Generative Digital Twin-Enabled Prognostics and Health Management: A Systematic Review of Concepts, Architectures, and Key Technologies

Prognostics and health management (PHM) for industrial equipment is still constrained by scarce fault samples, incomplete observation of degradation processes, limited cross-condition generalization, and fragmented maintenance knowledge. Digital twins (DTs) provide a physical–virtual integration basis for PHM through state synchronization, model updating, and service feedback. However, conventional DT-enabled PHM remains limited in data supplementation, degradation inference, knowledge organization, and maintenance decision-making. The rapid development of generative artificial intelligence (GenAI) offers a new pathway for enhancing DT-enabled PHM. This review focuses on generative digital twin (GDT)-enabled PHM. It clarifies the conceptual boundaries and functional enhancement logic of GDT-enabled PHM, summarizes the roles of key GenAI models, and reviews technical progress across health data synchronization, twin data processing, virtual modeling, prognostic reasoning, and maintenance decision feedback. A cross-process perspective is introduced to trace how data, model, and reasoning errors propagate into PHM reliability risks. Representative applications in aero-engines, rotating and transmission equipment, and energy equipment are then examined with attention to their validation boundaries. Finally, the review discusses generated-data validity, physical consistency, lifecycle updating, uncertainty, and engineering deployment under Industrial Internet of Things constraints.