Industry 5.0 emphasizes human-centric manufacturing and creates demand for maintenance systems that move beyond passive failure prediction toward economically grounded prescriptive decisions. Prescriptive maintenance strategies aim not only to anticipate failures but also to optimize intervention policies under practical economic and operational constraints. While deep reinforcement learning has shown promise in automating such decisions, existing approaches rely primarily on numerical data and cannot process the qualitative diagnostic information stored in maintenance logs. By missing these crucial historical fault patterns, they operate as opaque systems that are difficult for practitioners to interpret and trust. This paper addresses these gaps by proposing a large language model (LLM) enhanced hierarchical multi-agent reinforcement learning framework for prescriptive maintenance with three main contributions. First, it adopts a heterogeneous architecture where a policy-gradient Manager handles station-level maintenance prioritization and a value-based Worker selects from a predefined taxonomy of prescriptive maintenance actions, reflecting how real maintenance teams are organized. Second, locally deployed LLMs translate textual maintenance context, including short-form maintenance reports and technician notes, into diagnostic semantic state variables with no dependence on cloud services. Third, the framework introduces a six-component cost-model reward based on production economics, covering tool, material, equipment, downtime, maintenance, and handling costs, so that automated decisions can be understood in straightforward economic terms and the effect of reward design on maintenance behavior can be examined explicitly. The framework is evaluated on a calibrated digital shadow of a lab-scale drone assembly line. Semantic state augmentation is associated with a 25.1% higher converged reward and 42% lower learning variance compared to the numerical-only baseline for the strongest LLM configuration. Under joint noise at σ = 0.25, the LLM-augmented agent still retains a performance advantage over the baseline, indicating that text-derived semantic variables provide a relatively stable auxiliary signal when numerical observations become unreliable. The cost-model formulation also reduces total operational cost compared to the overall equipment effectiveness based alternative while making the resulting policies easier to interpret in economic terms. These results show that integrating human diagnostic knowledge, reflecting real maintenance workflows, and improving economic transparency can build the necessary trust to support effective human-AI teaming in industrial settings.

PurposeEffective knowledge transfer in multi-site organizations using a company-specific production system (XPS) in environments with pervasive digital technologies relies on understanding diverse learning preferences. This study examines the integration of mobile micro-learning and gamified elements into the professional maintenance (PM) education framework of a Scandinavian automotive company, aiming to address limited instructor availability and improve engagement.Design/methodology/approachA sequential mixed-method approach was conducted in three phases. The first phase involved a pre-study with interviews to gather maintenance directors' strategic perspectives, complemented by questionnaires from maintenance blue-collar (technicians and engineers) and white-collar (managers) staff to identify current challenges in maintenance education. In the second phase, a questionnaire collected key inputs for designing an application, which was validated through a proof of concept. The third and final phase evaluated user engagement, training usefulness and satisfaction, using performance metrics for assessment.FindingsGamified elements enhance employee engagement by creating a competitive and interactive learning environment, leading to higher user satisfaction. The platform's flexibility supports self-paced learning for a global workforce and addresses translation issues. The study concludes that mobile learning should complement classroom training, as it improves overall training effectiveness by providing continuous learning opportunities and standardized, engaging solutions, filling gaps in industrial maintenance departments.Originality/valueWhile E-learning, mobile compatibility, gamification, microlearning, spaced repetition, interactivity and collaboration are commonly discussed in corporate education, their combined integration into PM education and training for XPS has not been addressed.

The rise of advanced digitalization in Industry 4.0 has enabled manufacturers to leverage data through AI and ML solutions for various manufacturing challenges. However, integrating these models into factory settings remains challenging, as models that perform well on static datasets struggle with dynamic shop floor data. MLOps is an emerging discipline focused on bridging the gap between ML models and production environments; however, in the manufacturing domain, questions remain about how to effectively deploy ML models using MLOps. This article addresses these gaps by conducting a systematic literature review combined with thematic analysis to explore architectures and frameworks used to adopt MLOps in real-world industrial applications, referred to here as industrial MLOps. The study identifies key architectural requirements and outlines seven implementation challenges, with recommendations and architecture mappings to overcome them. Results show that fully automated MLOps frameworks remain underdeveloped, and that modular, scalable architectures are recommended to address model drift, data quality, and integration challenges.

Flexible manufacturing requires industrial robots to be reprogrammed rapidly as product variants change. This paper presents a language-model-based workflow that generates, validates, and iteratively corrects ABB RAPID robot programs from natural language task descriptions. A dual-stream retrieval-augmented generation (RAG) pipeline grounds code generation in verified technical documentation and production templates, reducing domain-specific errors produced by ungrounded language models. A custom Model Context Protocol (MCP) server connects the language-model client directly to ABB RobotStudio for automated code upload, simulation execution, and diagnostic feedback. The evaluation combines a 30-query retrieval benchmark, scoped code-generation checks, and RobotStudio case studies in a simulated pickand- place manufacturing cell. The simulation loop exposes execution failures that static and semantic checks alone cannot catch, including suction release-height errors, unreachable placement targets, and configuration-dependent recovery motions. The results show how RAG and MCP can connect grounded code generation with executable feedback from industrial robot simulation software, while reducing but not eliminating expert setup and final supervision.

PurposeThe need for smart maintenance (SM) is increasing as the manufacturing industry digitalizes. To facilitate the transformation of maintenance in digitalized manufacturing, scholars have developed a strategy development process (SDP) for implementing SM. However, the SDP must be tested and evaluated, as manufacturing companies and industries need explicit guidance and empirical evidence on how to use it. Design/methodology/approachThis study employed action research to facilitate collaboration between researchers and maintenance professionals in a large Swedish manufacturing company, testing and evaluating the SDP for SM implementation. The study was conducted in multiple phases over two years, focusing on the real-world implementation of key activities in an industrial setting. FindingsImplementing SM in the manufacturing industry resulted in a refined SDP. The study revealed synergies between the implementation steps, from concept to practice. This refined process advances benchmarking (Activities 1.1–1.4), streamlines goal setting, prioritization and planning of key activities (Activities 2.1–4.1) and ensures authorized elevation and cross-functional communication (Activities 5.1–5.2) for digitalization of maintenance. Practical implicationsThe theoretical implications refine the SDP and confirm the value of creating, acquiring and transferring knowledge within maintenance organizations, thereby facilitating SM implementation with empirical evidence. The practical implications offer recommendations for factory and maintenance management, providing explicit guidance to manufacturing companies in developing maintenance for digitalized manufacturing. Originality/valueThe refined SDP is an evolutionary process that requires continuous learning. This reinforces the focus on organizational development rather than solely technological transformation, i.e. becoming a learning organization when implementing SM.

Digital Twins (DTs) research still lacks of management frameworks covering their entire lifecycle. This gap has hindered their successful development and sustainment in real-world industrial settings. This paper addresses this need by proposing a first-of-its-kind DT Lifecycle Management Framework that aims to offer a comprehensive technical and managerial guide for supporting DTs across their complete lifecycle. By providing this comprehensive structure, this paper aims to contribute to the ongoing international standardisation efforts, particularly of the ISO 23247 series on a "DT Framework for Manufacturing", and offer practitioners and researchers a guide for the methodological adoption and implementation of DTs.

Production Disturbances (PDs) negatively impact efficiency and productivity in manufacturing systems by consuming resources and causing losses. It is intuitive that production equipment maintenance, and thereby PDs handling, is tightly connected to sustainability by contributing to the extended longevity of technical systems. However, there is a lack of knowledge regarding how PDs impact Environmental Sustainability Indicators (ESIs). This missing connection makes it challenging for decision-makers to motivate green investments and to align work procedures to reduce environmental impacts. This study aims to bridge the gap between PDs and ESIs through a questionnaire-based survey to collect data and establish an integrated picture of PDs and ESIs in the context of Overall Equipment Effectiveness (OEE). The findings within our study indicated that quality and maintenance organizations should prioritize addressing defects, reduced yield and idling/minor stops to improve ESIs. Targeted strategies to reduce these PDs can lead to improvements in energy efficiency and waste reduction, ultimately contributing to net-zero emissions.

Opportunistic Maintenance (OM) remains underutilized in manufacturing despite being introduced over half a century ago. To break new ground, this article seeks to provide concept clarity and demonstrate the potential of artificial intelligence techniques for OM in a manufacturing context. Through an analysis of the existing OM literature, we provide clarity in the OM theory and distinguish two separate views of OM that we dub the 'component view' and the 'flow view'. We then integrate the two views into a new and unified conceptual definition of OM followed by expanding the OM concept by embedding four distinct time constructs: frequency, duration, sequence, and timing. To pave the way for novel OM tools, we demonstrate a real-world application of data-driven prediction of maintenance opportunity windows in an automotive manufacturing line using a long short-term memory algorithm. Evaluated against a na & iuml;ve benchmark, our model showed quantitatively superior predictive performance on precision, recall, and F1 score. Our theoretical and practical implications relate to increasing the coherence in OM scholarship, making OM research easily understandable by working professionals, and creating new directions for OM tools capable of learning, adapting, and responding to changing production dynamics. We thereby offer a unified foundation for creating impactful OM theory and tools, aiming to inspire maintenance scholars to pursue the OM topic in their own research to deepen the understanding of OM and fully unlock its productivity potential in manufacturing.

The convergence of artificial intelligence (AI) and digital twin technology is reshaping maintenance strategies in the era of Industry 4.0. However, gaps persist between academic advancements and industrial adoption and expectation. This study systematically investigates the landscape of AI-enhanced digital twins for maintenance by integrating a systematic literature review (SLR) of related studies with in-depth interviews from industry practitioners. Our analysis reveals that while academia demonstrates robust applications of supervised, deep, and reinforcement learning to optimize digital twin models and prescribe data-driven actions, industrial implementation remains limited by challenges such as high scale dimension, data integration complexities, and insufficient workforce readiness. We identified and articulated three critical gap dimensions, scale, data, and model between academic research and industrial implementation and expectation. To bridge these gaps, we proposed a comprehensive five-layer framework for AI-enhanced digital twins, encompassing physical assets, data transmission, digital twins, AI analytics, and maintenance services. Actionable recommendations are provided, including the adoption of modular architectures, standardized data protocols, hybrid edge-cloud solutions, and targeted workforce upskilling. Our findings not only clarify the current state and challenges of AI-driven digital twins in maintenance but also offer a practical roadmap for accelerating their industrial implementation. This work advances the field by integrating insights from both academic research and industrial practice, offering concrete recommendations to support the practical realization of smart and sustainable maintenance practices.

Ever-evolving energy market landscape together with higher attention towards environmental sustainability, is leading manufacturing companies to reshape and re-evaluate their internal processes so to favour a more energy-efficient management of productions systems. In light of this, being maintenance one of the main contributors to this goal, the introduction of energy within the maintenance strategy definition process may change the final outcome by balancing costs and production lost with energy-related performance. Therefore, this research aims at investigating how energy is currently incorporated into the maintenance strategy definition by carrying out a bibliometric study. Grounded on the results, the process for energy-efficient maintenance strategy definition can be decomposed into data input, energy-related indicators modelling and finally the maintenance strategy definition. Scientific literature confirmed that barriers and challenges are still present to exploit envisioned benefits of introducing energy while defining maintenance strategies. For this reason, this work identifies a roadmap to embed energy into maintenance strategy definition by increasing the complexity in terms of asset behavioural modelling, energy indicators modelling and achievable benefits.

Machine tools are essential to manufacturing for precise and efficient component production. With Industry 4.0, abundant machine condition data enables data-driven maintenance decisions. However, deploying condition-based maintenance solutions is challenging due to the diverse configurations of equipment, complex failure modes, and compatibility issues with the digital infrastructure. While machine tool health monitoring relies on detailed tests like Ballbar measurements, they consume valuable production time. To address these challenges, this article presents a human-centric development and deployment of a condition-based data-driven maintenance dashboard. The solution uses data from the controller system to improve machine tool testing in a Swedish heavy-duty vehicle powertrain facility. (c) Copyright 2025 The Authors.

The European battery industry is rapidly evolving due to demands for sustainability and digitalization. Large-scale battery production is essential for the energy transition but presents significant challenges, including in maintenance operations. By ensuring uptime and productivity, effective maintenance is key to industrialization. This study adopts a socio-technical lens to examine how sociological, technological, and organizational factors influence maintenance operations in battery production. Through ethnographic research within a real-world gigafactory, we gathered in-depth data on the socio-technical interactions of maintenance to identify critical challenges and establish important development needs. Thematic analysis resulted in the formulation of 31 distinct and relevant research avenues, providing industry and academia with strategic guidance and actionable blueprint for advancing maintenance operations in battery production.

Accurate anomaly detection and localization in sheet metal glue line applications are crucial for quality assurance in automotive manufacturing. Most current vision-based inspection systems that rely on geometric deviations from a predefined shape often suffer from high false-positive rates, leading to unnecessary interventions and operational inefficiencies. This research investigates the potential of unsupervised deep learning models to significantly reduce false positives in the analysis of sheet metal glue line images, even with limited datasets. We conducted a comparative evaluation of 17 unsupervised deep learning models covering different categories with 28 backbones on datasets of approximately 300 industrial glue line images per part from a Swedish vehicle manufacturer. A data synthesis method was applied to balance the glue line dataset, further enhancing the reliability of the models. To address the challenge of limited training data and improve model generalization, we incorporated data augmentation techniques and performed robustness experiments to ensure applicability to real-world industrial conditions. Our findings demonstrate that deep learning approaches can effectively detect and localize anomalies, significantly reducing false positives and gluing machine downtimes compared to the existing system. Moreover, we proposed a multi-criteria decision-making based approach for model selection, enabling decision-makers to achieve optimal trade-offs between accuracy and inference time, thus improving operational efficiency. These advancements highlight that even with limited training data, unsupervised deep learning models can enhance anomaly detection reliability, streamline the automotive production process, and reduce unnecessary resource expenditures.

The field of data science is an emerging area of study that arises in the context of the production of a large volume of data in recent years. The objective of this area is to obtain valuable information that is extracted through data processing. In the industrial context, the identification of failures and bottlenecks in production lines is essential to increase the productivity of the evaluated systems. However, manual analysis can be time-consuming and costly. Process discovery is a set of techniques that includes the use of algorithms to extract a process model from the event log, which can be used as a basis for developing Digital Twins. Therefore, this paper proposes the use of an artificial production line generator so that process mining algorithms can be tested with a large number of samples and different network characteristics. Thus, the main contribution will be the testing of hypotheses to assist in choosing the best algorithms in a practical context.

Maintenance and quality control are typically disjoint areas in a production system and even though interactions between them do exist, they are limited. In some cases, the quality deviations are reported directly by the client the product is sold to before maintenance actions are taken to repair the faulty machines and prevent these specific deviations. In this paper, we claim that by using machine and quality data in combination, it is possible to generate information about the process and the resulting product, that will allow to detect deviations in earlier stages, likely before the product reaches the client, possibly even before it is produced. We analyze a production process over a period of two years, during which operational parameters of the machines executing the process are reported, as well as the quality deviations of the parts produced. The data gathered is used to establish whether there exists a correlation between the machine status and the quality deviations of the products. Experiments show that the correlation increases when adjustments to the machines are made. This evidence supports our hypothesis of the possibility of using quality and machine data in combination in the development of future predictive maintenance solutions.

Advanced manufacturing research for sustainable battery life cycles is of utmost importance to reach net zero carbon emissions ( European Commission, 2023a ) as well as several of the United Nations Sustainable Development Goals (UNSDGs), for example: 30% reduction of CO 2 emission, 10 million job opportunities and access to electricity for 600 million people ( World Economic Forum, 2019 ). This editorial paper highlights international motivations for pursuing more sustainable manufacturing practices and discusses key research topics in battery manufacturing. Batteries will be central to our sustainable future as generation and storage become key components to on-demand energy supply. Four underlying themes are identified to address industrial needs in this field: 1. Digitalizing and automating production capabilities: data-driven solutions for production quality, smart maintenance, automation, and human factors, 2. Human-centric production: extended reality for operator support and skills development, 3. Circular battery life cycles: circular battery systems supported by service-based and other novel business models, 4. Future topics for battery value chains: increased industrial resilience and transparency with digital product passports, and next-generation battery chemistries. Challenges and opportunities along these themes are highlighted for transforming battery value chains through circularity and more sustainable production, with a particular emphasis on lithium-ion batteries (LIB). The paper concludes with directions for further research to advance a circular and sustainable battery value chain through utilizing the full potential of digitalization realising a cleaner, more energy-efficient society.

Europe's emerging lithium-ion battery production sector faces immense challenges with Supply Chain Complexity (SCC). This article explores sources of and responses to SCC within maintenance operations of battery production. Using an engaged scholarship approach within automotive original equipment manufacturers (OEMs) and battery cell manufacturers in Sweden, qualitative data were collected and analysed using SCC theory. Our core findings reveal 37 sources of SCC classified by origin and type and 40 responses across four practice clusters. This study offers in-depth insights into the potential impact of SCC on maintenance operations in battery production and provides actionable guidance for managing complexity. It also identifies five future research avenues: investigating complexity interactions, matching sources and responses, exploring complexity-inducing responses, identifying internal-external interfaces and examining social aspects of SCC. In effect, the study sets the agenda for research on battery production maintenance and positions SCC as a versatile theoretical lens for understanding the emerging battery sector.

Industry increasingly focuses on data-driven digital twins of production lines, especially for planning, controlling and optimising applications. However, the lack of open data on manufacturing systems presents a challenge to the development of new data-driven strategies. To fill this gap, the paper aim to introduce a strategy for generating random production lines and simulating their behaviour, thus enabling the generation of synthetic data. So far, such data can be recorded in event logs or machine status format, with the latter adopted for the use cases. To do so, the production lines are modelled using complex network concepts, with the system's behaviour simulated via an algorithm in Python. Three use cases were assessed, in order to present possible applications. Firstly, the stabilisation of working, starved and blocked machines was investigated until a steady state was reached. The system behaviour was then investigated for different model parameters and simulation intervals. Finally, the production bottleneck behaviour (a phenomenon that can harm the production capacity of manufacturing systems) was statistically studied and described. The authors anticipate that this artificial and parametric data benchmark will enable the development of data-driven techniques without prior need for a real dataset.

Industry increasingly focuses on Digital Shadows and Twins of production lines, especially for planning, controlling, and optimizing operations. In parallel, shop floor processes can be described using Discrete Event Simulation (DES) models, which are ranked among the top tools for manufacturing system decision support. Although, Process Mining (PM) and model-driven Digital Twins (DT) were investigated in separate research communities. The integration of these two research fields is essential for advancing industrial applications by reducing time and efforts to model and describe processes. Thus, the objective of this paper is to propose a data integration pipeline to enhance realistic event logs and support the early stages of Data-driven Modelling of DT through PM techniques. This paper is expected to provide three relevant contributions. The first contribution is the enhancement of the production system event logs through the implementation of data integration techniques. The second contribution is to enable machine learning techniques to be applied by trace profiling the enhanced event logs, generating an attribute-value database. The third contribution is to extract value from a process-centered analysis, increasing the data value from a practical perspective.