Abstract
Industry 5.0 asks more of manufacturing systems than Industry 4.0 did: beyond automation and efficiency, they are expected to be resilient, sustainable, and human-centric. Meeting these expectations requires advances across the system life cycle, from physics-based process modeling and intelligent inspection to health monitoring, predictive maintenance, and robust operational optimization. This Editorial introduces the Special Issue “Modeling, Design, Optimization and Maintenance of Intelligent Manufacturing Towards Industry 5.0”, which brings together nine contributions: eight research articles and one review covering molten-jet breakup modeling, automated dye penetrant testing, sensor fault detection in lithium-ion batteries, bearing diagnosis under imbalanced data, remaining useful life prediction, hidden-parameter estimation, distribution-network optimization with renewable integration, and synchronized microgrid control. The methods range from linear-stability analysis and time-frequency transforms to generative learning, attention networks, Bayesian regression, and robust optimization. What links them is a concern with reliability under imperfect information; it is from this angle that the collection speaks most directly to the goals of Industry 5.0, while making clear that cross-domain generalization and the integration of human and sustainability objectives require further work.
Manufacturing is moving beyond the automation-centered logic of Industry 4.0 toward a broader Industry 5.0 vision, in which advanced technologies are judged not only by productivity but also by their contribution to resilience, sustainability, and human well-being [1,2,3]. Artificial intelligence, cyber-physical systems, digital twins, and industrial sensing remain essential enablers. Their value, however, depends on whether they can be translated into reliable models, trustworthy monitoring information, and robust operating decisions under real industrial constraints.
This translation is difficult in practice. Physical mechanisms are often incompletely known, sensor measurements are noisy or incomplete, and the variables that best describe degradation may be inaccessible during operation. Decisions, meanwhile, must accommodate changing loads, renewable-energy variability, and limited computational resources. Intelligent manufacturing therefore needs a life-cycle perspective: process design, equipment health, quality inspection, control, energy use, and maintenance cannot be treated as isolated activities if Industry 5.0 objectives are to be realized in practice.
Motivated by these challenges, the Special Issue “Modeling, Design, Optimization and Maintenance of Intelligent Manufacturing Towards Industry 5.0” provides a forum for recent theoretical and application-oriented developments in smart manufacturing, industrial fault diagnosis and prognosis, predictive maintenance, reliability assessment, artificial-intelligence-driven optimization, digital twins, industrial Internet of Things, and data-driven decision-making. Following peer review, nine papers were published: eight research articles and one review. We group them below into three connected themes: process modeling, inspection, and fault detection [4,5,6,7]; condition monitoring and predictive maintenance [8,9,10]; and robust control and optimization of energy systems [11,12].
The first theme concerns physically interpretable modeling of manufacturing processes. Sklabinskyi et al. [4] develop a theoretical framework for predicting molten-jet breakup in a rotating granulation system without forced excitation. By adapting the Rayleigh-Plateau and Tomotika linear-stability descriptions to a viscous melt and expressing the regime through the Ohnesorge, Weber, Reynolds, and Capillary numbers, the study derives the dominant wavelength, perturbation growth rate, breakup time, and characteristic droplet diameter. Evaluation for molten urea shows close agreement among classical, viscosity-corrected, and empirical predictions. The result is an engineering baseline for granule-size prediction in unforced rotating granulation, grounded in the instability physics of the process itself.
Quality inspection is addressed by Haupts et al. [5], who compare established and machine-learning-assisted binarization methods for automated dye penetrant testing of stainless-steel pipes. In addition to global, adaptive, and histogram-based thresholding, the authors evaluate Soft Binarization, Delta Binarization, and Convolutional Autoencoder Binarization using pixel-level and region-level indicators. Particularly instructive is the analysis of defect saturation: different methods exhibit distinct trade-offs between detecting low-saturation defects and suppressing false positives. The results emphasize that an inspection algorithm should be selected according to the risk tolerance and quality requirements of the production process, rather than by a single aggregate metric. This viewpoint is especially relevant to human-centric decision support, where automated outputs must remain aligned with inspection priorities and operational consequences.
For lithium-ion battery management systems, Wang et al. [6] study the rapid detection of multiple current- and voltage-sensor faults using time-frequency analysis. Six sensor-fault forms are modeled mathematically and standardized so that different detection techniques can be compared on equal footing. The proposed multi-window short-time Fourier transform overcomes the time- and frequency-resolution limits of the conventional single-window formulation and detects all fault types considered in the experiments. Because the method dispenses with a detailed process model, the very thing that makes most model-based battery-fault schemes difficult to apply in engineering, it is comparatively easy to deploy.
Data imbalance, which is common because industrial systems generate many normal samples but relatively few fault examples, is investigated by Tu et al. [7]. Their attention-enhanced conditional Wasserstein generative adversarial network incorporates attention into both the generator and discriminator, while Wasserstein distance improves training stability and reduces mode collapse. The generated bearing-fault samples are combined with wavelet decomposition and a residual network for diagnosis. Experiments on the Case Western Reserve University bearing dataset demonstrate high-fidelity sample generation and strong classification performance, and ablation studies confirm that the attention, Wasserstein, and generative components each contribute to the result. How well the approach transfers across machines and operating conditions is a question the authors leave open.
The second theme focuses on prognostics and health management (PHM). Wang et al. [8] pursue the problem with a multi-channel hybrid-attention temporal fusion channel network: a dual-channel mechanism combining self-attention and channel attention captures relationships across sensors and time, while causal dilated convolutions and enhanced residual structures model long-term degradation dependencies. On the PHM 2012 dataset, the method achieves an average root-mean-square error of 0.091 and outperforms representative alternatives. The authors are candid about this accuracy coming at a computational price and still requiring validation across bearing types and operating conditions.
Where Wang et al. [8] opt for a deep architecture, Wang et al. [9] stay within a Bayesian regression framework and make it more robust. They introduce a Student’s t-distribution into the relevance vector machine to reduce sensitivity to outliers and embed Q-learning into the feature-processing stage. The resulting Q-learning-incorporated robust relevance vector machine predicts degradation trends and supports rolling remaining-life estimation for capacitors in a high-speed-train traction system, improving substantially on conventional relevance vector machine variants. The pairing of probabilistic modeling with reinforcement learning is instructive, although the nested optimization involved would need to be simplified before deployment on constrained hardware.
Han et al. [10] address a complementary health-monitoring problem: important degradation parameters may be hidden because critical sensors cannot be installed or accessed during operation. The authors formulate hidden-parameter estimation as an observation-trajectory alignment problem and propose a Lyapunov Actor-Critic Tutor framework. The Lyapunov-based formulation improves estimation stability, and a tutor component guides learning across multiple resolutions. A control-moment-gyroscope case study estimates the motor torque coefficient under six degradation modes and demonstrates lower estimation error than comparison models. The framework can serve as a hidden-state observer feeding downstream fault detection and remaining-life prediction, though the accuracy is paid for in latency and computational cost.
The third theme extends intelligent decision-making from individual assets to energy and power systems. Guo et al. [11] consider the siting and operation of static var generators in renewable-integrated stochastic distribution networks. They develop a multi-objective voltage-loss sensitivity model and use information gap decision theory to quantify uncertainty in multiple renewable-energy sources. In the IEEE 33-bus case study, the proposed strategy reduces voltage fluctuation and network loss by 29.53% and 7.75%, respectively, relative to the non-intervention case. The approach provides decision-makers with operating strategies for different renewable-to-load growth scenarios and demonstrates how robust optimization can strengthen system resilience under uncertain energy supply.
Li et al. [12] provide the collection’s review contribution, surveying master–slave and peer-to-peer microgrid control, converter droop strategies, and synchronized fixed-frequency control based on Global Positioning System timing. The review explains how synchronized droop control can support frequency coordination among distributed inverters without direct inter-inverter communication, thereby reducing hardware requirements. It also identifies unresolved issues in energy-storage sizing, backup timing when satellite signals are unavailable, AC/DC hybrid microgrids, real-time compensation under load disturbances, and black-start support. These directions connect control-system design with the resilience expected of future distributed manufacturing and energy infrastructures.
Across these nine papers, three patterns stand out. Physics-based and data-driven methods prove complementary rather than competing: linear-stability theory supplies interpretability and regime boundaries, while learning-based approaches extract information from high-dimensional, partially observable systems. Reliability under imperfect information emerges as a shared requirement, since imbalanced fault samples, outliers, uncertain renewable generation, and inaccessible sensors all push toward methods that are robust by design. And read in sequence, the papers trace a continuous engineering chain—process modeling informs design, sensing and inspection establish operational awareness, diagnosis and prognosis characterize risk, and optimization converts that knowledge into control or maintenance decisions.
The collection most directly advances the resilience dimension of Industry 5.0, through better monitoring, prediction, and uncertainty-aware decision-making. The next step is to make the other two dimensions equally explicit. Human-centricity calls for interfaces and confidence estimates that let domain experts supervise and, when necessary, correct intelligent systems. Sustainability calls for life-cycle measures that connect predictive accuracy and equipment availability to energy use, material consumption, and emissions. Progress on all three fronts will also depend on multi-site industrial validation, standardized benchmarks, and models light enough for edge and real-time deployment.
We hope that this Special Issue encourages further collaboration among researchers and practitioners in the fields of process engineering, intelligent control, industrial artificial intelligence, reliability, energy systems, and maintenance. If these nine studies show one thing, it is that first-principles knowledge and data-driven learning are strongest when they are made to work together. A second edition of this Special Issue is now open for submissions, and we look forward to work that carries these ideas further.
Author Contributions
Conceptualization, Y.L. and C.W.; validation, Y.L., C.W., Y.S. and L.J.; formal analysis, C.W.; investigation, Y.S.; resources, L.J.; data curation, Y.L. and C.W.; writing—original draft preparation, Y.L.; writing—review and editing, C.W., Y.S. and L.J.; visualization, C.W.; supervision, L.J.; project administration, Y.L. All authors have read and agreed to the published version of the manuscript.
Acknowledgments
The Guest Editors sincerely thank all authors who contributed their work to this Special Issue and the reviewers who devoted their time and expertise to the peer-review process. We are grateful to the Editor-in-Chief of Processes, Giancarlo Cravotto, for the opportunity to organize this Special Issue and for his continuous support. We also thank the entire editorial team of Processes for their support during preparation and publication.
Conflicts of Interest
The authors declare no conflict of interest.
References
- Xu, X.; Lu, Y.; Vogel-Heuser, B.; Wang, L. Industry 4.0 and Industry 5.0-Inception, Conception and Perception. J. Manuf. Syst. 2021, 61, 530–535. [Google Scholar] [CrossRef] [Scilit]
- Ivanov, D. The Industry 5.0 framework: Viability-Based Integration of the Resilience, Sustainability, and Human-centricity Perspectives. Int. J. Prod. Res. 2023, 61, 1683–1695. [Google Scholar] [CrossRef] [Scilit]
- Xiang, W.; Yu, K.; Han, F.; Fang, L.; He, D.; Han, Q.-L. Advanced Manufacturing in Industry 5.0: A Survey of Key Enabling Technologies and Future Trends. IEEE Trans. Ind. Inform. 2024, 20, 1055–1068. [Google Scholar] [CrossRef] [Scilit]
- Sklabinskyi, V.; Liaposhchenko, O.; Ostroha, R.; Zabitsky, D.; Myshchenko, D.; Kozii, I.; Bocko, J. Theoretical Analysis of Molten Jet Breakup in a Rotating Granulation System Under Unforced Conditions. Processes 2026, 14, 1077. [Google Scholar] [CrossRef] [Scilit]
- Haupts, P.J.; Al-Joumaa, H.; Al-Shrouf, L.; Jelali, M. Comparative Analysis of Binarization Approaches for Automated Dye Penetrant Testing. Processes 2025, 13, 1212. [Google Scholar] [CrossRef] [Scilit]
- Wang, Q.; Chen, H.; Tian, E. Comparative Study of Multiple-Sensor-Fault-Detection Based Time-Frequency Analysis Methods on Lithium-Ion Batteries. Processes 2025, 13, 929. [Google Scholar] [CrossRef] [Scilit]
- Tu, H.; Zhang, Y.; Wang, X.; Li, Y. Attention-Enhanced Conditional Wasserstein GAN with Wavelet-ResNet for Fault Diagnosis Under Imbalanced Data. Processes 2025, 13, 3531. [Google Scholar] [CrossRef] [Scilit]
- Wang, C.; Jiang, J.; Qi, H.; Zhang, D.; Han, X. A Novel Temporal Fusion Channel Network with Multi-Channel Hybrid Attention for the Remaining Useful Life Prediction of Rolling Bearings. Processes 2024, 12, 2762. [Google Scholar] [CrossRef] [Scilit]
- Wang, X.; Li, Z.; Wang, X.; Hu, X. Q-Learning-Incorporated Robust Relevance Vector Machine for Remaining Useful Life Prediction. Processes 2024, 12, 2536. [Google Scholar] [CrossRef] [Scilit]
- Han, D.; Yang, J.; Tang, D.; Yu, J. LAC-T: A Tutored Reinforcement Learning Framework of Hidden-Parameter Estimation for Health Monitoring of Rotating Systems. Processes 2026, 14, 1902. [Google Scholar] [CrossRef] [Scilit]
- Guo, Y.; Fu, Y.; Li, J.; Chen, J. Optimizing Reactive Compensation for Enhanced Voltage Stability in Renewable-Integrated Stochastic Distribution Networks. Processes 2025, 13, 303. [Google Scholar] [CrossRef] [Scilit]
- Li, K.; Li, Y.; Fan, R.; Liu, Z.; Lu, M. A Review of Synchronous Fixed-Frequency Microgrid Droop Control Systems Based on Global Positioning System. Processes 2025, 13, 54. [Google Scholar] [CrossRef] [Scilit]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license.