑐𝑡𝑠 𝐵𝑒𝑓𝑜𝑟𝑒 𝐴𝑑𝑗𝑢𝑠𝑡𝑚𝑒𝑛𝑡− 𝐷𝑒𝑓𝑒𝑐𝑡𝑠 𝐴𝑓𝑡𝑒𝑟 𝐴𝑑𝑗𝑢𝑠𝑡𝑚𝑒𝑛𝑡 𝐷𝑒𝑓𝑒𝑐𝑡𝑠 𝑏𝑒𝑓𝑜𝑟𝑒 𝐴𝑑𝑗𝑢𝑠𝑡𝑚𝑒𝑛𝑡 × 100 A high defect reduction rate indicates that the Digital Tw...

  9. [9]

    Results and Discussion The evaluation of surface crack detection performance in existing studies has typically relied on proprietary datasets, emphasizing the scarcity of publicly accessible datasets with detailed bounding box and segmentation annotations [47, 53 –60]. To address this gap, our study presents an IoT -enhanced deep learning approach for cra...

  10. [10]

    Additionally, multimodal data fusion comb ining thermal imaging, ultrasonic sensing, and optical microscopy will further enhance defect characterization

    Future Work and Industry 4.0 Implications Looking ahead, our framework will expand to incorporate advanced architectures such as R -CNN and U- Net, specifically optimized for detecting multi -scale defects like pinholes, spatters, keyholes, and graded cracks. Additionally, multimodal data fusion comb ining thermal imaging, ultrasonic sensing, and optical ...

  11. [11]

    Conclusion This study introduced an IoT-enhanced deep learning framework leveraging CNNs for high-accuracy crack detection in AM surfaces. By integrating real -time IoT monitoring, CNN-based defect classification, and DT simulations, our approach significantly improve s the accuracy, efficiency, and scalability of defect detection in metal -based AM proce...

  12. [12]

    (No Title)

    Wohlers TWA Wohlers report 2021 : 3D printing and additive manufacturing global state of the industry. (No Title)

    2021

  13. [14]

    Ilani MA, Banad YM (2024) Modeling Melt Pool Geometry in Metal Additive Manufacturing Using Goldak’s Semi-Ellipsoidal Heat Source: A Data-driven Computational Approach

    2024

  14. [15]

    Chinese Journal of Mechanical Engineering: Additive Manufacturing Frontiers 1:100055

    Wei Q, Xie Y , Teng Q, et al (2022) Crack Types, Mechanisms, and Suppression Methods during High-energy Beam Additive Manufacturing of Nickel-based Superalloys: A Review. Chinese Journal of Mechanical Engineering: Additive Manufacturing Frontiers 1:100055. https://doi.org/10.1016/J.CJMEAM.2022.100055

    work page doi:10.1016/j.cjmeam.2022.100055 2022

  15. [16]

    J Alloys Compd 864:158803

    Iveković A, Montero-Sistiaga ML, Vleugels J, et al (2021) Crack mitigation in Laser Powder Bed Fusion processed Hastelloy X using a combined numerical-experimental approach. J Alloys Compd 864:158803. https://doi.org/10.1016/J.JALLCOM.2021.158803

    work page doi:10.1016/j.jallcom.2021.158803 2021

  16. [17]

    Superalloys 2012 577–586

    Carter LN, Attallah MM, Reed RC (2012) Laser Powder Bed Fabrication of Nickel-Base Superalloys: Influence of Parameters; Characterisation, Quantification and Mitigation of Cracking. Superalloys 2012 577–586. https://doi.org/10.1002/9781118516430.CH64

    work page doi:10.1002/9781118516430.ch64 2012

  17. [18]

    Mater Des 89:770–784

    Cloots M, Uggowitzer PJ, Wegener K (2016) Investigations on the microstructure and crack formation of IN738LC samples processed by selective laser melting using Gaussian and doughnut profiles. Mater Des 89:770–784. https://doi.org/10.1016/J.MATDES.2015.10.027

    work page doi:10.1016/j.matdes.2015.10.027 2016

  18. [19]

    Opt Laser Technol 124:105984

    Han Q, Gu Y , Soe S, et al (2020) Effect of hot cracking on the mechanical properties of Hastelloy X superalloy fabricated by laser powder bed fusion additive manufacturing. Opt Laser Technol 124:105984. https://doi.org/10.1016/J.OPTLASTEC.2019.105984

    work page doi:10.1016/j.optlastec.2019.105984 2020

  19. [21]

    A., & Khoshnevisan, M

    Ilani MA, Khoshnevisan M (2021) Study of surfactant effects on intermolecular forces (IMF) in powder-mixed electrical discharge machining (EDM) of Ti-6Al-4V. International Journal of Advanced Manufacturing Technology 116:1763–1782. https://doi.org/10.1007/S00170-021- 07569-3/FIGURES/30

    work page doi:10.1007/s00170-021- 2021

  20. [22]

    International Journal of Advanced Manufacturing Technology 120:5117–5129

    Ilani MA, Khoshnevisan M (2022) An evaluation of the surface integrity and corrosion behavior of Ti-6Al-4 V processed thermodynamically by PM-EDM criteria. International Journal of Advanced Manufacturing Technology 120:5117–5129. https://doi.org/10.1007/S00170-022-09093- 4/FIGURES/18

    work page doi:10.1007/s00170-022-09093- 2022

  21. [23]

    Materials 2018, Vol 11, Page 106 11:106

    Marchese G, Basile G, Bassini E, et al (2018) Study of the Microstructure and Cracking Mechanisms of Hastelloy X Produced by Laser Powder Bed Fusion. Materials 2018, Vol 11, Page 106 11:106. https://doi.org/10.3390/MA11010106

    work page doi:10.3390/ma11010106 2018

  22. [24]

    Materials Science and Engineering: A 732:228–239

    Han Q, Mertens R, Montero-Sistiaga ML, et al (2018) Laser powder bed fusion of Hastelloy X: Effects of hot isostatic pressing and the hot cracking mechanism. Materials Science and Engineering: A 732:228–239. https://doi.org/10.1016/J.MSEA.2018.07.008

    work page doi:10.1016/j.msea.2018.07.008 2018

  23. [25]

    2024 Conference on AI, Science, Engineering, and Technology (AIxSET) 119–122

    Ilani MA, Banad YM (2024) XGBoost Algorithm for Interpretable AI Prediction of Melt Pool Geometry in IoT-Enabled Additive Manufacturing within Industry 4.0 Utilizing NSGA-II. 2024 Conference on AI, Science, Engineering, and Technology (AIxSET) 222–227. https://doi.org/10.1109/AIXSET62544.2024.00043

    work page doi:10.1109/aixset62544.2024.00043 2024

  24. [26]

    2024 Conference on AI, Science, Engineering, and Technology (AIxSET) 119–122

    Ilani MA, Banad YM (2024) Brain Tumor Detection Using U-Net Architectures in Industry 4.0 IoT Healthcare industries. 2024 Conference on AI, Science, Engineering, and Technology (AIxSET) 119–122. https://doi.org/10.1109/AIXSET62544.2024.00022

    work page doi:10.1109/aixset62544.2024.00022 2024

  25. [27]

    Eng Fract Mech 268:108467

    Joshi D, Singh TP , Sharma G (2022) Automatic surface crack detection using segmentation-based deep-learning approach. Eng Fract Mech 268:108467. https://doi.org/10.1016/J.ENGFRACMECH.2022.108467

    work page doi:10.1016/j.engfracmech.2022.108467 2022

  26. [28]

    CIRP Annals 65:417–420

    Weimer D, Scholz-Reiter B, Shpitalni M (2016) Design of deep convolutional neural network architectures for automated feature extraction in industrial inspection. CIRP Annals 65:417–420. https://doi.org/10.1016/J.CIRP .2016.04.072

    work page doi:10.1016/j.cirp 2016

  27. [29]

    Rapid Prototyp J 25:530–540

    Zhang B, Jaiswal P , Rai R, et al (2019) Convolutional neural network-based inspection of metal additive manufacturing parts. Rapid Prototyp J 25:530–540. https://doi.org/10.1108/RPJ-04- 2018-0096/FULL/XML

    work page doi:10.1108/rpj-04- 2019

  28. [30]

    J Intell Manuf 31:2003–2017

    Li X, Jia X, Yang Q, Lee J (2020) Quality analysis in metal additive manufacturing with deep learning. J Intell Manuf 31:2003–2017. https://doi.org/10.1007/S10845-020-01549-2/METRICS

    work page doi:10.1007/s10845-020-01549-2/metrics 2020

  29. [31]

    IEEE Trans Industr Inform 18:3820–3830

    Dang H, Tatipamula M, Nguyen HX (2022) Cloud-Based Digital Twinning for Structural Health Monitoring Using Deep Learning. IEEE Trans Industr Inform 18:3820–3830. https://doi.org/10.1109/TII.2021.3115119

    work page doi:10.1109/tii.2021.3115119 2022

  30. [32]

    Nature Communications 2023 14:1 14:1–11

    Gao S, Li Z, Van Petegem S, et al (2023) Additive manufacturing of alloys with programmable microstructure and properties. Nature Communications 2023 14:1 14:1–11. https://doi.org/10.1038/s41467-023-42326-y

    work page doi:10.1038/s41467-023-42326-y 2023

  31. [33]

    IISE Trans 51:437–455

    Khanzadeh M, Chowdhury S, Tschopp MA, et al (2019) In-situ monitoring of melt pool images for porosity prediction in directed energy deposition processes. IISE Trans 51:437–455. https://doi.org/10.1080/24725854.2017.1417656

    work page doi:10.1080/24725854.2017.1417656 2019

  32. [34]

    Expert Syst Appl 255:124678

    Wu Y , Sicard B, Gadsden SA (2024) Physics-informed machine learning: A comprehensive review on applications in anomaly detection and condition monitoring. Expert Syst Appl 255:124678. https://doi.org/10.1016/J.ESWA.2024.124678

    work page doi:10.1016/j.eswa.2024.124678 2024

  33. [35]

    Addit Manuf 24:273–286

    Scime L, Beuth J (2018) A multi-scale convolutional neural network for autonomous anomaly detection and classification in a laser powder bed fusion additive manufacturing process. Addit Manuf 24:273–286. https://doi.org/10.1016/J.ADDMA.2018.09.034

    work page doi:10.1016/j.addma.2018.09.034 2018

  34. [36]

    Progress in Additive Manufacturing 2024 10:1 10:171–185

    Farrag A, Yang Y , Cao N, et al (2024) Physics-Informed Machine Learning for metal additive manufacturing. Progress in Additive Manufacturing 2024 10:1 10:171–185. https://doi.org/10.1007/S40964-024-00612-1

    work page doi:10.1007/s40964-024-00612-1 2024

  35. [37]

    Comput Mater Sci 207:111262

    Qin Y , Bao Y , DeWitt S, et al (2022) Dendrite-resolved, full-melt-pool phase-field simulations to reveal non-steady-state effects and to test an approximate model. Comput Mater Sci 207:111262. https://doi.org/10.1016/J.COMMATSCI.2022.111262

    work page doi:10.1016/j.commatsci.2022.111262 2022

  36. [38]

    Addit Manuf 58:103007

    Pandiyan V, Masinelli G, Claire N, et al (2022) Deep learning-based monitoring of laser powder bed fusion process on variable time-scales using heterogeneous sensing and operando X-ray radiography guidance. Addit Manuf 58:103007. https://doi.org/10.1016/J.ADDMA.2022.103007

    work page doi:10.1016/j.addma.2022.103007 2022

  37. [39]

    J Manuf Process 133:524–555

    Faegh M, Ghungrad S, Oliveira JP , et al (2025) A review on physics-informed machine learning for process-structure-property modeling in additive manufacturing. J Manuf Process 133:524–555. https://doi.org/10.1016/J.JMAPRO.2024.11.066

    work page doi:10.1016/j.jmapro.2024.11.066 2025

  38. [40]

    Krull, P

    Krull A, Hirsch P , Rother C, et al (2020) Artificial-intelligence-driven scanning probe microscopy. Communications Physics 2020 3:1 3:1–8. https://doi.org/10.1038/s42005-020-0317-3

    work page doi:10.1038/s42005-020-0317-3 2020

  39. [41]

    ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering 8:

    Mahadevan S, Nath P , Hu Z (2022) Uncertainty Quantification for Additive Manufacturing Process Improvement: Recent Advances. ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering 8:. https://doi.org/10.1115/1.4053184/1129174

    work page doi:10.1115/1.4053184/1129174 2022

  40. [42]

    IEEE Robot Autom Lett 6:6032–6038

    Chen R, Lu Y , Witherell P , et al (2021) Ontology-Driven Learning of Bayesian Network for Causal Inference and Quality Assurance in Additive Manufacturing. IEEE Robot Autom Lett 6:6032–6038. https://doi.org/10.1109/LRA.2021.3090020

    work page doi:10.1109/lra.2021.3090020 2021

  41. [43]

    https://doi.org/10.32657/10356/180195

    (2024) Multi-sensor monitoring for in-situ defect detection and quality assurance in laser-directed energy deposition. https://doi.org/10.32657/10356/180195

    work page doi:10.32657/10356/180195 2024

  42. [44]

    Kaufmann, D

    Wang Z, Jiang C, Liu P , et al (2020) Uncertainty quantification and reduction in metal additive manufacturing. npj Computational Materials 2020 6:1 6:1–10. https://doi.org/10.1038/s41524- 020-00444-x

    work page doi:10.1038/s41524- 2020

  43. [46]

    Compos Struct 348:118514

    Li Y , Wan D, Wang Z, Hu D (2024) Physics-constrained deep learning approach for solving inverse problems in composite laminated plates. Compos Struct 348:118514. https://doi.org/10.1016/J.COMPSTRUCT.2024.118514

    work page doi:10.1016/j.compstruct.2024.118514 2024

  44. [47]

    Sensors (Basel) 22:494

    McGowan E, Gawade V, Guo W (2022) A Physics-Informed Convolutional Neural Network with Custom Loss Functions for Porosity Prediction in Laser Metal Deposition. Sensors (Basel) 22:494. https://doi.org/10.3390/S22020494

    work page doi:10.3390/s22020494 2022

  45. [48]

    Ahsan MM, Liu Y , Raman S, Siddique Z (2024) Digital Twins in Additive Manufacturing: A Systematic Review

    2024

  46. [49]

    Air force vehicles

    Glaessgen EH, Stargel DS (2012) The digital twin paradigm for future NASA and U.S. Air force vehicles. Collection of Technical Papers - AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference. https://doi.org/10.2514/6.2012-1818

    work page doi:10.2514/6.2012-1818 2012

  47. [50]

    A review of the roles of digital twin in CPS-based production systems

    Negri E, Fumagalli L, Macchi M (2017) A Review of the Roles of Digital Twin in CPS-based Production Systems. Procedia Manuf 11:939–948. https://doi.org/10.1016/J.PROMFG.2017.07.198

    work page doi:10.1016/j.promfg.2017.07.198 2017

  48. [51]

    IFAC-PapersOnLine 51(11), 1016–1022 (2018) https://doi.org/10.1016/j.ifacol.2018.08.474

    Kritzinger W, Karner M, Traar G, et al (2018) Digital Twin in manufacturing: A categorical literature review and classification. IFAC-PapersOnLine 51:1016–1022. https://doi.org/10.1016/J.IFACOL.2018.08.474

    work page doi:10.1016/j.ifacol.2018.08.474 2018

  49. [52]

    Industry 5.0: Prospect and retrospect

    Osho J, Hyre A, Pantelidakis M, et al (2022) Four Rs Framework for the development of a digital twin: The implementation of Representation with a FDM manufacturing machine. J Manuf Syst 63:370–380. https://doi.org/10.1016/J.JMSY .2022.04.014

    work page doi:10.1016/j.jmsy 2022

  50. [53]

    Procedia CIRP 118:717–722

    Oettl F, Hörbrand S, Wittmeir T, Schilp J (2023) Method for evaluating the monetary added value of the usage of a digital twin for additive manufacturing. Procedia CIRP 118:717–722. https://doi.org/10.1016/J.PROCIR.2023.06.123

    work page doi:10.1016/j.procir.2023.06.123 2023

  51. [54]

    Addit Manuf 48:102388

    Yi L, Glatt M, Ehmsen S, et al (2021) Process monitoring of economic and environmental performance of a material extrusion printer using an augmented reality-based digital twin. Addit Manuf 48:102388. https://doi.org/10.1016/J.ADDMA.2021.102388

    work page doi:10.1016/j.addma.2021.102388 2021

  52. [55]

    The South African Journal of Industrial Engineering 32:37–43

    Anderson AM, Van Der Merwe A (2021) TIME-DRIVEN ACTIVITY-BASED COSTING RELATED TO DIGITAL TWINNING IN ADDITIVE MANUFACTURING. The South African Journal of Industrial Engineering 32:37–43. https://doi.org/10.7166/32-1-2271

    work page doi:10.7166/32-1-2271 2021

  53. [56]

    Int J Prod Res 59:4811–

    Wang Z, Liu Q, Chen H, Chu X (2021) A deformable CNN-DLSTM based transfer learning method for fault diagnosis of rolling bearing under multiple working conditions. Int J Prod Res 59:4811–

    2021

  54. [57]

    https://doi.org/10.1080/00207543.2020.1808261

    work page doi:10.1080/00207543.2020.1808261 2020

  55. [58]

    Int J Comput Integr Manuf 34:1177–

    Li X, Li M, Wu Y , et al (2021) Accurate screw detection method based on faster R-CNN and rotation edge similarity for automatic screw disassembly. Int J Comput Integr Manuf 34:1177–

    2021

  56. [59]

    https://doi.org/10.1080/0951192X.2021.1963476

    work page doi:10.1080/0951192x.2021.1963476 2021

  57. [60]

    Sensors 2011, Vol 11, Pages 9628-9657 11:9628–9657

    Gavilán M, Balcones D, Marcos O, et al (2011) Adaptive Road Crack Detection System by Pavement Classification. Sensors 2011, Vol 11, Pages 9628-9657 11:9628–9657. https://doi.org/10.3390/S111009628

    work page doi:10.3390/s111009628 2011

  58. [61]

    J Healthc Eng 2022:

    Hosseini Rad R, Baniasadi S, Yousefi P , et al (2022) Presented a Framework of Computational Modeling to Identify the Patient Admission Scheduling Problem in the Healthcare System. J Healthc Eng 2022:. https://doi.org/10.1155/2022/1938719

    work page doi:10.1155/2022/1938719 2022

  59. [62]

    2019 19th International Conference on Advanced Robotics, ICAR 2019 420–425

    Lemos CB, Farias PCMA, Filho EFS, Conceicao AGS (2019) Convolutional neural network based object detection for additive manufacturing. 2019 19th International Conference on Advanced Robotics, ICAR 2019 420–425. https://doi.org/10.1109/ICAR46387.2019.8981618

    work page doi:10.1109/icar46387.2019.8981618 2019

  60. [63]

    Results in Engineering 22:102003

    Awan MR, Chan CW, Murphy A, et al (2024) Deep Learning and Image data-based surface cracks recognition of laser nitrided Titanium alloy. Results in Engineering 22:102003. https://doi.org/10.1016/J.RINENG.2024.102003

    work page doi:10.1016/j.rineng.2024.102003 2024

  61. [64]

    Eng Fail Anal 119:104908

    Alqahtani H, Bharadwaj S, Ray A (2021) Classification of fatigue crack damage in polycrystalline alloy structures using convolutional neural networks. Eng Fail Anal 119:104908. https://doi.org/10.1016/J.ENGFAILANAL.2020.104908

    work page doi:10.1016/j.engfailanal.2020.104908 2021

  62. [65]

    Materials 2018, Vol 11, Page 2467 11:2467

    Konovalenko I, Maruschak P , Prentkovskis O, Junevičius R (2018) Investigation of the Rupture Surface of the Titanium Alloy Using Convolutional Neural Networks. Materials 2018, Vol 11, Page 2467 11:2467. https://doi.org/10.3390/MA11122467

    work page doi:10.3390/ma11122467 2018

  63. [66]

    Pandian AP , Palanisamy R, Ntalianis K (2021) Proceedings of International Conference on Intelligent Computing, Information and Control Systems. 1272:. https://doi.org/10.1007/978-981- 15-8443-5

    work page doi:10.1007/978-981- 2021

  64. [67]

    Pattern Recognit Lett 33:227–238

    Zou Q, Cao Y , Li Q, et al (2012) CrackTree: Automatic crack detection from pavement images. Pattern Recognit Lett 33:227–238. https://doi.org/10.1016/J.PATREC.2011.11.004

    work page doi:10.1016/j.patrec.2011.11.004 2012

  65. [68]

    https://doi.org/101177/1475921719883202 19:1440–1452

    Mohtasham Khani M, Vahidnia S, Ghasemzadeh L, et al (2019) Deep-learning-based crack detection with applications for the structural health monitoring of gas turbines. https://doi.org/101177/1475921719883202 19:1440–1452. https://doi.org/10.1177/1475921719883202

    work page doi:10.1177/1475921719883202 2019

  66. [69]

    M., & Srivastava, A

    Banadaki, Y . M., & Srivastava, A. (2013, August). A novel graphene nanoribbon field effect transistor for integrated circuit design. In 2013 IEEE 56th International Midwest Symposium on Circuits and Systems (MWSCAS) (pp. 924-927). IEEE. 10.1109/MWSCAS.2013.6674801

    work page doi:10.1109/mwscas.2013.6674801 2013

  67. [70]

    Banadaki, Y . M. M. (2016). Physical modeling of graphene nanoribbon field effect transistor using non-equilibrium green function approach for integrated circuit design. Louisiana State University and Agricultural & Mechanical College

    2016

  68. [71]

    Banadaki, Y ., & Sharifi, S. (2019). Graphene nanostructures: modeling, simulation, and applications in electronics and photonics. Jenny Stanford Publishing

    2019

  69. [72]

    Banadaki, Y ., Razaviarab, N., Fekrmandi, H., Li, G., Mensah, P ., Bai, S., & Sharifi, S. (2021). Automated quality and process control for additive manufacturing using deep convolutional neural networks. Recent Progress in Materials, 4(1). https://doi.org/10.21926/rpm.2201005

    work page doi:10.21926/rpm.2201005 2021

  70. [73]

    Banadaki, Y . (2019). On the use of machine learning for additive manufacturing technology in industry 4.0. J. Comput. Sci. Inf. Technol, 7, 61-68. https://doi.org/10.15640/jcsit.v7n2a7

    work page doi:10.15640/jcsit.v7n2a7 2019