ABSTRACT To effectively mitigate stochastic disturbances during production and ensure the on‐time delivery of orders, manufacturing systems must demonstrate resilience by rapidly generating revised production schemes within critical time windows. Although traditional scheduling methods facilitate system recovery, existing centralised approaches often suffer from computational paralysis and decision failure due to excessive rescheduling frequencies. Conversely, distributed multi‐agent methods are frequently constrained by limited local observability, leading to suboptimal convergence. To address these challenges, this study proposes a multi‐agent deep reinforcement learning (MADRL) scheduling method based on dynamic information entropy‐driven QMIX (DE‐QMIX). First, the flexible job‐shop scheduling problem is formulated as a partially observable markov decision process (POMDP), empowering individual machine agents to respond to disturbances based on localised perceptions. Second, a self‐attention mechanism is embedded within the agent architecture to optimise feature weight allocation, effectively suppressing policy degradation in high‐dimensional observation spaces. Finally, the cascading effects of disturbance events are inferred using a shop‐floor job network adjacency matrix. By integrating node‐level information entropy to quantify global state fluctuations, the proposed method significantly enhances the operational robustness of the system under multi‐disturbance scenarios.
ABSTRACT Digital twin technologies are gaining importance in modern intelligent manufacturing to facilitate increased efficiency, smartness in manufacturing, and sustainability. Nevertheless, the choice of a suitable digital twin solution is a complicated long‐term decision because of the high rate of technological development, integration issues, and financial limitations. To solve these problems, this research proposes a multi‐criteria decision‐making model based on the circular q‐rung orthopair fuzzy compromise of ideal solution (CqROF‐CoCoFISO). The proposed model uses circular q‐rung orthopair fuzzy sets, which is an effective model for capturing the uncertainty, hesitation and ambiguity that exists in smart manufacturing environments. The most important assessment criteria will be the technological flexibility, compatibility with the system, the cost of implementation, functionality of the system, intellectual property protection, adherence to security and sustainability. The CoCoFISO mechanism will allow ranking competing digital twin alternatives reliably based on conflicting criteria, which will improve the strength and stability of decision‐making results. The usefulness of the proposed framework can be proved through a comparative analysis. The findings offer useful information to manufacturing engineers, decision‐makers and policymakers aiming at informed and credible digital twin adoption policies in changing industrial conditions.
ABSTRACT In the realm of Industry 5.0, the concept of Operator 5.0 emphasises the role of technologies enhancing human capabilities and adaptability. Augmented Reality, which shows potential to improve productivity, cognitive support, and workplace efficiency is one promising technology, being able to provide real‐time access to instructions, facilitate assistance, and enhance operator performance. Nevertheless, its industrial adoption faces several challenges, including the need for specialised expertise in system design and concerns regarding the technology’s usability. To fulfil the gaps related to the development of human‐centric AR solutions, this study explores the impact of AR‐based guidance in assembly processes, through an experimental approach. A test conducted in laboratory involving 24 users demonstrates the AR's effectiveness in reducing errors and achieving high rates of usability, even if time increases in the tasks' completion were observed. The findings offer theoretical and practical insights to foster successful implementation of AR in manufacturing, understanding its benefits and challenges and highlighting its potential to optimise assembly operations and enhance workforce wellbeing and resilience.
ABSTRACT With the context of the human‐centric paradigm of Industry 5.0, this paper empirically tests the extent to which human–machine collaboration modes contribute to the optimisation of decision‐making in manufacturing. It determines the impact of different types of collaboration on economic efficiency, resilience and sustainability and discusses the role of AI capabilities in this process and the moderating role played by the level of digital transformation of enterprises. A mixed methodology was used. A large‐scale questionnaire survey ( n = 519) of manufacturing managers across different industries and case‐based interviews were adopted. The theoretical framework was examined through structural equation modelling (SEM) because hypotheses were formulated based on paradox theory, and direct, mediating and moderating effects were analysed. Findings affirm meaningful dimension‐specific direct effects; task‐division collaboration increases economic efficiency, real‐time interaction increases resilience and decision‐complementarity collaboration increases sustainability. In part, these relationships are mediated by AI abilities (efficiency in data processing, adaptive learning speed, friendliness of the interface). Although the moderating role of the stage of digital transformation is statistically not significant, the analysis of heterogeneity shows significant differences across industries. The paper has three contributions: it creates a rigorous, mutually exclusive set of typologies of three modes of collaboration to fill classification gaps; defines the criteria of selection of instrumental AI capabilities and their association with collaborative decision‐making and enumerates the factory‐level evaluation standards of decision optimisation. It provides a new understanding of the micro‐mechanisms of representation of collaborative practices into multi‐dimensional optimisations that would bridge the Industry 4.0 and 5.0 transformations.
ABSTRACT Automated Guided Vehicles (AGVs) operating in flexible manufacturing systems are exposed to uncertain travel conditions and random failures that significantly affect system performance. Traditional AGV routing methods often rely on additive cost models or deterministic reliability measures, which may fail to capture the probabilistic nature of failure events and the trade‐off between efficiency and survivability. This paper proposes a unified stochastic fractional routing algorithm for AGV networks in which arc failures are modelled as exponential random processes. Each arc traversal is associated with an expected travel time and an expected cumulative hazard representing failure risk. Path reliability is interpreted probabilistically through a survival function, leading to a fractional objective that minimises expected travel time per unit of expected hazard. Based on this formulation, a new labelling‐based algorithm is developed that incrementally constructs feasible paths, applies deterministic Pareto dominance on expectations rules to eliminate suboptimal partial routes and identifies the optimal solution without exhaustive enumeration. Computational experiments on randomly generated AGV networks demonstrate that the proposed algorithm consistently identifies the optimal fractional path whilst significantly reducing computational effort compared with complete Pareto enumeration and other benchmark approaches. Sensitivity analyses further illustrate the algorithm's robustness to variations in failure rates and network density. The results confirm that the proposed stochastic formulation provides a physically consistent and computationally efficient framework for reliability‐aware AGV routing in manufacturing environments.
ABSTRACT This article presents and experimentally validates DT‐virtual‐flow, a human‐centric VR environment designed for production flow simulation and operator training based on a digital twin. The study addresses the key problem of temporal misalignment between physical processes and their DT during data transmission over unstable networks, increasing the risk of critical errors in VR control. To solve this and align with the human‐centric principles of Industry 5.0, the authors propose a hybrid human‐in‐the‐loop architecture combining event‐driven WIP synchronisation, a VR interface with tactile feedback and a dead reckoning + LERP delay compensation module. This fundamentally distinguishes the solution from traditional snapshot approaches. The integration of heterogeneous equipment is ensured by an HAL abstraction layer based on OPC UA, normalising telemetry for the digital processing pipeline. In a randomised controlled experiment ( n = 30), eliminating visual latency reduced critical errors by 76.4% and shortened skill acquisition time by 20.1%. Additionally, the results show a 9.4% increase in OEE, whereas ablation analysis confirmed that disabling network delay compensation leads to a 3.8‐fold increase in accident rates. This emphasises the decisive role of algorithmic synchronisation for the safe ergonomic use of VR in next‐generation industrial systems.
ABSTRACT 3D printed concrete technology has revolutionised construction methods by enabling the fabrication of complex structures. However, complex nonlinear coupling relationships exist among its process parameters, such as extrusion velocity, printing speed and printing height, which critically influence forming accuracy. Therefore, achieving precise prediction of geometric dimensions is essential for high‐precision construction. This paper first establishes a nonlinear mapping model between process parameters and geometric dimensions through experiments, effectively addressing the challenge of multi‐parameter coupling effects. To overcome the limitations of traditional optimisation algorithms—namely slow convergence in high‐dimensional parameter spaces and susceptibility to local optima—this study proposes an (ECOA). This algorithm integrates two key innovations: the exploration phase of the Gorilla Troop Optimiser and the vertical crossover strategy. The ECOA‐BPNN model for geometric dimension prediction is constructed using the BP neural network (BPNN) optimised by ECOA. Experimental results demonstrate that among 23 standard test functions and the CEC2019 test function, ECOA achieved global optimal solutions for 18 functions, whereas the fitness values for the remaining functions also outperformed other comparison algorithms, validating the optimisation capability of ECOA. The ECOA‐BPNN model significantly outperformed traditional BPNN in prediction accuracy, reducing MAPE and RMSE by 1.2862% and 0.6828%, respectively, and demonstrating its effectiveness in predicting 3D printed concrete geometric dimensions. Based on this model, optimised process parameter combinations were further derived, enabling precise control and high‐quality forming.
ABSTRACT This study identifies and models the key challenges that hinder Industry 4.0 (I4.0) adoption, supporting organisations in prioritising mitigation efforts while advancing sustainable production goals. A review of the literature led to the selection of 13 major challenges and 57 related sub‐challenges. Using Interpretive Structural Modelling (ISM), an eight‐level hierarchical structure was developed to map the interrelationships among these challenges. MICMAC analysis indicates that most challenges are linkage variables, characterised by strong interdependence and high system sensitivity. The Interpretive Ranking Process (IRP) was employed to rank the challenges based on their dominance across performance dimensions. Results from both ISM and IRP reveal that insufficient advanced technologies and supporting processes constitute the most critical barrier to I4.0 implementation. The combined ISM–IRP framework offers a practical, decision‐oriented tool that enables managers to identify priority areas, allocate resources effectively, and design more coherent strategies for I4.0 adoption in alignment with organisational needs.
The transition towards Industry 5.0 demands manufacturing systems that seamlessly integrate digital technologies with human-centric and sustainability objectives, enabling resilient and data-driven operations. This paper introduces an integrated industrial metaverse platform that unifies real-time Internet of things (IoT) sensing, dynamic digital twin modelling and immersive collaborative environments to support multistakeholder interaction, operational awareness and sustainability monitoring in manufacturing systems. The platform enables users to interact with a realistic 3D virtual representation of the factory via desktop and extended reality (XR) interfaces, accessing contextualised data on machine performance, energy use, environmental conditions and emissions. By embedding sustainability key performance indicators-such as energy consumption, CO2 emissions and waste metrics-directly within the immersive environment, the platform moves beyond conceptual frameworks and offers an operational solution that supports predictive maintenance, process optimisation and collaborative decision-making aligned with human-centric manufacturing goals. A pilot implementation in a textile manufacturing setting demonstrates improvements in operational transparency, reduced downtime, enhanced remote collaboration and greater energy efficiency. Results illustrate the potential of industrial metaverse technologies to advance sustainable, resilient and human-centric manufacturing, providing a scalable framework suitable for broader industrial adoption.
This study examines the transformative impact of artificial intelligence (AI) on supply chain management (SCM) within the automotive manufacturing sector, with a focus on AI adoption, its influence on key performance indicators (KPIs) and integration challenges. A mixed-methods research design was employed, combining quantitative survey data collected from supply chain professionals with qualitative insights derived from semi-structured interviews with industry experts. The quantitative data were analysed using statistical techniques to assess the relationship between AI adoption and SCM performance, whereas thematic analysis of interview data provided deeper contextual understanding of organisational and technological dynamics. The results indicate that AI adoption significantly improves operational performance, particularly in inventory turnover, delivery lead times and order accuracy. However, the findings also reveal critical barriers, including data quality limitations, system integration complexities, ethical concerns related to algorithmic bias and data privacy and potential workforce displacement. The study further identifies organisational readiness-driven by leadership commitment, innovation-oriented culture and robust technological infrastructure-as a key enabler of successful AI implementation. Based on these findings, the study recommends the development of structured AI adoption frameworks, including phased implementation strategies, investment in digital infrastructure, and targeted workforce reskilling initiatives. It also emphasises the need for ethical governance mechanisms to ensure transparency, accountability, and responsible AI deployment. These insights provide actionable guidance for practitioners and policymakers seeking to enhance supply chain performance while addressing the broader socio-technical implications of AI integration.
ABSTRACT The evolution of industrial robotics is increasingly shaped by the transition from Industry 4.0 to Industry 5.0, where manufacturing extends beyond automation efficiency toward intelligent, human‐centric, sustainable and resilient systems. This study presents a systematic literature review of 137 peer‐reviewed articles, conducted following PRISMA guidelines, to examine the changing role of industrial robotics across technological, human and value‐oriented dimensions. The findings show that industrial robots have evolved from isolated automation units into cyber‐physical, data‐driven and intelligent components embedded within smart manufacturing architectures. Key enabling technologies, including digital twins, artificial intelligence, IIoT and cloud edge computing, are identified as primary drivers of this transformation. Beyond technological advancements, the review highlights a shift toward human‐centric manufacturing, where collaborative robotics, ergonomic design, worker augmentation and explainable AI redefine human–machine interaction. The results further indicate that Industry 5.0 introduces new evaluation criteria, emphasising lifecycle sustainability, energy efficiency, circular manufacturing and system resilience. However, challenges remain, including interoperability limitations, high implementation costs, organisational resistance and governance gaps in AI‐enabled systems. This review develops an integrated framework linking architectures, applications, human factors and Industry 5.0 values and outlines key research directions for advancing responsible and sustainable industrial robotics.
In modern manufacturing value chains, achieving innovation and optimal performance often requires strong collaboration across organizational boundaries. Tackling the multi-party optimization problems arising from these partnerships necessitates methodologies that can learn from complementary datasets, where each partner provides unique partial features for a shared product or process. Although Data-Driven Evolutionary Optimization (DDEO) is an effective paradigm for solving complex optimization problems, its potential is limited by the need to centralise sensitive proprietary data. Although initial research on privacy-preserving DDEO has emerged, these efforts focus exclusively on horizontally partitioned data and rely on weight aggregation methods that are structurally incompatible with vertically partitioned data. Consequently, secure collaborative optimization across value chains-where data are vertically distributed among partners with distinct features-remains largely unaddressed. To bridge this gap, we propose a novel protocol enabling secure DDEO on feature-partitioned datasets. The architecture protects information in transit via standard encryption and uses Trusted Execution Environments (TEEs) to securely aggregate and process the combined datasets within a hardware-isolated enclave. This creates a versatile, model-agnostic computational environment that accommodates diverse methodologies, allowing practitioners to leverage different algorithmic advantages. Empirical evaluations using four state-of-the-art DDEO algorithms show that the proposed protocol achieves solution quality statistically identical to an insecure centralised baseline. Ultimately, the framework provides a practical solution for multi-party optimization without compromising data confidentiality.
The development of a reliable and cost-effective detection system of tool wear is essential for monitoring the machining process. Researchers in recent years have proposed many methods to implement tool condition monitoring (TCM) with the use of timeseries signals acquired by high resolution sensors. However, the high installation and operation costs of these sensors incurs high difficulties for small and medium sized enterprises, whereas using a low-cost alternative like the smartphone is a viable option. This work introduces a smartphone-enabled approach that utilises the smartphone vibration sensor with deep learning approach to detect the tool wear in two categories: Fresh and Worn. An extended Kalman filter noise reduction technique is applied to reduce undesired noise effects. Then, the Euclidean norm (l2-norm) is applied to determine the overall magnitude of vibration and the resulted data is z-normalisation and transformed by using the sliding window method into a suitable input shape for the 1D CNN model. The implemented method is validated with real-world datasets from the milling machining of Al 6061.
As concrete 3D printing emerges as a transformative technology in construction, optimising printing path planning has become crucial for enhancing manufacturing efficiency and structural integrity. Conventional path planning methods often struggle with complex geometries, leading to discontinuous paths, excessive directional changes and inefficient start-stop cycles. To address these limitations, this study introduces a novel reinforcement learning (RL)-based framework incorporating a dual-strategy coordination mechanism. The proposed approach employs two specialised deep Q-network (DQN) models: one for continuous path optimisation and the other for start-stop sequence refinement. The state and action spaces are designed based on sliced geometric features of concrete components, while a multi-objective reward function penalises nonprintable movements, abrupt turns and unnecessary interruptions. In addition, a path merging strategy and a G-code generation module are incorporated to ensure path continuity and operational feasibility. Experimental evaluations across diverse complex topological structures demonstrate that the proposed approach achieves complete edge coverage, minimises idle motions and path oscillations, and significantly improves printing efficiency and geometric fidelity. The results underscore the potential of the dual-strategy RL method as an intelligent and viable solution for advanced additive manufacturing path planning.
The article presents the concept of multi-level dynamic employee value (MLDEV) as an innovative approach to human resource management in flexible project-oriented maritime logistics enterprises. The purpose of this study is to develop a methodological framework for integrating human capital into digital management systems operating within the Industry 5.0 paradigm, with a focus on human-centred development. The paper proposes a model that combines three levels of assessment-individual, team and organisational, allowing to track the dynamics of employee performance over time. The key parameters (competence, motivation, adaptability, synergy of interaction and innovation) and relevant indicators that form an integrated assessment of the value of personnel are identified. Particular attention is paid to the digital implementation of the model through artificial intelligence tools, HR analytical dashboards and management decision support systems that provide forecasting of personnel risks and increase the efficiency of organisational interaction. The practical significance of the study lies in the formation of a scientific basis for the creation of intelligent human resource management systems in maritime logistics companies focused on adaptability, innovation and sustainable development.
Gas Metal Arc Welding is a widely used process in manufacturing due to its versatility, speed, and cost-effectiveness. However, variations in process parameters can lead to quality issues affecting production efficiency and increasing rework costs. Optimising these parameters is essential to ensure high-quality welds, reduce material waste and improve overall manufacturing productivity. Traditional optimisation models are often limited by specific welding conditions and constrained experimental data. This research employs machine learning and big data analytics to develop a generalised optimisation model for Gas Metal Arc Welding, leveraging literature data to overcome experimental limitations. Data preprocessing, including imputation for missing values, was applied to enhance data quality. Supervised machine learning algorithms were compared for predicting key weld characteristics, including tensile strength, penetration, and weld width. A Medium Gaussian SVM model predicted weld tensile strength with 85.60% accuracy, while an interactions linear model achieved 80.40% accuracy for weld penetration. A stepwise linear regression model provided 95.80% accuracy for weld width prediction. By optimising machine learning models for different manufacturing scenarios, this study offers a data-driven approach to parameter selection, improving weld quality and operational efficiency. The findings bridge machine learning-based welding optimisation and industrial applications, supporting data-driven decision-making for enhanced production performance.
This paper presents the development and validation of a digital twin system for a bakery production line, integrating real-time sensor data, physics-based process models and advanced predictive analytics with CNN + LSTM neural networks. The proposed architecture combines logistic growth, moisture evaporation and heat transfer equations with deep learning for accurate prediction and early detection of baking defects. Simulation and pilot implementation results demonstrate that the digital twin reproduces dough volume dynamics with an error below 3%, predicts humidity within +/- 2% and stabilises oven temperature in a narrow range (+/- 1.2 degrees C). The intelligent system enabled a 77% reduction in unplanned equipment downtime, decreased alarm events by over 60% and reduced the share of defective products from 8% to 2%. These outcomes highlight the practical impact and scalability of the hybrid digital twin framework for improving product quality, minimising losses and enhancing process reliability in food manufacturing.
This study addresses an integrated job shop scheduling problem with material handling robots (JSPMHR), aiming to minimise the maximum completion time (makespan). First, a mathematical model is developed to formulate the JSPMHR. Second, three meta-heuristics, genetic algorithm (GA), particle swarm optimisation (PSO) and artificial bee colony (ABC), are improved to solve the concerned problems. Based on the problem-specific features, seven local search strategies are designed to improve the convergence speed. Third, two reinforcement learning algorithms, that is, Q-learning and Sarsa, are employed to assist meta-heuristics in selecting the premium local search strategies during iterations. Finally, comprehensive experiments are conducted to evaluate the performance of the proposed algorithms by solving 82 benchmark instances. The proposed GA with Q-learning shows the strongest competitiveness among all compared algorithms solving JSPMHR.
Digital twins (DTs) and immersive extended reality (XR) interfaces offer new opportunities for intuitive, human-centred interaction in smart manufacturing. However, implementations often lack rigorous validation, quantitative performance analysis and assessment of scalability and robustness, especially in small-scale or resource-constrained manufacturing settings. This work proposes a modular metaverse framework integrating ROS2, Unity and Meta Quest devices to develop interactive, bidirectional DTs enhanced with gesture-based and mixed reality (MR) control. The framework is demonstrated through a lab-scale case study combining a robotic Wire Arc Additive Manufacturing (WAAM) system and a collaborative robot-based laser-cleaning station, showing broad applicability across industrial robotics. To evaluate usability, a preliminary user study with 17 participants was conducted, comparing a standard teach pendant with the proposed XR interface for a tool-inspection task. Results show an 80% reduction in programming time and significant decreases in perceived workload. Measured end-to-end gesture-to-action latency ranged from 430 to 450 ms, representing suitable timeframe for high-level interaction and task initiation. The study provides empirical user-centred evidence for metaverse-enabled interaction and discusses scalability, latency and industrial constraints. Aligned with the human-centric intelligence principles of Industry 5.0, the proposed approach improves accessibility, operator well-being and adaptability, contributing towards more inclusive and robust smart manufacturing systems.
The accelerating adoption of generative artificial intelligence (AI) is reshaping sustainable product design, yet current research remains fragmented across computational design, multi-objective optimisation, and smart manufacturing. This systematic review addresses this fragmentation by analysing 59 peer-reviewed studies (2010–2025) using PRISMA guidelines, advanced bibliometric mapping, and structural topic modelling to uncover how these domains converge to create superior sustainability outcomes. The study develops the Technology Convergence Framework, a unified theoretical model that integrates Advanced Computational Methods, Multi-Objective Optimisation, and Smart Manufacturing into an interconnected system capable of delivering emergent performance improvements. Findings show that when these domains operate synergistically—supported by mechanisms such as infrastructural maturation, empirical validation feedback loops, and standardisation-driven diffusion—manufacturers achieve 30%–65% gains in energy efficiency, waste reduction, and material optimisation, far exceeding improvements achieved through isolated technological efforts. The framework further incorporates human-AI collaboration principles aligned with Industry 5.0, emphasising the critical role of human judgement, contextual reasoning, and ethical oversight in complementing AI-driven decision systems. By bridging methodological, technological, and operational gaps, this review provides a holistic roadmap for transitioning from fragmented innovation to integrated sustainable product realisation, offering both scholars and industry leaders a coherent foundation for advancing next-generation sustainable manufacturing ecosystems.
