Song Li

The detection of sea mines in side-scan sonar imagery is hindered by small object size, heavy speckle noise, and severe data scarcity. Existing studies mostly transplant improvements designed for natural images to sonar scenarios, and their effectiveness under small-sample conditions lacks unified verification. This study presents a systematic benchmark of lightweight object detectors on a publicly available side-scan sonar mine-like object dataset, SSS-MLOD. Under a unified protocol, five YOLO variants are compared under from-scratch training and COCO-pretrained transfer learning, and 25 structural, loss-based, and data-side modifications are evaluated. The results show that differences among lightweight baselines are marginal (mAP50 of 0.732 to 0.753), transfer learning on only 243 training images yields the largest single gain (mAP50 from 0.742 to 0.856 for YOLOv13n), and none of the 25 modifications was observed to improve the baseline under from-scratch training on this dataset. Increasing the input resolution improves mAP50:95 by 0.044 to 0.049 for YOLOv5n and YOLOv8n but degrades the performance of YOLOv11n, indicating that the benefit of resolution is architecture-dependent; test-time augmentation with a tuned confidence threshold raises recall by 0.035 under a precision constraint of 0.85. The best configuration, YOLOv13n with transfer learning, achieves an mAP50 of 0.856 and a recall of 0.757. These findings provide an evidence-based reference for detector selection and configuration in resource constrained applications such as autonomous underwater vehicles.

The Random Ray Method (TRRM) inherits the well-established batch statistics framework of Monte Carlo codes and reports statistical uncertainties under the assumption that the batch samples are independent and identically distributed. In the iteration, however, the source of the next batch is generated by normalizing the flux estimate of the previous one, so inter-batch correlations are dictated by the iteration mechanism. Whether the statistical uncertainties reported by the code are trustworthy has therefore not previously been examined systematically. This work investigates the random ray solver of OpenMC through multi-seed measurements and mechanistic analyses on six cases, and develops a correction based on the effective sample size. The measurements show that the distortion factor of the statistical uncertainties ranges from 0.67 to 5.27, with both underestimation and conservatism present. The empirical coverage of the nominal 95% confidence intervals for the 𝑘eff series drops to as low as 32.0%. In the slab case examined in this work, the distortion of the tallies isabout twice that of the 𝑘ef f series. Mechanistically, what is propagated through the source distribution is the ray sampling noise of the previous batch itself. The degree of distortion can be characterized by the autocorrelation structure of the batch series itself. The dominance-ratio-based predictor from Monte Carlo, however, can either overestimate or underestimate the distortion in TRRM, with no consistent direction of bias. Within the two cases examined and over the range of rays per batch examined, the integrated autocorrelation time does not depend on the number of rays per batch, but is a constant determined by the structure of the problem. An effective-sample-size correction based on Geyer’s initial positive sequence estimator restores the interval coverage of the 𝑘ef f series in the underestimating cases to 90.0%–96.0%. All diagnostics and corrections are implemented as standalone post-processing scripts and apply directly to the output files of any OpenMC TRRM calculation.

The lead-cooled fast reactor (LFR), a Generation IV nuclear system candidate, presents unique neutronic characteristics distinct from pressurized water reactors. Its neutron spectrum spans wider energy ranges with fast neutron dominance, exhibiting resonance phenomena across energy regions. These features require a fine energy group structure for fuel lattice calculations, significantly increasing computational demands. To balance local heterogeneity modeling with computational efficiency, researchers across the world adopt fuel assembly equivalence methods using 1D cylindrical models through volume equivalence principles. This approach enables detailed energy group calculations in simplified geometries, followed by lattice homogenization for few-group parameter generation, effectively reducing whole-core computational loads. However, limitations emerge when handling strongly heterogeneous components like structural/control rods. This study investigates the 1D equivalence method’s accuracy in lead–bismuth fast reactors under various fuel assembly configurations. Through comprehensive analysis of material distributions and their equivalence impacts, the applicability of the one-dimensional equivalence approach to fuel assemblies of different geometries and material types is analyzed in this paper. The research further proposes corrective solutions for low-accuracy scenarios, enhancing computational method reliability. This paper is significant in its optimization of the physical calculation and analysis process of a new type of fast reactor component and has important engineering application value.

Motivated by the growth of global energy demand and the goal of carbon neutrality, lead-cooled fast reactors, which are core reactor types of fourth-generation nuclear energy systems, have become a global research hotspot due to their advantages of high safety, nuclear fuel breeding capability, and economic efficiency. However, its engineering implementation faces key challenges, such as material compatibility, closed fuel cycles, and irradiation performance of structures. This paper comprehensively reviews the latest progress in the core design of lead-cooled fast reactors in terms of the innovation of nuclear fuel, optimization of coolant, material adaptability, and design of assemblies and core structures. The research findings indicate remarkable innovation trends in the field of lead-cooled fast reactor core design, including optimizing the utilization efficiency of nuclear fuel based on the nitride fuel system and the traveling wave burnup theory, effectively suppressing the corrosion effect of liquid metal through surface modification technology and the development of ceramic matrix composites; replacing the lead-bismuth eutectic system with pure lead coolant to enhance economic efficiency and safety; and significantly enhancing the neutron economy and system integration degree by combining the collaborative design strategy of the open-type assembly structure and control drums. In the future, efforts should be made to overcome the radiation resistance of materials and liquid metal corrosion technology, develop closed fuel cycle systems, and accelerate the commercialization process through international standardization cooperation to provide sustainable clean energy solutions for basic load power supply, high-temperature hydrogen production, ship propulsion, and other fields.

Small nuclear power reactors have small core dimensions, frequent power changes, and more severe power distortion compared to nuclear power stations. However, their core has fewer measurement points, making it difficult to observe their core power distribution. High-precision physical calculation programs can accurately calculate the core power distribution, but the real-time performance of the calculation is poor, which is not conducive to online use. In this study, based on physical computing programs, the power distribution spectrum library of small nuclear power reactors under different operating conditions is calculated, and artificial intelligence algorithms are designed. A data-driven model for the proxy relationship between operating state parameters and core power distribution is trained and constructed to achieve rapid calculation and online support of core power distribution, which improves the level of online safety supervision of small power reactors. Numerical experiments show that this method has high accuracy and good robustness, and can meet the requirements of small nuclear power reactor operation safety support. This research is based on a data-driven proxy model and has achieved fast computation of power distribution in the fuel cores of small modular reactors. It addresses the issue of insufficient real-time performance of high-precision physical programs and has important significance for the safe operation of reactors.

Traditional computational methods for pressurized water reactors are unable to handle dispersed fuel particles as the double heterogeneity and the direct volumetric homogenization can result in significant errors. In contrast, reactivity-equivalent physical transformation techniques offer high precision for addressing the double heterogeneity introduced by dispersed fuel particles. This approach converts the double heterogeneity problem into a single heterogeneity problem, which is then subsequently investigated by using the conventional pressurized water reactor computational procedure. However, it is currently empirical and takes a lot of time to obtain the right k_∞. In this paper, we train the RPT model by using the existing dataset of plate-dispersed fuel and rod-dispersed fuel by a machine learning method based on a linear regression model, and we then use the new data to make predictions and derive the corresponding similarity ratios. The burnup verification, density verification, fission rate verification, and neutron energy spectrum analysis are calculated through the OpenMC program. For plate-type fuel elements, the method maintains an accuracy within 200 pcm during depletion, with deviations in the ²³⁵U density and ²³⁵U fission rate within 0.1% and neutron energy spectrum errors within 6%. For rod-type fuel elements, the method maintains an accuracy within 100 pcm during depletion, with deviations in ²³⁵U and ²³⁹Pu density within 1.5% and neutron energy spectrum errors within 1%. The numerical validation indicates that the reactivity-equivalent physical transformation method based on the linear regression model not only greatly improves the computational efficiency, but also ensures a very high accuracy to deal with double heterogeneity in nuclear reactors.

According to the construction requirements and structural characteristics of the water-purifying constructed wetland in Qiangwei Lake, this research clarifies the connotation of the constructed wetland's health, and analyzes the relevant factors affecting the operation effect of the wetland. Meanwhile, based on the principle of wetland health evaluation, this research applies the analytic hierarchy process to divide the health evaluation system of the water-purifying constructed wetland in Qiangwei Lake into the goal layer, the criterion layer, and the index layer. Furthermore, by selecting a host of evaluation indexes, encompassing water quality indexes, wetland structure, biodiversity, etc., this research discusses and establishes the corresponding index evaluation standard, ultimately constructing the comprehensive scoring system of the wetland.

The fine-mesh subgroup method (FSM) is proposed to treat the significant resonance self-shielding effect both effectively and accurately. Similar to the ultra-fine group method, the fine-mesh subgroup method adopts a fine group structure on the resonance energy range to avoid the extra resonance interference effect correction. To improve the efficiency, on the one hand, the one-group micro-level optimization is adopted, so the subgroup fixed-source equations will be only calculated on a certain number of pre-determined subgroup levels, and an interpolation process is employed to obtain the actual subgroup flux. On the other hand, the slowing-down calculation is carried out for group condensation for multigroup transport calculation. The main theory and feasible improvements of the fine-mesh subgroup method are analyzed in this paper. Several pin cell and lattice problems are applied to test the performance of the fine-mesh subgroup method, and the particle swarm optimization method is adopted to find the better group structure. The numerical results indicate a good performance both for accuracy and efficiency.