Abstract Predictive maintenance (PdM) leverages artificial intelligence (AI) and data analytics to forecast equipment failures in smart manufacturing, enabling timely interventions that minimize downtime and operational costs. This literature review examines recent advancements in PdM, focusing on three interrelated dimensions: (1) data challenges and limitations, (2) role of advanced AI models, and (3) actionable decision-making with human-AI collaboration. Unlike previous studies that often address these aspects in isolation, our review synthesizes them to provide a comprehensive understanding of current capabilities and limitations. We highlight how emerging AI technologies such as generative models, large language models (LLMs), and hybrid frameworks enhance predictive accuracy, enable synthetic data generation, and support interpretable, human-centered maintenance strategies. By evaluating both strengths and gaps across existing approaches, this work offers a comprehensive foundation for developing more scalable, reliable, and adaptable PdM systems aligned with Industry 5.0 principles through the integrative data–model–human (DMH) framework.
Diego Reforgiato Recupero (2025) studied this question.