基于注意力机制的堆叠LSTM网络雷达HRRP序列目标识别方法

传统的雷达高分辨距离像(high resolution range profile, HRRP)序列识别方法依赖于人工特征提取, 并且现有的深度学习方法存在梯度消失问题, 导致收敛速度慢, 识别精度低。针对上述问题, 提出一种基于注意力机制的堆叠长短时记忆(attention-based stacked long short-term memory, Attention-SLSTM)网络模型, 该模型通过堆叠多个长短时记忆(long short-term memory, LSTM)网络层, 实现了HRRP序列更深层次抽象特征的提取; 通过替换模型的激活函数, 减缓了堆叠LSTM(stacked LSTM, SLSTM)模型梯度消失问题; 引入注意力机制计算特征序列的分配权重并用于分类识别步骤, 增强了隐藏层特征的非线性表达能力。模型在雷达目标识别标准数据集MSTAR上多种不同目的的实验结果表明, 所提方法具有更快的收敛速度和更好的识别性能, 与多种现有方法对比具有更高的识别率, 证明了所提方法的正确性和有效性。     

基于复数卷积-残差网络的雷达杂波幅度统计模型分类

雷达杂波幅度统计模型分类是进行杂波背景下检测目标的重要步骤。雷达杂波原始数据通常是复数数据, 但现有杂波幅度统计模型分类研究都是在实数数据上完成的。复数数据同时包含幅度和相位信息, 更丰富的信息量有助于雷达杂波幅度统计模型分类。为此, 引入复数神经网络, 利用仿真杂波高分辨距离像(high resolution range profile, HRRP)复数数据, 对雷达杂波幅度统计模型分类问题进行研究, 完成了以下工作: 一是为构建复数最大池化层, 定义并改进了复数最大池化算法, 通过复数卷积神经网络(complex-valued convolutional neural networks, CV-CNN) 对杂波幅度统计模型的分类实验, 对比了两种复数最大池化算法和复数平均池化算法的分类效果, 实验结果表明复数最大池化算法的分类效果更好, 分类准确率为97.29%;二是为进一步提高分类准确率, 构建了复数卷积-残差网络(complex-valued convolution-ResNet, CV-CRN), 通过实验对比分析了CV-CRN的性能, 实验结果表明, CV-CRN的分类性能优于CV-CNN, 分类准确率达到98.84%, 并具有较好的鲁棒性。     

A pilot study on wastewater treatment by constructed wetlands with high dissolved oxygen vegetated submerged bed

It is necessary to treat wastewater in a more environmentally friendly fashion, vegetated submerge bed constructed wetlands (CWs) has become the focus of the research in this field. In this paper, the overall capacity of vertical subsurface flow CWs in the removal of pollutants from municipal wastewater effluent and the effects of the depth of vegetated submerged bed on the removal efficiency were studied. The results showed that the COD, NH4+-N and TN could be removed effectively in the vertical subsurface flow CWs, the best efficiency was obtained from the 10 cm run. However, the TP removal was not effective in all vertical CWs. A shallow depth of vegetated submerged bed may avoid the risk of substrate plugging effectively, and may keep a stable operation of CWs in long term. Decreasing the depth of vegetated submerged bed could create a sufficient aerobic circumstance in which the concentration of DO in bed was super saturation.     

光纤光栅技术在玻璃幕墙边缘检测应用

针对玻璃幕墙结构胶老化、失效以及高风压载荷导致的玻璃脱落等安全事故问题,提出了一种基于光纤光栅传感技术检测玻璃幕墙边缘动态应力变化的方法,构造了玻璃面板边缘应变与结构胶之间的多模态耦合模型,从而预测玻璃幕墙结构安全状况.通过对比分析多模态应变的仿真与实验结果数据可以得到：准分布光纤光栅能够给出结构胶的失效位置,提前对玻璃幕墙结构胶的健康状况进行安全性能评估与反馈.由此可及时更换处理失效玻璃,减少玻璃幕墙脱落事故,在减少经济损失的同时提升了玻璃幕墙的安全性,具有实际价值.     

计算机虚拟仿真技术在数控加工中的应用研究

为有效预防数控加工中机床、刀具和夹具相互之间的干涉碰撞问题,采用计算机虚拟仿真技术,构建了某数控车床虚拟加工仿真系统,重点介绍了该数控车床虚拟装配模型的建模过程．通过设置虚拟机床参数,完成了虚拟环境中的典型零件加工过程仿真,同时进行了数控加工过程中可能发生的碰撞和干涉检查,分析了零件的可加工性和工序合理性,以及检验数控程序的正确性．仿真结果表明：该数控加工虚拟仿真系统的装配建模方法可行,加工仿真过程可有效判断该机床工艺系统可能存在的相互干涉问题,对预防现实加工中的碰撞事故的发生具有指导意义,为数控车床虚拟加工系统的后续开发提供了一种新的视角和途径．     

四阶时滞微分方程边值问题的正解

随着泛函微分方程理论的发展以及其在物理、力学、自动控制理论、生物学、经济学等众多学科中的应用,时滞微分方程边值问题成为关注的一个热点.运用锥上的不动点指数理论研究了四阶时滞微分方程边值问题{u(4)(t)+au″(t)-bu(t)=f(t,ut),t∈[0,1],u(t)=Ф(t),t∈[-τ,0],u(0)=u(1)=u″(0)=u″(1)=0正解的存在性,其中,f:[0,1]×C+→[0,+∞)连续,C+={φ∈C|φ(θ)≥0,θ∈[-τ,0]},Ф(t)∈C([-τ,10],[0,+∞)),Ф(0)=0,对t∈[0,1],ut(θ)=u(t+θ),θ∈[-τ,0],0≤τ,且a,b∈R,满足a2π2,b-a2/4,b/π4+a/π21.所得结果推广和改进了现有结果.     

具卷积核和Cauchy核的奇异积分方程的求解

本文综述介绍了关于具卷积核和Ｃａｕｃｈｙ核的奇异积分方程的求解研究，提出了几个需在今后研究中进一步讨论的问题．     

蓄电池剩余容量在线检测方法研究

针对蓄电池充放电的特点,指出了几种蓄电池剩余容量检测方法的缺点,接着对用内阻法测量蓄电池剩余容量的方法从原理上进行分析,提出一种新的蓄电池剩余容量在线检测方法,并设计了监控系统软件.实验结果表明了这种改进方法的有效性.     

Nonuniform stretch processing for the range profile of a target with micro-motion

Complex motions such as micro-motions induce complex range-Doppler coupling effect in radar imaging. Hence, it is a challenge to reconstruct a clear range profile of a target with micro-motion. Stretch processing with fine motion compensation cannot perform well while it ties to eliminate the distortion of range profile caused by Doppler effect, because it is difficult to accurately estimate each scattering center velocity during coherent processing interval. Based on nonuniform stretch processing (NSP), a new technology is proposed to decouple range, time and Doppler and then achieve range profile of target with micro-motion. The new technology uses a new synthetic wide-band waveform and has an innovative mathematic interpretation. Statistical properties of the new approach are primarily examined. Typical models of target with micro-motion for high resolution radar are presented. Monte Carlo simulations demonstrate the performance of the new technology and the simulation results confirm our expectations.     

阿基米德基本蜗杆齿轮滚刀的偏差分析

针对齿轮加工中的齿轮滚刀造成所加工齿轮副接触精度较低,直接影响到齿轮传动的平稳性问题,经过对滚刀齿廓的造型工艺出现的偏差进行分析研究,找出产生偏差的原因,提出了使用新型刀具材料,采用铲削工艺来取代一直沿用的铲磨工艺可对齿轮滚刀进行精加工,从而使齿轮滚刀的设计方法和加工方法在造型原理上相一致,以达到提高齿轮滚刀制造精度的目的。