Fault diagnosis for small modular reactor (SMR) faces challenges such as limited fault sample quantities and variations in data distribution across different operating conditions, which constrain the performance of cross-domain diagnosis models. To address these issues, an improved fault diagnosis method is presented based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network with a Squeeze-and-Excitation (SE) attention mechanism to enable SMR cross-domain diagnosis performance. To enhance the adaptability of the fault diagnosis model under variable power levels, multiple CNN-LSTM-SE transfer learning strategies by using different parameter fine-tuning approaches are designed and evaluated. Furthermore, to mitigate the scarcity of target domain samples, a particle swarm optimization and support vector regression (PSO-SVR) model is used to generate representative synthetic fault samples for data augmentation. These synthetic fault samples are combined with real data to promote adaptive capability of the transfer learning model in the target domain. Gaussian noise is introduced to simulate actual sensor noise environments, and a wavelet denoising preprocessing strategy is integrated, which enhances the model’s engineering applicability of actual SMR plant. The calculation results demonstrate that the proposed fault diagnosis framework has higher diagnosis accuracy and better generalization performance than other deep learning approaches under scarce fault samples and cross-domain diagnosis conditions.

With the rapid advancement of artificial intelligence (AI), intelligent diagnostics have seen broad application across industries. To address the limitations of traditional data-driven methods in accurately identifying faults in nuclear power reactor systems, this study proposes a hybrid model combining Transformer and XGBoost for diagnosing faults in CPR1000 pressurized water reactors. Fault-related data were automatically collected using the self-developed AutoSave-PCTRAN software from the PCTRAN simulator, with key nuclear parameters selected as features. A 10-fold cross-validation with recursive feature elimination was used for feature selection. The Transformer model extracted temporal features via its self-attention mechanism, and an RWOA was employed to tune XGBoost hyperparameters for fault classification. The model effectively identified faults such as Loss of Coolant Accident(LOCA), Steam Line Break Inside Containment(SLBLC), and steam generator B-tube rupture(SGTR-B), achieving 99.46% accuracy, confirming its reliability and practicality.

Fast reactors are crucial for promoting the sustainable development of fission energy. However, high-fidelity simulation of effective delayed neutron fraction(/3eff) in fast reactors are constrained by nuclear data uncertainties, which directly affects the reliability of safety assessment and design optimization. Conducting sensitivity and uncertainty(S/U) analyses on /3eff enables scientific quantification of system safety margins and guides the refinement of nuclear data. This study focuses on the S/U of /3eff in fast reactors. First, covariances were generated based on the ENDF/B-VIII.1, JENDL-5.0, and JEFF-3.3 nuclear data libraries. Subsequently, a S/ U analysis code for /3eff was developed using first-order perturbation theory. Finally, systematic analyses were performed on representative fast reactors(ABTR, MET-1000 and CiADS). The results show significant discrepancies in /3eff uncertainty obtained from different covariances, ranging from 1% to 4%. Compared to JENDL-5.0, the use of covariances from ENDF/B-VIII.1 and JEFF-3.3 leads to a considerable underestimation of the /3eff uncertainty. The main sources of uncertainty originate from delayed fission neutron yields(ud) and prompt fission neutron yields(up), with major contributions concentrated in the keV to MeV energy region.

Metamorphic testing is a valuable approach for addressing the oracle problem, with the identification of metamorphic relations being a crucial task. Despite the availability of metamorphic relations in previously studied programs, many current studies do not leverage them, resulting in inefficiencies and reliability challenges when identifying metamorphic relations in new programs. SimiMR proposes the recommendation of verified metamorphic relations for new code by leveraging similarities with existing programs. This approach is based on the assumption that akin programs exhibit akin metamorphic relations, thus linking program similarity with metamorphic relations classification: semantic similarity corresponds with physical and computational model metamorphic relations, whereas syntactic similarity corresponds with code model ones. Experiments demonstrate that metamorphic relations in code models are effectively reusable among programs with similar syntax, whereas those in physical or computational models can be applied to semantically similar programs. SimiMR surpasses AutoMR and similar methods by enhancing identification efficiency, extending applicability, and minimizing redundancy. It operates with minimal domain knowledge, leverages existing metamorphic relations, and keeps identification costs low.

Dose calculation is the foundation of boron neutron capture therapy (BNCT). MagicDose, a dose calculation program for the BNCT treatment planning system, is developed based on the Monte Carlo method. First, the voxel phantom of the modified Snyder head with 16 and 8 mm is constructed, and the results from MagicDose and MCNP are presented as two-dimensional coordinate points (Xn, Yn), comparing their relationship relative to the y = x linear function, while analyzing their respective calculation time. A modified Snyder head phantom with a tumor at three different spatial resolutions of 16, 8, and 1 mm was constructed, and the depth-dose rate curves and spatial distribution maps are analyzed. Finally, the patients’ head CT data were used for the application. The results indicate that the calculations from MagicDose and MCNP exhibit high consistency and demonstrate that MagicDose offers superior computational efficiency compared to MCNP, with improvements of approximately 31.24

Dose calculation is the foundation of boron neutron capture therapy (BNCT). MagicDose, a dose calculation program for the BNCT treatment planning system, is developed based on the Monte Carlo method. First, the voxel phantom of the modified Snyder head with 16 and 8 mm is constructed, and the results from MagicDose and MCNP are presented as two-dimensional coordinate points (Xn , Yn ), comparing their relationship relative to the y=x linear function, while analyzing their respective calculation time. A modified Snyder head phantom with a tumor at three different spatial resolutions of 16, 8, and 1 mm was constructed, and the depth-dose-rate curves and spatial distribution maps are analyzed. Finally, the patients’ head CT data were used for the application. The results indicate that the calculations from MagicDose and MCNP exhibit high consistency anddemonstrate that MagicDose offers superior computational efficiency compared to MCNP, with improvements of approximately 31.24% and 28.65% at spatial resolutions of 16 and 8 mm, respectively. As the spatial resolution increased, the variability in the dose rate results decreased. The voxel size and number of threads are both inversely proportional to the calculation time. For the CT model, a voxel phantom with a spatial resolution of 1mm×1 mm×1 mm is successfully constructed. The calculation results showed that the boron dose rate contribution significantly exceeds that of the other dose components, with the spatial distribution of the total relative biological effect dose rate clearly delineating the boundaries between the high- and low-dose-rate regions. The above results verify the correctness of MagicDose, which also provides a reference for optimizing the design ofvoxel phantoms for clinical treatment.

Boron neutron capture therapy (BNCT) has progressively gained recognition as a primary modality for cancer treatment, attributed to its advantages of precise targeting, minimal adverse effects, and expedited recovery periods. However, the dynamic quantitative assessment of boron concentration remains a formidable challenge in evaluating the efficacy of BNCT. Conventional methodologies, such as single-point immersion sampling and in vivo blood concentration inversion, are inadequate for monitoring the dynamic fluctuations of boron concentration within tumor tissues. Consequently, the precise administration of neutron and boron doses is compromised, thereby affecting the quality of treatment. This study proposes the utilization of single-photon emission computed tomography (SPECT) as a technique for non-invasive, in vivo, real-time quantitative evaluation of boron concentration during BNCT, facilitating dynamic monitoring. Investigations into the efficiency of detector crystals for the 478 key prompt gamma rays emitted during BNCT indicate that Gadolinium Aluminum Gallium Garnet (GAGG) crystals exhibit superior detection efficiency and meet precision requirements. Through a rational assessment of collimator penetration requirements, the dimensions of the collimator were established at 4.5 & times; 4.5 & times; 78 mm3. The spatial resolution obtained via reconstruction with the Ordered Subset Expectation Maximization (OSEM) algorithm was 6-8 mm, effectively meeting the clinical detection standards and enhancing the accuracy of BNCT treatment.

The interaction and feedback between 3D neutronics and thermal hydraulics are of great significance in reactor safety analyses, particularly for the TRIGA reactor. Owing to the TRIGA reactor’s pulse-transient operation status, the power changes by six to eight orders of magnitude within an extremely short duration; this operation is significantly different from PWRs and imposes some challenges for conventional neutronics methods. To describe the transient status of rod insertion or withdrawal, a novel time-dependent particle transport algorithm based on the combined and moving geometry methods is developed and integrated into the neutronics code MagicMC, which is a Monte Carlo particle transport code developed by the Nuclear Energy and Application Laboratory. Combined with the subchannel model, this work presents neutronics and thermal-hydraulics coupling methods for high-fidelity simulation of the TRIGA reactor. First, a steady-state coupling method is established based on over-relaxation iteration, and the number of neutrons in the Monte Carlo simulation is adaptively controlled according to convergence. Subsequently, a transient coupling method is proposed based on the semi-implicit coupling strategy, and a dynamically changing time-step strategy is designed for the coupling iterative process to achieve reasonable convergence. The parameter mapping strategy between neutronics and thermal hydraulics was constructed using one-to-one mapping and volume weight methods. To verify the reliability of the methods, a JSI TRIGA Mark II reactor was selected as the validation benchmark. The coupling results were in good agreement with the experimental data of the JSI TRIGA Mark II reactor, and the coupling methods achieved a high-fidelity numerical simulation of TRIGA reactor. Therefore, the coupling methods proposed in this paper can provide technical support for reactor experiments and the safe operation of the TRIGA reactor.

Small Modular Reactors (SMRs) are widely used to supply energy in special scenarios such as remote mountainous areas, oceanic scientific expeditions, and space exploration due to their flexible installation characteristics. A critical challenge to address is designing a miniaturized, lightweight radiation shielding system that ensures dual-stage radiation safety during both reactor operation and shutdown, allowing nuclear devices to adapt to a variety of complex environments. To address these challenges, this paper proposes a dual-stage radiation-shielding optimization design (DROD) method. DROD integrates evolutionary algorithms with the Monte Carlo method under massive parallelism to optimize radiation-shielding designs for both reactor operation and shutdown stage. Additionally, a multi-objective evolutionary algorithm based on hypervolume guidance and reference-point association (HV-RP-MOEA) is proposed as a solution to the expensive constrained multi-objective optimization problem in radiation shielding. The performance of HV-RP-MOEA was verified through multi-objective test problems, demonstrating its fast convergence and multi-objective optimization capabilities. Furthermore, DROD was applied to optimize the radiation-shielding design of the small pressurized water reactor KLT-40. The results indicate that DROD can efficiently explore a wide range of shielding solutions, outperforming conventional multi-objective radiation-shielding optimization methods in terms of both depth and breadth of optimization. This work provides new insights into the optimization design of radiation-shielding.

The absence of a universal and effective Global Variance Reduction (GVR) methodology poses significant challenges for accurate and efficient radiation field calculations in large-scale nuclear facilities using Monte Carlo methods. This study introduces a novel Space self-adaption Global Variance Reduction(SsGVR) method for large-scale radiation field simulations. The method dynamically adjusts the lower limit of the window weight during the iteration process by statistically considering the empty mesh rate and the average relative error. Using the NUREG-CR-6115 1/4 core benchmark problem, it was found that the computational efficiency of the SsGVR method was improved by more than three orders of magnitude compared to standard Monte Carlo simulation. Beyond its computational advantages, the SsGVR method exhibits high versatility and robustness, leveraging a straightforward approach to construct global information that circumvents intricate mathematical derivations or geometric modeling. These attributes highlight its substantial potential for advanced applications in large-scale radiation analysis.

During nuclear reactor operations, neutron activation reactions generate significant quantities of radionuclides from the structural materials, directly impacting shielding design, maintenance planning, and decommissioning strategies. This is a critical component of radiation safety analysis. As advanced nuclear reactor technology evolves, the increasing complexity of reactor geometry, material configurations, and neutron spectra complicates activation analysis. Consequently, there is a pressing need for high-resolution activation analysis of nuclear reactors. This paper presents a high-resolution activation analysis method for large-scale complex structural materials, utilizing the Monte Carlo global variance reduction particle transport technique. A fully automated coupled high-resolution activation analysis program is developed, enabling the calculation of high-resolution decay source distributions and precise evaluation of decay photon sources for extensive complex structural materials. The methodology is benchmarked against the shutdown dose rate benchmark released by the International Thermonuclear Experimental Reactor (ITER) program. Additionally, an application study of highresolution activation analysis is conducted on a standard pressurized water reactor (PWR). The methodology demonstrated in this paper holds significant engineering value, enhancing the accuracy of activation calculations for large-scale nuclear reactor structural materials and it provides guidance for optimizing shielding design to reduce radiation exposure during operation.

Boiling heat transfer has long been a promising thermal management technique across various industrial sectors. In the context of the energy crisis, recovering low-grade thermal energy from boiling environments is crucial for industrial advancement. Meanwhile, triboelectric nanogenerators (TENGs) are notable for low-grade energy harvesting due to their simple fabrication, adaptability to low mechanical frequencies, and lightweight nature. This study introduces a magnet gear-based TENG, designed to harness energy from boiling environments. The magnet gear facilitates remote force transmission from the rotor to the TENG module, effectively mitigating the adverse effects of high temperature and humidity. Rabbit fur is used to create a soft contact, enhancing coupling with the low-grade energy source, significantly reducing frictional resistance, and improving device durability. Tests demonstrate that the M-TENG does not impact the thermal management of boiling heat transfer, showing potential for integration with other thermal energy harvesting techniques. Its operational range spans the boiling curve from nucleation to critical heat flux, maintaining performance after 15,000 cycles. Furthermore, selfpowered applications have been successfully demonstrated, such as powering small calculators, lighting at least 44 LEDs, and monitoring boiling heat flux, offering a viable strategy for industrial upgrading in boiling heat transfer scenarios.

The sustained development of global nuclear energy underscores the importance of nuclear decommissioning as a critical measure for environmental and public safety. This article systematically reviews recent advances in remediation technologies for nuclear facilities. Significant progress has been made in surface decontamination techniques, including mechanical, chemical, laser, and electrochemical methods, as well as novel stripping materials, with approaches such as laser ablation achieving removal rates exceeding 99%. The integration of intelligent systems, such as robotic operations and AI-driven monitoring, has substantially improved operational safety and efficiency. The study also summarizes relevant regulations pertaining to safety management, radioactive waste management, and nuclear protection. While current remediation technologies show a trend toward integrated multi-method solutions, challenges remain in areas such as complex structure decontamination and long-lived radionuclide immobilization. Future developments are directed toward smarter, more sustainable approaches, where international cooperation will be crucial in advancing nuclear decommissioning and nuclear protection capabilities.

Dose calculation is the foundation of Boron Neutron Capture Therapy (BNCT). MagicDose, a dose calcula- tion program for the BNCT treatment planning system, is developed based on the Monte Carlo method. Firstly, the voxel phantom of the modified Snyder head with 16, 8 mm are constructed, and the deviation of each result on y=x and the calculation time are statistical.The modified Snyder head phantom with tumor at three different spatial resolutions of 16, 8, and 1 mm are constructed, and the depth-dose-rate curves and spatial distribution maps are analyzed. Finally, the patient's head CT data is used as an application. The results show that the results calculated by MagicDose and MCNP are in good consistency, demonstrating that the computational efficiency of MagicDose is also better than that of MCNP. As the spatial resolution increases, the variability of the dose rate results is smaller. The voxel size and the number of threads are both inversely proportional to the time. For the CT model, the voxel phantom is successfully constructed and the calculation results are reasonable. The above results verify the correctness of MagicDose, which also provides a reference for optimizing the design of the voxel phantom in clinical treatment.

In recent years, the development of new types of nuclear reactors, such as transportable, marine, and space reactors, has presented new challenges for the optimization of reactor radiation-shielding design. Shielding structures typically need to be lightweight, miniaturized, and radiation-protected, which is a multi-parameter and multi-objective optimization problem. The conventional multi-objective (two or three objectives) optimization method for radiation-shielding design exhibits limitations for a number of optimization objectives and variable parameters, as well as a deficiency in achieving a global optimal solution, thereby failing to meet the requirements of shielding optimization for newly developed reactors. In this study, genetic and artificial bee-colony algorithms are combined with a reference-point-selection strategy and applied to the many-objective (having four or more objectives) optimal design of reactor radiation shielding. To validate the reliability of the methods, an optimization simulation is conducted on three-dimensional shielding structures and another complicated shielding-optimization problem. The numerical results demonstrate that the proposed algorithms outperform conventional shielding-design methods in terms of optimization performance, and they exhibit their reliability in practical engineering problems. The many-objective optimization algorithms developed in this study are proven to efficiently and consistently search for Pareto-front shielding schemes. Therefore, the algorithms proposed in this study offer novel insights into improving the shielding-design performance and shielding quality of new reactor types.

Small pressurized water reactor (PWR) usually operates under flexible conditions with rapid power variations, the resulting uncertainty in the core nonlinear model and core parameters cannot be ignored, which is a main challenge for the optimal design of the core power control system and the safety operation of the reactor. This paper designs a core power flexible switching (CPFS) control system of small PWR to achieve full power range control under arbitrary initial power conditions. For quantitative assessment of the robustness of the CPFS control system, based on the normal cloud model and the Bootstrap Interval Estimation (Cloud-BIE) method, a new uncertainty quantification method with the Cloud-BIE framework is proposed, which is suitable in the case of small sample size and unknown parameter distribution function. In the 5% uncertainty of reactivity temperature coefficients under reactivity disturbances on the CPFS control system. The uncertainties of all output parameters are within the engineering safety limit, the results show that the Cloud-BIE framework is suitable for evaluating the robustness of the control system.

The highly nonlinear spatiotemporal evolution characteristics of short-lived products during the course of water radiolysis render their prediction of concentration distribution immensely important with respect to radiation chemistry investigations and nuclear safety assessments. Thus, this paper proposes a spatiotemporal prediction model of water radiolysis product concentrations based on Swin Transformer and a hierarchical window attention mechanism. Keeping the advantages of the sliding window mechanism in the original design of the Swin structure, a temporally enhanced window self-attention module is constructed so that long-term dependency features in the products evolution process can be recognized. Subsequently, then, a multi-scale reaction-diffusion embedding module is designed to model the micro-to-macro concentration gradients effectively and rapidly. Finally, a physics-informed positional decoder is introduced to add physics consistency and predictive stability to the attention mechanism. The experimental results show that on particular benchmark data sets, this model gives better results in predictions than traditional methods with significant qualitative improvements in accuracy and also in temporal generalization capability.