Conference Agenda

The analysis of severe accidents in nuclear power plants is critical due to their potentially catastrophic impacts on public safety and the environment, underscoring the need for severe accident analysis codes like MELCOR. However, MELCOR faces two major challenges: (1) difficulty in solving multi-physics problems and (2) instability caused by complex computational schemes. To address these issues, this study investigates the feasibility of a physics informed neural network (PINN)-based MELCOR code by designing and evaluating a module for the CVH/FL package. MELCOR's governing equations were first implemented as a Python-based module for a thorough understanding, followed by the development of a PINN-based module applied to simplified 2-tank and 3-tank gravity problems. While both models approximated height and velocity well across most regions, discrepancies emerged when the height of the last tank approached the pipe. Notably, the PINN module struggles to accurately predict physical phenomena, particularly in scenarios involving singularities. We believe that our benchmarking study of PINN modules against the MELCOR CVH/FL package is very useful for examining its feasibility in severe accident analysis.

Accurate prediction of thermodynamic properties near the critical point is crucial for safety analysis in nuclear reactors, especially during severe accidents involving steam explosions. Existing methods face challenges in this region due to rapid and nonlinear changes in physical properties, leading to numerical instability and unreliable results. To address these limitations, we developed a deep learning-based standalone model that predicts physical properties near the critical point with high accuracy and computational efficiency. Utilizing data from the International Association for the Properties of Water and Steam (IAPWS), the model is trained to take specific internal energy and specific volume as inputs and outputs the corresponding pressure and temperature. The neural network employs a multilayer perceptron architecture with Leaky ReLU activation functions and is optimized using the mean squared error loss function and the Adam optimizer. Hyperparameter tuning, including adjustments to batch size and learning rate, was performed to enhance model performance. The developed model successfully captures the complex thermodynamic behavior near the critical point, overcoming the deficiencies of previous approaches. Integration of this model into the SCAR (Steam Explosion Code for Associated Risk) module, which is currently under development, enhances its predictive capabilities, providing more reliable inputs for severe accident analysis. This work demonstrates the potential of deep learning approaches in improving thermodynamic property predictions and paves the way for their application in other areas of nuclear safety analysis.

This study aims to reinforce safety measures for pressurized water reactors, ensuring more effective mitigation of severe accidents. Through uncertainty and sensitivity analyses, the research evaluates the effectiveness of reactor coolant system and steam generator injection strategies during an Extended Station Blackout scenario in the OPR-1000 nuclear reactor. Uncertainty analysis focuses on both code-related uncertainty parameters and the human reliability of executing time, critical factors that influence accident mitigation. Additionally, sensitivity analysis is performed to examine the injection rate of the Multi-barrier Accident Coping Strategies equipment, providing insights into the optimization of mobile equipment performance. The research evaluates the effects of these variables on key outcomes, including core cooling, reactor coolant system depressurization, and integrity of the reactor vessel. Utilizing the MAAP5 code, the study provides relevant data to enhance Severe Accident Management Guidance and improve accident management strategies.

As an integrated SMR, IP200 has the advantages of compact structure and high safety, and can be applied to floating nuclear power platforms through certain improved designs. The inherent characteristics and safety facility design of IP200 make its accident sequence slightly different from that of land-based PWR. The complete accident process from the initiation of the accident to the early occurrence of the reactor phenomenon, and then to the IVR and even the reactor reaction after the pressure vessel damage in the late stage of the serious accident, as well as the thermal and hydraulic effects of the safety facility input, are worth further research. A complete and detailed simulation model including the main coolant system and safety facilities of severe accident is established based on the mechanical severe accident analysis program Melcor and the integrated PWR thermal model, and the floating nuclear power platform IP200 is taken as the research object. The research results indicate the complete accident development sequence, key physical response characteristics of the core, and response characteristics of thermal and hydraulic parameters inside and outside the floating nuclear power platform under the condition of DVI pipeline rupture accidents before and after the failure of safety facilities, verifying the effectiveness of safety facility design.

One of the challenges in severe accident evaluation of Generation IV Sodium-cooled Fast Reactors (SFR) is the eutectic reaction between boron carbide (B4C) and stainless steel (SS), leading to boron migration in a molten pool within the core, which increases neutron absorption. To investigate this phenomenon, high-resolution radiative heating was employed to observe boron migration, eutectic behaviour, and melt structure. Experiments replicating control rod designs were conducted using B4C pellets in SS tubes at temperatures up to 1372°C. Two melting mechanisms were identified: SS separating from the B4C pellet and forming a melt drop, and B4C pellets fracturing due to thermal stress The use of visualisation techniques allowed for the detection of the eutectic onset, and the resulting eutectic melt was further analysed using material characterisation techniques. X-ray diffraction (XRD) and X-ray photoelectron spectroscopy (XPS) confirmed the formation of metal borides and metal carbides, attributed to the high chromium, iron, and carbon content. The current paper's findings confirm the relocation of the B4C-SS eutectic mixture and the formation of diverse boride phases, conditions likely to occur under extreme reactor conditions.