This paper proposes a case study in the control of a heavy oil pyrolysis/cracking furnace with a newly extended U-model based pole placement controller(U-PPC). The major work of the paper includes: 1) establishing a c...This paper proposes a case study in the control of a heavy oil pyrolysis/cracking furnace with a newly extended U-model based pole placement controller(U-PPC). The major work of the paper includes: 1) establishing a control oriented nonlinear dynamic model with Naphtha cracking and thermal dynamics; 2) analysing a U-model(i.e., control oriented prototype) representation of various popular process model sets; 3)designing the new U-PPC to enhance the control performance in pole placement and stabilisation; 4) taking computational bench tests to demonstrate the control system design and performance with a user-friendly step by step procedure.展开更多
U-Net因结构简单且高效被广泛应用于医学分割领域。然而,U-Net的跳跃连接不能很好地弥补编码器和解码器之间的语义差距。而医学分割数据的标注要求严格,使得数据集数量和规模都较小。针对上述问题,设计多尺度注意力融合(MSAF)模块,旨在...U-Net因结构简单且高效被广泛应用于医学分割领域。然而,U-Net的跳跃连接不能很好地弥补编码器和解码器之间的语义差距。而医学分割数据的标注要求严格,使得数据集数量和规模都较小。针对上述问题,设计多尺度注意力融合(MSAF)模块,旨在利用注意力机制可调整网络学习方向的特点和多尺度特征融合来有效缓解语义偏差。MSAF模块在前2个阶段使用通道注意力来捕获全局特征;在后2个阶段使用空间注意力来捕获局部特征;最后将多个阶段提取的特征进行融合以增强特征信息。此外,提出基于傅里叶变换的数据增强(FTDA)方法解决医学分割数据集稀少的问题。FTDA通过扰动输入图像在频域中的幅度信息实现其相位信息的数据增强。在MoNuSeg、CryoNuSeg和2018 Data Science Bowl数据集上的实验结果表明,提出方法的mIoU和Dice指标比其他先进方法表现出更好的性能。此外,提出的FTDA方法对小规模数据集也具有较好的增益效果。展开更多
基金partially supported by the National Natural Science Foundation of China(61273188,61473312)Taishan Scholar Construction Engineering Special Funding of Shandong
文摘This paper proposes a case study in the control of a heavy oil pyrolysis/cracking furnace with a newly extended U-model based pole placement controller(U-PPC). The major work of the paper includes: 1) establishing a control oriented nonlinear dynamic model with Naphtha cracking and thermal dynamics; 2) analysing a U-model(i.e., control oriented prototype) representation of various popular process model sets; 3)designing the new U-PPC to enhance the control performance in pole placement and stabilisation; 4) taking computational bench tests to demonstrate the control system design and performance with a user-friendly step by step procedure.
文摘U-Net因结构简单且高效被广泛应用于医学分割领域。然而,U-Net的跳跃连接不能很好地弥补编码器和解码器之间的语义差距。而医学分割数据的标注要求严格,使得数据集数量和规模都较小。针对上述问题,设计多尺度注意力融合(MSAF)模块,旨在利用注意力机制可调整网络学习方向的特点和多尺度特征融合来有效缓解语义偏差。MSAF模块在前2个阶段使用通道注意力来捕获全局特征;在后2个阶段使用空间注意力来捕获局部特征;最后将多个阶段提取的特征进行融合以增强特征信息。此外,提出基于傅里叶变换的数据增强(FTDA)方法解决医学分割数据集稀少的问题。FTDA通过扰动输入图像在频域中的幅度信息实现其相位信息的数据增强。在MoNuSeg、CryoNuSeg和2018 Data Science Bowl数据集上的实验结果表明,提出方法的mIoU和Dice指标比其他先进方法表现出更好的性能。此外,提出的FTDA方法对小规模数据集也具有较好的增益效果。
