We presented a numerical examination of the effect of microtexturing negative rings' structure on the tribological performance of parallel bearing couples' cell. Three mcirotexturing rings, which are circle, s...We presented a numerical examination of the effect of microtexturing negative rings' structure on the tribological performance of parallel bearing couples' cell. Three mcirotexturing rings, which are circle, square and ellipse, were chosen and the analysis model were established. We used the Reynolds equation coupled with finite difference method and successive over relaxation Gauss-Seidel iterative method to solve the Newtonian flow's hydrodynamics within a bearing couple. The effect of texture density and radius ratio(thickness) of the microtexturing rings were investigated on the tribological performance under the similar operating conditions. The numerical simulation reveals that: 1) The microtexturing rings' structure can homogenize the local pressure much uniformly within the bearing cell. 2) The tribological performance is determined mainly by the microtexturing rings' geometry and texture density, and the thickness of the rings' structure can help to change the quantitative values. 3) The square and circle rings' s microtexturing surface can slightly improve the frictional performance with the bearing cells' gap, while the ellipse ring's surface may decrease the frictional performance. 4) The ellipse rings' microtexturing surface can achieve the minimum spacing gap but the maximum friction coefficient of the bearing couple, and then the circle and square rings' structure take the second and third place, respectively.展开更多
传统的应急响应及搜救行动的指挥控制(Command and Control,C2)活动以经典理论OODA环为指导。该模式可以快速、有效地指导救援现场的组织指挥活动。但当事件规模较大,响应层级较高时,该模式失去效用。论文针对以上问题,结合C2活动过程...传统的应急响应及搜救行动的指挥控制(Command and Control,C2)活动以经典理论OODA环为指导。该模式可以快速、有效地指导救援现场的组织指挥活动。但当事件规模较大,响应层级较高时,该模式失去效用。论文针对以上问题,结合C2活动过程机理的尺度关联性,运用PREA环&OODA环理论,建立了多域多尺度分布式C2模型框架。该模型在不同尺度(层级)的C2活动采用不同的理论指导,完整地描述了搜救C2活动从宏观到微观的指挥决策过程。论文将该模型应用分析至海上搜救领域,指导海上搜救行动。海上搜救的应用分析表明,多域多尺度分布式C2模型框架对海上应急救援指挥体系建设具有现实指导意义。展开更多
针对深度学习图像分类场景中多GPU并行后传输效率低的问题,提出一种低时间复杂度的Ring All Reduce改进算法。通过分节点间隔配对原则优化数据传输流程,缓解传统参数服务器并行结构的带宽损耗。基于数据并行难以支撑大规模网络参数及加...针对深度学习图像分类场景中多GPU并行后传输效率低的问题,提出一种低时间复杂度的Ring All Reduce改进算法。通过分节点间隔配对原则优化数据传输流程,缓解传统参数服务器并行结构的带宽损耗。基于数据并行难以支撑大规模网络参数及加速延缓的问题,根据深度学习主干网络所包含的权重参数低于全连接层权重参数、同步开销小、全连接层权重大与梯度传输开销过高等特点,提出GPU混合并行优化算法,将主干网络进行数据并行,全连接层进行模型并行,并通过改进的Ring All Reduce算法实现各节点之间的并行后数据通信,用于基于深度学习模型的图像分类。在Cifar10和mini ImageNet两个公共数据集上的实验结果表明,该算法在保持分类精度不变的情况下可以获得更好的加速效果,相比数据并行方法,可达到近45%的提升效果。展开更多
基金Funded by the National Natural Science Foundation of China(No.51675453)Technology Projects of Shenzhen(No.JCYJ 20160517103720819)
文摘We presented a numerical examination of the effect of microtexturing negative rings' structure on the tribological performance of parallel bearing couples' cell. Three mcirotexturing rings, which are circle, square and ellipse, were chosen and the analysis model were established. We used the Reynolds equation coupled with finite difference method and successive over relaxation Gauss-Seidel iterative method to solve the Newtonian flow's hydrodynamics within a bearing couple. The effect of texture density and radius ratio(thickness) of the microtexturing rings were investigated on the tribological performance under the similar operating conditions. The numerical simulation reveals that: 1) The microtexturing rings' structure can homogenize the local pressure much uniformly within the bearing cell. 2) The tribological performance is determined mainly by the microtexturing rings' geometry and texture density, and the thickness of the rings' structure can help to change the quantitative values. 3) The square and circle rings' s microtexturing surface can slightly improve the frictional performance with the bearing cells' gap, while the ellipse ring's surface may decrease the frictional performance. 4) The ellipse rings' microtexturing surface can achieve the minimum spacing gap but the maximum friction coefficient of the bearing couple, and then the circle and square rings' structure take the second and third place, respectively.
文摘传统的应急响应及搜救行动的指挥控制(Command and Control,C2)活动以经典理论OODA环为指导。该模式可以快速、有效地指导救援现场的组织指挥活动。但当事件规模较大,响应层级较高时,该模式失去效用。论文针对以上问题,结合C2活动过程机理的尺度关联性,运用PREA环&OODA环理论,建立了多域多尺度分布式C2模型框架。该模型在不同尺度(层级)的C2活动采用不同的理论指导,完整地描述了搜救C2活动从宏观到微观的指挥决策过程。论文将该模型应用分析至海上搜救领域,指导海上搜救行动。海上搜救的应用分析表明,多域多尺度分布式C2模型框架对海上应急救援指挥体系建设具有现实指导意义。
文摘针对深度学习图像分类场景中多GPU并行后传输效率低的问题,提出一种低时间复杂度的Ring All Reduce改进算法。通过分节点间隔配对原则优化数据传输流程,缓解传统参数服务器并行结构的带宽损耗。基于数据并行难以支撑大规模网络参数及加速延缓的问题,根据深度学习主干网络所包含的权重参数低于全连接层权重参数、同步开销小、全连接层权重大与梯度传输开销过高等特点,提出GPU混合并行优化算法,将主干网络进行数据并行,全连接层进行模型并行,并通过改进的Ring All Reduce算法实现各节点之间的并行后数据通信,用于基于深度学习模型的图像分类。在Cifar10和mini ImageNet两个公共数据集上的实验结果表明,该算法在保持分类精度不变的情况下可以获得更好的加速效果,相比数据并行方法,可达到近45%的提升效果。
