Optimized Prognostics Approach for Early Anomaly-Detection Methodology Used in Small Modular Reactors

Small modular reactors (SMRs) must be developed when nuclear power can be generated from each unit of smaller reactors, particularly in off-grid or remote areas. As the operation and maintenance costs of current nuclear power plants increase, distributed fleets will become more expensive if remote monitoring and maintenance are unavailable. To economize the power-plant cost, prognostics and health management via proactive maintenance for early warning and anomaly detection with high reliability can be employed actively. In particular, prognostics and diagnostics are prioritized in the nuclear discipline, thus necessitating on-line calibration. This paper suggests an optimized prognostics approach for an early anomaly-detection methodology used in SMRs, which has been discussed by nuclear researchers. Additionally, several expected challenging issues regarding SMRs are presented, and the suggested approach for an anomaly-detection system using on-line monitoring and a data-driven artificial-intelligence process for SMRs is introduced. The proposed approach enables autonomous operation and offers many advantages over previous systems. The application of a transformer layer and multiple modular modeling results in outstanding features such as robust feature extraction and favorable real-time detection. To exploit these advantages, the early anomaly-detection methodology should be discussed and established in the early design stage.