Preventive to intelligent maintenance: uncovering research frontiers of maintenance using BERTopic modeling

Purpose Maintenance and reliability engineering have evolved from reactive, asset-focused practices toward intelligent, data-driven, and human-centric systems. Despite extensive research, limited attention has been given to understanding the long-term thematic evolution of this field. This study aims to identify dominant and emerging research themes, assess their intellectual maturity, and examine their evolution. Design/methodology/approach The study retrieved 3,967 articles indexed in Scopus database in accordance with the PRISMA protocol, covering the period between 1969 and 2025. We then applied BERTopic modeling to the titles and abstracts to identify latent thematic structures within the literature. Findings Our findings reveal 11 major research themes, indicating a transition from traditional reliability analysis toward intelligent, system-oriented and data-driven maintenance strategies. Predictive maintenance technologies emerge as the most mature and influential theme, while advanced fault diagnosis and IoT sensor networks enabled maintenance to represent rapidly expanding research frontiers. Practical implications The results emphasize the importance of adopting predictive, risk-informed and human-aware maintenance strategies supported by sensing technologies and advanced analytics. Originality/value Our study offers an AI-enabled, long-term thematic synthesis of maintenance research, providing a comprehensive intellectual map and guidance for future research and practice.