XAI-PdMNet-Bench: an explainable generative AI framework for leakage-safe predictive maintenance in industry 5.0 manufacturing
The International Journal of Advanced Manufacturing Technology · 2026-08 · 145 (9-10) · 6837-6862
DOI: 10.1007/s00170-026-18674-6 ↗Kurz gefasst
Fachartikel in der Zeitschrift The International Journal of Advanced Manufacturing Technology, online erschienen am 14.07.2026. Verfasst von Abdullah Almasad und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 50 Literaturangaben. Gefördert von: King Saud University.
Abstract (Originalsprache)
As manufacturing transitions toward Industry 5.0, the demand for trustworthy, operator-centric predictive maintenance has intensified; yet today’s CNC milling models often fall short on the very qualities that make them deployable: leakage-free evaluation, calibrated performance under imbalance, and explainability. This paper introduces XAI-PdMNet-Bench, a leakage-safe and explainable PdM framework validated on two CNC milling benchmarks: AI4I 2020 (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n{=}9{,}973$$\end{document}; 3.31 % failures) and PHM 2010 (real cutting signals; tool-independent split). A 25-feature physics-informed matrix is constructed after removing all leaked indicator columns, and minority-class augmentation is performed via statistically audited CTGAN (Frobenius \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= 0.083$$\end{document}; all KS tests passed). The proposed XGBoost achieves F1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= 0.991$$\end{document} and PR-AUC \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= 0.994$$\end{document} under leakage-safe cross-validation, exceeding the best leakage-unsafe baseline (F1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= 0.983$$\end{document}). TabTransformer achieves F1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= 0.987$$\end{document}, the best deep-learning result. On PHM 2010, the proposed RCNN (Conv1D + BiLSTM) attains F1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= 0.921$$\end{document} and PR-AUC \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= 0.941$$\end{document}; staged transfer learning from AI4I 2020 yields F1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= 0.894$$\end{document}. SHAP attribution identifies RPM\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times $$\end{document}wear as the dominant fault driver, and a structured operator report supports operator-facing maintenance interpretation. Ablation studies confirm CTGAN contributes the largest gain under leakage-safe balanced cross-validation (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta $$\end{document}F1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= +0.160$$\end{document} over no augmentation; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta $$\end{document}F1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= +0.140$$\end{document} on the imbalanced held-out set), followed by physics-informed feature engineering (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta $$\end{document}F1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$= +0.050$$\end{document}).
Förderung
- King Saud University
Zitieren
APA
Almasad, A., Rehman, T., Mustafa, M., Azim, S., & Alkahtani, M. S. (2026). XAI-PdMNet-Bench: an explainable generative AI framework for leakage-safe predictive maintenance in industry 5.0 manufacturing. The International Journal of Advanced Manufacturing Technology, 145(9-10), 6837-6862. https://doi.org/10.1007/s00170-026-18674-6
Vancouver
Almasad A, Rehman T, Mustafa M, Azim S, Alkahtani MS. XAI-PdMNet-Bench: an explainable generative AI framework for leakage-safe predictive maintenance in industry 5.0 manufacturing. The International Journal of Advanced Manufacturing Technology. 2026;145(9-10):6837-6862. doi:10.1007/s00170-026-18674-6
BibTeX
@article{Almasad2026,
author = {Almasad, Abdullah and Rehman, Tayyab and Mustafa, Muzzamil and Azim, Shakir and Alkahtani, Mohammed Saad},
title = {XAI-PdMNet-Bench: an explainable generative AI framework for leakage-safe predictive maintenance in industry 5.0 manufacturing},
journal = {The International Journal of Advanced Manufacturing Technology},
year = {2026},
volume = {145},
number = {9-10},
pages = {6837--6862},
doi = {10.1007/s00170-026-18674-6}
}
EUVIMED