This study proposes a multi-net trash barrier system to ensure water intake safety in nuclear power plants,comprising various planar nets and net bags supported by piles and anchor blocks.A comprehensive physical mode...This study proposes a multi-net trash barrier system to ensure water intake safety in nuclear power plants,comprising various planar nets and net bags supported by piles and anchor blocks.A comprehensive physical model experiment was conducted in a three-dimensional wave basin to investigate the mooring forces on this barrier.The research analyzes the effect of blockage levels on the dynamic responses of pile-supported and floating nets within blockage ratios of 10%−70%under current-only,wave-only,and wave-current conditions.For optimal net span design,the study examines curvatures ranging from 0.2%to 18.6%,representing the extension proportion of arc length compared to straight-line distance in a span.Results indicate that combined water intake and wave action effects generate substantial forces on the net barrier,requiring careful consideration during design.Higher blockage ratios and lower curvatures correspond to increased net barrier tensions.For practical engineering applications,the recommended curvature for floating net barriers ranges from 7%to 10%for optimal net arc length design.展开更多
为解决低轨遥感星座协同任务规划面临的计算复杂度高、通信开销大、动态响应能力弱等问题,提出一种基于任务聚类与禁忌搜索的改进合同网算法(improved contract net protocol based on task clustering and tabu search,CN-TCTS).该算...为解决低轨遥感星座协同任务规划面临的计算复杂度高、通信开销大、动态响应能力弱等问题,提出一种基于任务聚类与禁忌搜索的改进合同网算法(improved contract net protocol based on task clustering and tabu search,CN-TCTS).该算法采用“单星调度-全局分配”的分层求解框架.首先,通过任务聚类合并元任务,有效缩减解空间;其次,设计了动态约束禁忌搜索算法(dynamic constraint tabu search,DCTS),通过价值导向的邻域搜索策略实现单星任务序列的快速规划;最后,在全局分配阶段,引入多种策略对传统合同网协议进行改进,实现任务的高效分配与冲突消解.仿真结果表明,本文所提CN-TCTS算法在400个任务的大规模场景下,任务完成率仍保持82.0%,且平均通信轮次仅为6.6轮.此外,在卫星突发失效的动态场景下,该算法表现出更强的鲁棒性,收益损失率更低.此外,局部规划算法仿真中验证了DCTS算法在收敛速度与解质量方面的优势.展开更多
以Mamba为代表的状态空间模型(state space models,SSM)凭借其出色的长程依赖建模能力与较低的线性计算复杂度,在医学图像分割领域展现出广阔的应用前景。但该类方法对图像逐像素展平的处理方式会破坏图像空间结构,导致局部细节提取不足...以Mamba为代表的状态空间模型(state space models,SSM)凭借其出色的长程依赖建模能力与较低的线性计算复杂度,在医学图像分割领域展现出广阔的应用前景。但该类方法对图像逐像素展平的处理方式会破坏图像空间结构,导致局部细节提取不足,且缺乏对器官与病灶的聚焦机制,在复杂背景下易引入冗余背景信息。为解决上述问题,本文提出了一种融合谱域动态特征与注意力机制的医学图像分割网络——频域动态注意力U型Mamba(spectral dynamic attention U-Mamba,SDA U-Mamba)。该网络采用分层U型结构设计,分别从空间连续性建模和区域聚焦能力两方面对Mamba进行优化。在网络浅层设计Mamba空频注意力模块,通过融合空域卷积、频域变换和金字塔自注意力结构以提升模型的局部信息感知能力与多尺度上下文建模效果;在网络深层引入双极路由注意力模块,通过动态路由选择与稀疏激活机制增强模型对医学图像器官或病灶的表征。实验结果表明,SDA U-Mamba在BUSI、CVCClinicDB与CHAOS-Liver这3个公开医学数据集上的分割性能显著优于当前主流方法,平均交并比(intersection over union,IoU)提升2.61%。本文所提算法可用于临床医学图像分割。展开更多
基金financially supported by the Foundation for Research Development Project of China Communications Construction Company(Grant No.2020-ZJKJ-PTJS08)Research Development Project of CCCC First Harbor Engineering Company(Grant No.20230612).
文摘This study proposes a multi-net trash barrier system to ensure water intake safety in nuclear power plants,comprising various planar nets and net bags supported by piles and anchor blocks.A comprehensive physical model experiment was conducted in a three-dimensional wave basin to investigate the mooring forces on this barrier.The research analyzes the effect of blockage levels on the dynamic responses of pile-supported and floating nets within blockage ratios of 10%−70%under current-only,wave-only,and wave-current conditions.For optimal net span design,the study examines curvatures ranging from 0.2%to 18.6%,representing the extension proportion of arc length compared to straight-line distance in a span.Results indicate that combined water intake and wave action effects generate substantial forces on the net barrier,requiring careful consideration during design.Higher blockage ratios and lower curvatures correspond to increased net barrier tensions.For practical engineering applications,the recommended curvature for floating net barriers ranges from 7%to 10%for optimal net arc length design.
文摘为解决低轨遥感星座协同任务规划面临的计算复杂度高、通信开销大、动态响应能力弱等问题,提出一种基于任务聚类与禁忌搜索的改进合同网算法(improved contract net protocol based on task clustering and tabu search,CN-TCTS).该算法采用“单星调度-全局分配”的分层求解框架.首先,通过任务聚类合并元任务,有效缩减解空间;其次,设计了动态约束禁忌搜索算法(dynamic constraint tabu search,DCTS),通过价值导向的邻域搜索策略实现单星任务序列的快速规划;最后,在全局分配阶段,引入多种策略对传统合同网协议进行改进,实现任务的高效分配与冲突消解.仿真结果表明,本文所提CN-TCTS算法在400个任务的大规模场景下,任务完成率仍保持82.0%,且平均通信轮次仅为6.6轮.此外,在卫星突发失效的动态场景下,该算法表现出更强的鲁棒性,收益损失率更低.此外,局部规划算法仿真中验证了DCTS算法在收敛速度与解质量方面的优势.
文摘以Mamba为代表的状态空间模型(state space models,SSM)凭借其出色的长程依赖建模能力与较低的线性计算复杂度,在医学图像分割领域展现出广阔的应用前景。但该类方法对图像逐像素展平的处理方式会破坏图像空间结构,导致局部细节提取不足,且缺乏对器官与病灶的聚焦机制,在复杂背景下易引入冗余背景信息。为解决上述问题,本文提出了一种融合谱域动态特征与注意力机制的医学图像分割网络——频域动态注意力U型Mamba(spectral dynamic attention U-Mamba,SDA U-Mamba)。该网络采用分层U型结构设计,分别从空间连续性建模和区域聚焦能力两方面对Mamba进行优化。在网络浅层设计Mamba空频注意力模块,通过融合空域卷积、频域变换和金字塔自注意力结构以提升模型的局部信息感知能力与多尺度上下文建模效果;在网络深层引入双极路由注意力模块,通过动态路由选择与稀疏激活机制增强模型对医学图像器官或病灶的表征。实验结果表明,SDA U-Mamba在BUSI、CVCClinicDB与CHAOS-Liver这3个公开医学数据集上的分割性能显著优于当前主流方法,平均交并比(intersection over union,IoU)提升2.61%。本文所提算法可用于临床医学图像分割。
