Physics-Informed Digital Twins for Rotating Machinery Maintenance: Data Synchronization, Degradation, and Decision Reliability

Digital twins are increasingly proposed for rotating machinery, yet many implementations are offline models with limited connection to operational decisions. This critical review differentiates static digital models, automatically updated digital shadows, and bidirectionally integrated digital twins in the context of bearings, pumps, and industrial drives. It examines sensing, state estimation, physics-informed degradation modeling, remaining useful life prediction, and maintenance decision logic. Published frameworks for condition-based maintenance and machinery prognosis are synthesized with digital-twin literature to identify which capabilities require online feedback and which can be assessed offline. A conceptual test protocol considers operating regime changes, sparse fault labels, delayed maintenance records, and asymmetric costs of early and late interventions. The resulting agenda emphasizes synchronization quality, uncertainty and human oversight instead of maximizing a single prognostic accuracy score. No new maintenance experiment or machine-life estimate is claimed.