A submodel method was proposed that works from computational models of marine gear cases to verify that the proposed bolts will give it sufficient structural integrity. Calculations for marine equipment using this sys...A submodel method was proposed that works from computational models of marine gear cases to verify that the proposed bolts will give it sufficient structural integrity. Calculations for marine equipment using this system accorded well with conventional results. As an example, an anti-shock computation was processed for a gear case, and the submodel was then employed to check the strength of individual components. The results showed that the gear case connecting structure can satisfy relative anti-shock requirements, and the dynamic response characteristics seen in the bolt structures had a close relationship with the method used for attaching the bolt. This provides a new means for checking the strength of connecting structures on large-scale equipment and thus has significant reference value.展开更多
When a hydrogen storage vessel is subjected to a local impact load, damage may occur in the liner and result in hydrogen leakage and other catastrophic consequences. When predicting liner damage of a hydrogen storage ...When a hydrogen storage vessel is subjected to a local impact load, damage may occur in the liner and result in hydrogen leakage and other catastrophic consequences. When predicting liner damage of a hydrogen storage vessel using the finite element method(FEM), although large element size is required to achieve a desired computational efficiency, it oftentimes causes inaccuracy in the damage model. To remedy this problem, in this study a novel approach which calculates the material damage based on the GISSMO(Generalized Incremental Stress State dependent damage Model) damage model and employs a submodeling strategy is proposed. According to this approach,the global model is discretized to large elements to increase the efficiency, while the submodel is meshed to much smaller elements to accurately reflect the material damage. Employing the established approach and material parameters calibrated from a large set of notched aluminum alloy 5083 specimens, the liner damage of a type Ⅲ hydrogen storage vessel subjected to a local compressive load was simulated. This way, the study reveals how the characteristics of the stress and material damage interact with each other. In addition, the study also demonstrates that the proposed approach can be used as a viable means to evaluate the damage within hydrogen storage vessels.展开更多
为研究浮式生产储卸油装置(floating production storage and offloading,FPSO)浮筒关键节点的疲劳强度问题,本文基于FPSO三舱段、浮筒锥体、浮筒转塔、锥形对接模块(mating cone module,MCM)的整体模型,通过应力分析确定出应力较为集...为研究浮式生产储卸油装置(floating production storage and offloading,FPSO)浮筒关键节点的疲劳强度问题,本文基于FPSO三舱段、浮筒锥体、浮筒转塔、锥形对接模块(mating cone module,MCM)的整体模型,通过应力分析确定出应力较为集中的区域并作为热点,建立精细化网格的子模型以获得热点应力。基于力传递函数的方法,根据规范开展浮筒和MCM的疲劳简化分析,研究系泊和立管载荷对浮筒局部结构的疲劳影响。结果表明,浮筒锥体、浮筒转塔和MCM中关键节点的疲劳强度均满足设计疲劳强度要求,其中最薄弱的热点结构HS1的疲劳寿命为3262 a,大于20 a的设计寿命,满足FPSO单点系泊浮筒的设计预期要求。展开更多
传统联邦学习(FL)未考虑协作公平性,导致客户端获得的奖励与它的实际贡献不匹配。针对这一问题,提出一种基于个性化子模型和K均值聚类的联邦学习公平性算法(FedPSK)。首先,根据神经网络中神经元的激活模式对神经元聚类,且仅对聚类后的...传统联邦学习(FL)未考虑协作公平性,导致客户端获得的奖励与它的实际贡献不匹配。针对这一问题,提出一种基于个性化子模型和K均值聚类的联邦学习公平性算法(FedPSK)。首先,根据神经网络中神经元的激活模式对神经元聚类,且仅对聚类后的簇中心神经元进行重要性评估,并使用簇中心神经元的评分代表簇中其他神经元的评分,从而降低神经元评估的耗时;其次,使用层次选取方式选择客户端子模型中包含的神经元数量及编号,并为每个客户端建立具有完整神经网络结构的子模型;最后,通过为客户端下发子模型,实现协作公平性。在不同数据集上的实验结果表明,在公平度量的相关系数方面,FedPSK比FedSAC(Federated learning framework with dynamic Submodel Allocation for Collaborative fairness)提高了2.70%;在时间开销方面,FedPSK比FedSAC至少降低了84.12%。可见,FedPSK在提升FL算法公平性的同时,极大地降低了算法运行的时间开销,验证了所提算法的高效性。展开更多
基金Supported by the Shipbuilding Industry of National Defense Science and Technology Research Projects in Advance (07J1.5.3)
文摘A submodel method was proposed that works from computational models of marine gear cases to verify that the proposed bolts will give it sufficient structural integrity. Calculations for marine equipment using this system accorded well with conventional results. As an example, an anti-shock computation was processed for a gear case, and the submodel was then employed to check the strength of individual components. The results showed that the gear case connecting structure can satisfy relative anti-shock requirements, and the dynamic response characteristics seen in the bolt structures had a close relationship with the method used for attaching the bolt. This provides a new means for checking the strength of connecting structures on large-scale equipment and thus has significant reference value.
基金Supported by National Natural Science Foundation of China (Grant No. 52172353)。
文摘When a hydrogen storage vessel is subjected to a local impact load, damage may occur in the liner and result in hydrogen leakage and other catastrophic consequences. When predicting liner damage of a hydrogen storage vessel using the finite element method(FEM), although large element size is required to achieve a desired computational efficiency, it oftentimes causes inaccuracy in the damage model. To remedy this problem, in this study a novel approach which calculates the material damage based on the GISSMO(Generalized Incremental Stress State dependent damage Model) damage model and employs a submodeling strategy is proposed. According to this approach,the global model is discretized to large elements to increase the efficiency, while the submodel is meshed to much smaller elements to accurately reflect the material damage. Employing the established approach and material parameters calibrated from a large set of notched aluminum alloy 5083 specimens, the liner damage of a type Ⅲ hydrogen storage vessel subjected to a local compressive load was simulated. This way, the study reveals how the characteristics of the stress and material damage interact with each other. In addition, the study also demonstrates that the proposed approach can be used as a viable means to evaluate the damage within hydrogen storage vessels.
文摘为研究浮式生产储卸油装置(floating production storage and offloading,FPSO)浮筒关键节点的疲劳强度问题,本文基于FPSO三舱段、浮筒锥体、浮筒转塔、锥形对接模块(mating cone module,MCM)的整体模型,通过应力分析确定出应力较为集中的区域并作为热点,建立精细化网格的子模型以获得热点应力。基于力传递函数的方法,根据规范开展浮筒和MCM的疲劳简化分析,研究系泊和立管载荷对浮筒局部结构的疲劳影响。结果表明,浮筒锥体、浮筒转塔和MCM中关键节点的疲劳强度均满足设计疲劳强度要求,其中最薄弱的热点结构HS1的疲劳寿命为3262 a,大于20 a的设计寿命,满足FPSO单点系泊浮筒的设计预期要求。
文摘传统联邦学习(FL)未考虑协作公平性,导致客户端获得的奖励与它的实际贡献不匹配。针对这一问题,提出一种基于个性化子模型和K均值聚类的联邦学习公平性算法(FedPSK)。首先,根据神经网络中神经元的激活模式对神经元聚类,且仅对聚类后的簇中心神经元进行重要性评估,并使用簇中心神经元的评分代表簇中其他神经元的评分,从而降低神经元评估的耗时;其次,使用层次选取方式选择客户端子模型中包含的神经元数量及编号,并为每个客户端建立具有完整神经网络结构的子模型;最后,通过为客户端下发子模型,实现协作公平性。在不同数据集上的实验结果表明,在公平度量的相关系数方面,FedPSK比FedSAC(Federated learning framework with dynamic Submodel Allocation for Collaborative fairness)提高了2.70%;在时间开销方面,FedPSK比FedSAC至少降低了84.12%。可见,FedPSK在提升FL算法公平性的同时,极大地降低了算法运行的时间开销,验证了所提算法的高效性。
