Changhong Peng

This study investigates the containment pressurization response of the International Reactor Innovative and Secure (IRIS) during a bounding 4-inch Small Break Loss-of-Coolant Accident (SBLOCA) on the Chemical and Volume Control System (CVCS) line. An explicitly coupled RELAP5/MOD3.4-CONTEMPT-LT framework is employed, enabling simultaneous evaluation of primary system and containment behaviour. Time-resolved break discharge parameters from RELAP5—including liquid and vapor mass flows and specific enthalpies—serve as input to the containment model, which incorporates the IRIS Containment Pressure Suppression System (CPSS) and detailed heat structures. The analysis predicts a peak containment pressure of 0.842 MPa at 444 s, remaining well below the 1.2 MPa design limit. The results confirm that the coupled approach captures key reactor–containment interactions and provides a robust basis for deterministic safety assessment of IRIS passive safety systems.

To enhance computational efficiency in estimating the posterior failure probability of passive systems, this study proposes an innovative framework integrating inverse uncertainty quantification (IUQ) with adaptive metamodel-based subset importance sampling (AM-SIS). This approach leverages updated input parameter distributions for precise posterior failure probability assessment. Within the IUQ framework, Bayesian neural networks is employed to assist established Markov Chain Monte Carlo methods for posterior sampling, significantly reducing computational costs. The coupled IUQ-AM-SIS framework is validated on polynomial benchmark cases and the equal height difference passive containment cooling system (EHDPCCS), examining scenarios with single and double calibration parameters. Results demonstrate that integrating the AM-SIS method significantly improves computational efficiency for posterior failure probability estimation, particularly in low failure probability regimes.

Small reactors have the advantages of compact structure and wide application scenarios, and can be deployed in remote areas. However, due to the complex structure and special application scenarios, it is more difficult for operators to diagnose and identify accident information. This article conducts accident diagnosis research using simulated data from IRIS reactors. An accident diagnosis model based on LSTM neural network has been established. The diagnostic ability of the model has been analyzed when there is an error in the time of the accident occurrence. When there are numerical and time errors in the dataset, the simulated diagnostic accuracy is 96.8 %. The method of using a diagnostic model based on shutdown signal partitioning has improved the diagnostic ability of transient operating events and minor accident conditions, achieving an accuracy of 96.1 % when there is ±5 % noise in the dataset. By optimizing the output form of the diagnostic model, the diagnosis of unknown accidents was achieved. When using LOCA and LOFA accidents as unknown accidents, the diagnostic accuracy of the model reached 98.5 %.

Recent advances in high-fidelity modeling, numerical computing, and data science have spurred interest in model-data integration for nuclear reactor applications. While machine learning often prioritizes data-driven predictions, this study focuses on data assimilation (DA) to synergize physical models with measured data, aiming to enhance predictive accuracy and reduce uncertainties. We implemented deterministic DA methods—Kalman filter (KF) and three-dimensional variational (3D-VAR)—in a one-dimensional single-phase natural circulation loop and extended 3D-VAR to RELAP5, a system code for two-phase loop analysis. Unlike surrogate-based or model-reduction strategies, our approach leverages full-model propagation without relying on computationally intensive sampling. The results demonstrate that KF and 3D-VAR exhibit robustness against varied noise types, intensities, and distributions, achieving significant uncertainty reduction in state variables and parameter estimation. The framework’s adaptability is further validated under oceanic conditions, suggesting its potential to augment baseline models beyond conventional extrapolation boundaries. These findings highlight DA’s capacity to improve model calibration, safety margin quantification, and reactor field reconstruction. By integrating high-fidelity simulations with real-world data corrections, the study establishes a scalable pathway to enhance the reliability of nuclear system predictions, emphasizing DA’s role in bridging theoretical models and operational demands without compromising computational efficiency.