Digital twins (DTs) are increasingly proposed for nuclear-reactor monitoring, diagnosis, prediction, control, and decision support, but the maturity of reported systems is difficult to compare because DT realization, physical anchoring, synchronization, adaptation, validation, real-time execution, and trustworthiness are often conflated. This systematic scoping
[...] Read more.
Digital twins (DTs) are increasingly proposed for nuclear-reactor monitoring, diagnosis, prediction, control, and decision support, but the maturity of reported systems is difficult to compare because DT realization, physical anchoring, synchronization, adaptation, validation, real-time execution, and trustworthiness are often conflated. This systematic scoping review maps substantive DT research for nuclear fission reactors and nuclear power plants using a strict substantive-DT definition and a multidimensional evidence-coding framework covering DT realization, physical anchoring, adaptation, validation, real-time capability, and trustworthiness. Scopus, Web of Science Core Collection, and IEEE Xplore yielded 845 database records and 590 unique records after deduplication. A multi-pass title/abstract workflow classified 182 records as Include, 397 as excluded from the primary-study stream, and 11 as Uncertain, yielding 193 reports sought for retrieval. Despite iterative multi-source retrieval attempts, 96 reports could not be retrieved. One additional Uncertain conference report was linked to an already retained journal report and was not treated as an independent report; 96 retrieved reports underwent primary full-text eligibility assessment. Seventy-one retrieved and eligible peer-reviewed primary reports met the inclusion criteria. Of these, 44/71 (62%) were Tier A implemented or integrated DT systems. Physical anchoring ranged from P0 simulation-only realization (39/71, 55%) to P3 operating-reactor or plant evidence (9/71, 13%); within P3, four reports used retrospective/offline operating data and five demonstrated live or near-live reactor/plant linkage. Validation evidence comprised 40/71 reports (56%) with development-linked validation (V1), 26/71 (37%) with separated empirical or benchmark validation (V2), and 5/71 (7%) with prospective, challenge-based, or independently confirmed validation (V3). Adaptation was limited: 51/71 reports (72%) demonstrated no adaptation (A0), 3/71 (4%) episodic or intermittent recalibration (A1), 5/71 (7%) online state synchronization/estimation (A2-S), 12/71 (17%) online parameter/model adaptation (A2-P), and none met the A3 criterion. Computational real-time capability was supported in 42/71 reports (59%), whereas integrated real- or near-real-time execution was demonstrated in 27/71 (38%); live physical-data ingestion, explicit recurring physical-to-digital synchronization, and automatic DT-to-physical actuation were demonstrated in 11/71 (15%), 5/71 (7%), and 1/71 (1%), respectively. Hybrid modelling was the dominant paradigm (34/71, 48%). Trustworthiness was asymmetric: calibration and robustness/OOD assessment were much more common than uncertainty quantification, explainability, cybersecurity, traceability, human oversight, or demonstrated regulatory readiness. A targeted data audit identified only four reports linked to clearly open and reusable benchmark or downloadable resources. Family-aware, quality, physical-evidence, temporal, and data-access sensitivity analyses did not materially change the central conclusions. Nuclear-DT maturity is best understood as multidimensional; future progress requires staged physical validation, bounded adaptive synchronization, trustworthy decision support, reproducible data/model provenance, and independent deployment-oriented validation.
Full article