In order to analyze the hydrodynamic performance of the ducted propeller with high precision, this paper proposes a new method which combines Multi-Block Hybrid Mesh and Reynolds Stress Model (MBHM & RSM). The cal...In order to analyze the hydrodynamic performance of the ducted propeller with high precision, this paper proposes a new method which combines Multi-Block Hybrid Mesh and Reynolds Stress Model (MBHM & RSM). The calculation errors of MBHM & RSM and standard two-equation model (standard k-ε model) on the ducted propeller JD7704 +Ka4-55 are compared. The maximum error of the total thrust coefficient KT, the duct thrust coefficient KTN, the torque coefficient KQ and the open-water efficiency η0 of MBHM & RSM are 2.98%, 4.01%, 1.46%, and 0.89%, respectively, which are lower than those of standard k-ε model. Indeed, the pressure distribution on the propeller surfaces, the pressure and the velocity vector distribution of the flow field are also analyzed, which are consistent with the theory. It is demonstrated that MBHM & RSM on the thruster dynamics analysis are feasible. This paper provides reference in the thruster designing of underwater robot.展开更多
从三维Mesh数据中分割建筑物立面以识别对象,是三维场景理解的关键,但现有方法多依赖高成本的精细标注数据。针对该问题,提出了一种半监督学习方法,引入一种基于对比学习和一致性正则化的半监督语义分割(semi-supervised semantic segme...从三维Mesh数据中分割建筑物立面以识别对象,是三维场景理解的关键,但现有方法多依赖高成本的精细标注数据。针对该问题,提出了一种半监督学习方法,引入一种基于对比学习和一致性正则化的半监督语义分割(semi-supervised semantic segmentation based on contrastive learning and consistency regularization,SS_CC)方法,用于分割三维Mesh数据的建筑物立面。在SS_CC方法中,改进后的对比学习模块利用正负样本之间的类可分性,能够更有效地利用类特征信息;提出的基于特征空间的一致性正则化损失函数,从挖掘全局特征的角度增强了对所提取建筑物立面特征的鉴别力。实验结果表明,所提出的SS_CC方法在F1分数、mIoU指标上优于当前一些主流方法,且在建筑物的墙面和窗户上的分割效果相对更好。展开更多
文摘In order to analyze the hydrodynamic performance of the ducted propeller with high precision, this paper proposes a new method which combines Multi-Block Hybrid Mesh and Reynolds Stress Model (MBHM & RSM). The calculation errors of MBHM & RSM and standard two-equation model (standard k-ε model) on the ducted propeller JD7704 +Ka4-55 are compared. The maximum error of the total thrust coefficient KT, the duct thrust coefficient KTN, the torque coefficient KQ and the open-water efficiency η0 of MBHM & RSM are 2.98%, 4.01%, 1.46%, and 0.89%, respectively, which are lower than those of standard k-ε model. Indeed, the pressure distribution on the propeller surfaces, the pressure and the velocity vector distribution of the flow field are also analyzed, which are consistent with the theory. It is demonstrated that MBHM & RSM on the thruster dynamics analysis are feasible. This paper provides reference in the thruster designing of underwater robot.
文摘从三维Mesh数据中分割建筑物立面以识别对象,是三维场景理解的关键,但现有方法多依赖高成本的精细标注数据。针对该问题,提出了一种半监督学习方法,引入一种基于对比学习和一致性正则化的半监督语义分割(semi-supervised semantic segmentation based on contrastive learning and consistency regularization,SS_CC)方法,用于分割三维Mesh数据的建筑物立面。在SS_CC方法中,改进后的对比学习模块利用正负样本之间的类可分性,能够更有效地利用类特征信息;提出的基于特征空间的一致性正则化损失函数,从挖掘全局特征的角度增强了对所提取建筑物立面特征的鉴别力。实验结果表明,所提出的SS_CC方法在F1分数、mIoU指标上优于当前一些主流方法,且在建筑物的墙面和窗户上的分割效果相对更好。
