Based on reasonable assumptions that simplified the calculational model,a simple and practical method was proposed to calculate the post-construction settlement of high-speed railway bridge pile foundation by using th...Based on reasonable assumptions that simplified the calculational model,a simple and practical method was proposed to calculate the post-construction settlement of high-speed railway bridge pile foundation by using the Mesri creep model to describe the soil characteristics and the Mindlin-Geddes method considering pile diameter to calculate the vertical additional stress of pile bottom.A program named CPPS was designed for this method to calculate the post-construction settlement of a high-speed railway bridge pile foundation.The result indicates that the post-construction settlement in 100 years meets the requirements of the engineering specifications,and in the first two decades,the post-construction settlement is about 80% of its total settlement,while the settlement in the rest eighty years tends to be stable.Compared with the measured settlement after laying railway tracks,the calculational result is closed to that of the measured,and the results are conservative with a high computational accuracy.It is noted that the method can be used to calculate the post-construction settlement for the preliminary design of high-speed railway bridge pile foundation.展开更多
该文基于现场可编程门阵列(field-programmable gate array,FPGA),为永磁同步电机驱动提出一种扩张控制集模型预测电流控制策略(model predictive current control,MPCC)。由于在每个控制周期内只有8个基本电压矢量可供选择,传统有限控...该文基于现场可编程门阵列(field-programmable gate array,FPGA),为永磁同步电机驱动提出一种扩张控制集模型预测电流控制策略(model predictive current control,MPCC)。由于在每个控制周期内只有8个基本电压矢量可供选择,传统有限控制集模型预测电流控制(finite control set MPCC,FCS-MPCC)稳态性能较低。为此,文中采用具有818个可选矢量的ECS来实现更精细的电压输出。为减轻因电压矢量大幅增加而带来的计算负担,设计一种简化的最优矢量搜索策略,且可推广用于其他多目标成本函数。基于算法固有并行性,将所提ECS-MPCC方法在FPGA中进行实现,使电流环总控制时间缩短至0.59μs,从而可以消除计算延迟,提高电流环动态性能。最后,通过仿真和实验,验证所提ECS-MPCC策略的有效性。实验结果表明,与传统FCS-MPCC相比,ECS-MPCC的相电流总谐波失真降低77%。展开更多
针对自动驾驶边缘计算场景中行人车辆检测任务面临的模型计算复杂度高、参数量大导致的部署难题,该文提出一种轻量化神经网络模型YOMANet(Yolo Model Adaptation Network),基于异构FPGA平台设计YOMANet加速器(YOMANet-Accel),实现边缘...针对自动驾驶边缘计算场景中行人车辆检测任务面临的模型计算复杂度高、参数量大导致的部署难题,该文提出一种轻量化神经网络模型YOMANet(Yolo Model Adaptation Network),基于异构FPGA平台设计YOMANet加速器(YOMANet-Accel),实现边缘端人车检测的算法加速。YOMANet算法的主干网络采用轻量型网络MobileNetv2以大幅压缩模型参数量,颈部网络使用深度可分离卷积来代替常规卷积以提升训练速度,并在头部网络嵌入基于归一化的注意力模块(NAM)以增强网络对细节信息的捕获能力。为将YOMANet算法部署到现场可编程门阵列(FPGA)平台,该文针对卷积运算在任务层设计循环分块以调整内循环和外循环的顺序,在运算层对处理引擎单元(PE)设计乘加树,使得多个乘加运算可以同时执行,提高数据的并行计算效率。同时,针对数据存储过程采用双缓存机制来减少数据传输时延,对权重参数和激活函数进行int8数据量化以降低资源消耗。实验结果表明,YOMANet算法在训练平台上的检测精度和检测速度表现优异,对小目标和遮挡目标具备较好的检测能力,有效减少了误检和漏检情况的发生。算法部署到硬件平台后,YOMANet-Accel的目标检测效果保持在较高水平,硬件资源的能效比表现良好,有效发挥了FPGA的并行优势。展开更多
基金Projects(2009G008-B,2010G018-E-3) supported by Key Projects of China Railway Ministry Science and Technology Research and Development ProgramProject(CX2013B076) supported by Hunan Provincial Innovation Foundation For Postgraduate,China
文摘Based on reasonable assumptions that simplified the calculational model,a simple and practical method was proposed to calculate the post-construction settlement of high-speed railway bridge pile foundation by using the Mesri creep model to describe the soil characteristics and the Mindlin-Geddes method considering pile diameter to calculate the vertical additional stress of pile bottom.A program named CPPS was designed for this method to calculate the post-construction settlement of a high-speed railway bridge pile foundation.The result indicates that the post-construction settlement in 100 years meets the requirements of the engineering specifications,and in the first two decades,the post-construction settlement is about 80% of its total settlement,while the settlement in the rest eighty years tends to be stable.Compared with the measured settlement after laying railway tracks,the calculational result is closed to that of the measured,and the results are conservative with a high computational accuracy.It is noted that the method can be used to calculate the post-construction settlement for the preliminary design of high-speed railway bridge pile foundation.
文摘该文基于现场可编程门阵列(field-programmable gate array,FPGA),为永磁同步电机驱动提出一种扩张控制集模型预测电流控制策略(model predictive current control,MPCC)。由于在每个控制周期内只有8个基本电压矢量可供选择,传统有限控制集模型预测电流控制(finite control set MPCC,FCS-MPCC)稳态性能较低。为此,文中采用具有818个可选矢量的ECS来实现更精细的电压输出。为减轻因电压矢量大幅增加而带来的计算负担,设计一种简化的最优矢量搜索策略,且可推广用于其他多目标成本函数。基于算法固有并行性,将所提ECS-MPCC方法在FPGA中进行实现,使电流环总控制时间缩短至0.59μs,从而可以消除计算延迟,提高电流环动态性能。最后,通过仿真和实验,验证所提ECS-MPCC策略的有效性。实验结果表明,与传统FCS-MPCC相比,ECS-MPCC的相电流总谐波失真降低77%。
文摘针对自动驾驶边缘计算场景中行人车辆检测任务面临的模型计算复杂度高、参数量大导致的部署难题,该文提出一种轻量化神经网络模型YOMANet(Yolo Model Adaptation Network),基于异构FPGA平台设计YOMANet加速器(YOMANet-Accel),实现边缘端人车检测的算法加速。YOMANet算法的主干网络采用轻量型网络MobileNetv2以大幅压缩模型参数量,颈部网络使用深度可分离卷积来代替常规卷积以提升训练速度,并在头部网络嵌入基于归一化的注意力模块(NAM)以增强网络对细节信息的捕获能力。为将YOMANet算法部署到现场可编程门阵列(FPGA)平台,该文针对卷积运算在任务层设计循环分块以调整内循环和外循环的顺序,在运算层对处理引擎单元(PE)设计乘加树,使得多个乘加运算可以同时执行,提高数据的并行计算效率。同时,针对数据存储过程采用双缓存机制来减少数据传输时延,对权重参数和激活函数进行int8数据量化以降低资源消耗。实验结果表明,YOMANet算法在训练平台上的检测精度和检测速度表现优异,对小目标和遮挡目标具备较好的检测能力,有效减少了误检和漏检情况的发生。算法部署到硬件平台后,YOMANet-Accel的目标检测效果保持在较高水平,硬件资源的能效比表现良好,有效发挥了FPGA的并行优势。
