A review of predictive maintenance strategies for Industry 5.0: AI, digital twins and human–machine collaboration

Purpose The purpose of this article is to provide a comprehensive review of predictive maintenance (PdM) strategies within the emerging Industry 5.0 paradigm, emphasizing the shift from automation-centric maintenance to human-centric, collaborative systems. The study aims to synthesize model-based, data-driven, artificial intelligence (AI)-enabled and hybrid PdM approaches, examine the role of key enabling technologies such as digital twins and edge computing and highlight human–machine collaboration in maintenance decision-making. By proposing a unified taxonomy, the article seeks to identify research gaps and guide the development of resilient, sustainable and explainable PdM systems. Design/methodology/approach This study adopts a structured systematic review methodology to analyze PdM strategies in the context of Industry 5.0. Peer-reviewed journal articles, standards and authoritative reports were identified through major scientific databases using predefined keywords related to PdM, condition-based maintenance, AI, digital twins and human–machine collaboration. The selected literature was critically examined and classified based on maintenance approach, enabling technologies, deployment architecture and degree of human involvement. Comparative analysis was employed to identify research gaps, challenges and emerging trends, culminating in the development of a novel taxonomy for Industry 5.0-oriented PdM systems. Findings The review reveals that PdM in Industry 5.0 is evolving toward hybrid frameworks that integrate data-driven intelligence with physics-based models and human expertise. While AI-based approaches improve diagnostic and prognostic accuracy, challenges related to data quality, interpretability and workforce acceptance persist. Digital twins, edge computing and collaborative robots emerge as key enablers for real-time, resilient and human-centric maintenance systems. The analysis highlights a lack of unified frameworks addressing human–machine collaboration and resilience, underscoring the need for explainable, lifecycle-oriented PdM solutions tailored to Industry 5.0 environments. Research limitations/implications This study is limited by its reliance on published literature, which may underrepresent proprietary industrial implementations and emerging practices not yet reported in academic sources. The review does not include empirical validation or quantitative benchmarking of PdM methods, focusing instead on conceptual and methodological synthesis. Nevertheless, the proposed taxonomy and identified research gaps provide a foundation for future empirical studies, experimental validation and large-scale industrial case analyses aimed at developing resilient, human-centric PdM frameworks aligned with Industry 5.0 principles. Practical implications The findings of this review offer practical guidance for industrial practitioners seeking to implement PdM within Industry 5.0 environments. The proposed taxonomy assists decision-makers in selecting appropriate PdM strategies based on system complexity, human involvement and technological readiness. Insights into digital twins, edge computing and human-in-the-loop AI support the design of maintenance systems that enhance reliability, reduce unplanned downtime and improve workforce acceptance. The study also highlights the importance of explainable and collaborative maintenance solutions to facilitate successful adoption in real-world manufacturing settings. Social implications By emphasizing human-centric and collaborative PdM, this study highlights the social impact of Industry 5.0 on the industrial workforce. The integration of human-in-the-loop decision-making and collaborative robots can improve workplace safety, reduce physically demanding tasks and support skill augmentation rather than job displacement. Transparent and explainable maintenance systems foster trust between workers and intelligent technologies, facilitating organizational acceptance and long-term adoption. Ultimately, Industry 5.0-oriented PdM contributes to more inclusive, resilient and sustainable industrial environments that align technological advancement with human well-being. Originality/value This article provides one of the first comprehensive reviews that explicitly situates PdM within the Industry 5.0 paradigm, moving beyond automation-centric perspectives. Its originality lies in the integration of PdM, condition-based maintenance, diagnostics and prognostics through a unified, human-centric lens. The study proposes a novel taxonomy that incorporates digital twins, edge computing and human–robot collaboration, offering a structured foundation for future research. By bridging technological, organizational and human factors, the article delivers both conceptual clarity and practical value for advancing next-generation maintenance systems.