开缝钝体尾迹的拟序结构

弄清开缝钝体尾迹的拟序结构是认识其火焰稳定机理的基础。在雷诺数R e为470000条件下,采用RNGk-ε模型对通道内的开缝钝体尾迹进行数值模拟来分析大涡尺度的拟序结构。模拟结果显示,偏向一侧的中缝流将近尾分成主回流区和次回流区,主回流区的漩涡脱落激发扰动,引起近尾的绝对不稳定。并提出单涡突然置于两剪切层间的漩涡脱落机理来解释拟序结构不稳定特性。为了验证上述结论,在闭式风洞中采用激光粒子测速技术(P IV)对开缝钝体的尾迹进行了实验研究,其结果与数值分析较好地吻合。     

卷积神经网络在装备磨损颗粒识别中的研究综述

铁谱法是用于装备故障诊断的1种重要方法,其中铁谱法的重点是铁谱图像的分析,即磨损磨粒分析. 卷积神经网络是当下最流行的深度学习算法之一,其广泛应用于图像识别领域,使得图像识别领域得到突破. 随着卷积神经网络的快速发展,磨损颗粒在智能识别方面的技术取得了重大的突破. 本文中首先简述了卷积神经网络与磨粒智能识别的发展历史,针对基于卷积神经网络的磨粒识别方法进行了从图像数据集处理到模型优化技术方面的介绍,并详细说明了这些技术在磨粒识别中的具体应用实例. 然后从现有网络和自设计网络两方面分类,整理了近年来卷积神经网络应用于磨粒智能识别的代表性文献,综述了这些工作所提出的模型结构和特点,分析并阐述了各个模型主要的识别原理,各个网络结构存在的优缺点,以及它们的数据采用情况等,并对未来磨粒智能识别的主要研究方向进行了展望. 最后肯定了卷积神经网络方法在磨粒智能识别方面的重要性,同时指出了基于此方法的磨粒识别模型的缺点,并提出了应紧跟图像识别领域的最新技术以促进磨粒智能识别水平提高等建议,对磨粒智能识别的发展具有一定的意义.      

返回舱高雷诺数再入过程底部流动稳定性

返回舱高雷诺数再入过程中存在肩部高热流、底部阻力无法准确预测以及非定常振动等问题,解决此类问题的关键是分离和转捩等物理现象的准确识别. 本文采用大涡模拟方法细致刻画了返回舱类钝体外形在高雷诺数再入过程中的分离和转捩等物理现象,获得了返回舱底部流动形态以及稳定性特征. 从肩部剪切失稳、底部流动结构失稳、尾迹发展区以及远尾迹区的耦合失稳等多个角度分析了返回舱外形的底部流动失稳机制.研究发现, 返回舱类外形底部流动稳定性主要存在两类失稳模式即肩部剪切失稳模式以及底部流动结构失稳模式,二种模式存在耦合效应, 同时在远尾迹湍流区域存在类卡门涡街的振荡行为.这些认识为理解返回舱外部扰动因素对底部流动的作用机理及返回舱稳定性控制提供了基础理论支撑.      

基于视觉技术的光纤陀螺闭合光路光纤缺陷检测方法

针对光纤陀螺内部光纤在装配后光路闭合,无有效手段进行光纤缺陷检查,对光纤陀螺长期稳定可靠工作带来潜在危险的问题,提出利用红外视觉检测技术检查光纤缺陷,分析光纤缺陷图像特征,采用最大熵法进行图像分割,提出了结合直方图特征、缺陷区域形状特征、缺陷边界形状特征提取方法,采用改进的神经网络分类方法,用两层网络进行缺陷分类,识别判断不同类别的光纤缺陷。针对光纤缺陷图像的处理结果表明,该方法能够有效地检测光纤缺陷,对不可接受的缺陷能够100%地正确判断。     

钝体尾流控制机理及方法研究进展

首先从涡脱落生成理论出发对钝体尾流控制方法进行了分类,并简单介绍了国内尾流控制研究情况. 之后介绍了我们用窄条或小方柱取代小圆柱后,对Strykowsky和Sreenivasan 控制方法的改进及其在高雷诺数下对圆柱和方柱尾流涡脱落的有效抑制情况, 并探讨了控制件钝度对抑制效果的影响.第3部分用实验数据对各个涡脱落生成模型做了分析与检验, 指出控制件方法的机理与改变钝体分离位置、减小钝体背压吸力、改变流动的展向相关性、 防止钝体两侧剪切层相互作用等无关,而与钝体近尾流速度剖面的局部修正及其稳定性的改变有关. 最后简单介绍了控制件方法今后研究工作展望及其工程应用前景.     

基于四元数卷积神经网络的移动机器人闭环检测

为提高复杂环境下移动机器人闭环检测的准确性,减少视觉里程计的累积误差,提出了四元数卷积神经网络的闭环检测算法。首先,利用超像素分割,提取多尺度路标,提升图像描述的视角和外观不变性;然后,扩展卷积神经网络的卷积层为四元数卷积层,增加红绿蓝三通道的关联性,提取彩色图像的深层信息,更好地体现彩色图像的整体性;最后,在图像相似度度量中,不仅计算路标距离,还考虑路标形状和空间分布信息,提高相似性度量的准确性。在卡内基梅隆大学和昆士兰大学公开数据集上对算法有效性进行了验证。实验结果表明,相比传统闭环检测算法,所提算法在保证较高召回率的同时,提高了闭环检测的准确率,在数据集上分别取得87.64%和90.12%的平均准确率,显著提高了光照、视角和季节变化等复杂环境中移动机器人闭环检测的准确性。     

基于并行卷积神经网络的地磁方向适配性分析

针对地磁方向适配性分析时人工特征提取主观性较强、所取特征难以表达深层的结构性特征的问题,并为了进一步提高方向适配性分析的准确率,提出了一种基于并行卷积神经网络的地磁方向适配性分析方法。首先,从不同角度建立了地磁场在6个代表方向上的适配性分析图;然后,从同一磁场的不同角度出发,利用卷积神经网络自动完成了特征学习,得到了更为全面的方向适配性特征描述;最后,在并行卷积神经网络所得特征的基础上,利用BP网络建立了地磁方向适配性的分析模型。仿真结果证明,该方法可以有效避免人工特征提取和计算等复杂步骤,实现了地磁方向适配性分析的自动化,而且可以获得优于传统网络和单路卷积神经网络的准确率。     

基于深度学习和近场动力学的层合板冲击工况识别

在冲击荷载作用下复合材料会产生断裂和分层等损伤。基于损伤数据对冲击工况进行识别，对改善复合材料的设计和确保其安全使用具有重要意义。基于此，本文提出一种基于深度学习和近场动力学（PD）理论的层合板冲击工况识别方法。首先使用改进的表面修正系数PD理论建立复合材料层合板刚体冲击损伤演化分析PD模型，PD模型数值模拟结果结合噪声数据增强技术构建层合板的冲击工况数据库；基于深度学习-卷积神经网络（CNN），对不同工况下的冲击损伤演化数据进行训练，实现对未知冲击工况的识别。结果显示，对于钢球冲击速度和角度的识别准确率均高于90%。     

基于经验小波变换和卷积神经网络的水泵水轮机无叶区流态特征提取及识别研究

监测和识别原型水泵水轮机无叶区的流动状态，对于保证抽水蓄能电站的运行安全性和稳定性有非常重要的意义。本文提出了一种基于经验小波变换、散布熵和卷积神经网络原理的流态特征提取和识别方法，首先使用经验小波变换对压力脉动信号进行分解，然后通过计算各分量的散布熵提取流态相关特征，最后通过利用特征–标签对训练卷积神经网络得到的智能识别模型，实现了无叶区流态识别。利用从国内某水泵水轮机采集到的发电、抽水和空转工况下实测压力脉动信号对该方法进行了测试，测试的平均准确率达到了94.84%，证明了该方法的有效性。     

基于卷积特征的光纤缺陷检测方法

针对人工难以实现熊猫型保偏光纤的缺陷检测问题,提出了一种小样本下基于卷积特征的图像检测方法。首先针对光纤缺陷特征将Inception V3模型微调,使用微调后模型提取光纤的2048维卷积特征;其次使用主元分析法将2048维特征降为74维;最后使用降维后的特征训练支持向量机分类器,同时使用粒子群算法对分类器参数寻优,实现对光纤缺陷的识别与分类。经实验证明,该方法对光纤涂覆层微小缺陷的识别率达到97%,涂覆层局部损伤和严重破损的识别率达到100%,对降低光纤环绕制中原纤损失、提升光纤环的精密性、研究光纤缺陷对光纤陀螺精度的影响有一定意义。     

Approach towards self-preservation of turbulent cylinder and screen wakes

Two-dimensional wakes generated from a circular cylinder and a 50% solidity screen have significantly different initial conditions. Accordingly, the approach towards self-preservation is quite different for the two wakes. For the cylinder wake, the normalized Reynolds stresses and spanwise vorticity decrease with increasing distance from the wake generator; the inverse occurs in the screen wake. Distributions of mean velocity, Reynolds stresses, and rms spanwise vorticity indicate that self-preservation is reached at a much smaller streamwise distance for the screen than for the cylinder wake. This result is consistent with the previously reported topological differences between these two flows.     

Vortex motions in stratified wake flows

The vortex wakes of obstacles (circular cylinder and sphere) moving through a linearly stratified fluid have been investigated, by means of flow visualization, at Reynolds numbers smaller than 800 and non-dimensional buoyancy frequencies smaller than 6. Vortex shedding from a horizontally suspended circular cylinder is suppressed when the fluid is stratified. The wake of a sphere is affected by lee waves when the Reynolds number exceeds about 200.     

Bluff body wakes with free, fixed, and discontinuous separation at low Reynolds numbers and low aspect ratio

Some comparative experimental results are presented of bluff body wakes with free, fixed, and discontinuous separation. The particular examples considered are a circular cylinder, a plain blunt trailing edge aerofoil and a segemented blunt trailing edge aerofoil. The evolution of the near wake and vortex shedding modes are compared and discussed.     

Orthogonal wavelet analysis of turbulent wakes behind various bluff bodies

The multi-scale structures of turbulent wakes generated by three kinds of bluff body, i.e. circular cylinder, square cylinder and compound of cylinder and square (CS) cylinders, have been experimentally investigated in this paper. Firstly, the instantaneous velocity fields and vorticity were measured by the high-speed PIV technique in a circulating water channel. The instantaneous streamlines and corresponding normalized vorticity contours are obtained at a Reynolds number of 5600. Then one- and two-dimensional wavelet multi-resolution technique was used to analyze the instantaneous velocities and vorticity measured by the high-speed PIV. The turbulence structures were separated into a number of subsets based on their central frequencies, which are linked with the turbulence scales. The instantaneous vorticity and Reynolds shear stresses of various scales were examined and compared between the three generators. It is found that the large-scale turbulent structure makes the largest contribution to the vorticity and Reynolds shear stresses for the three wake generators and exhibits a strong dependence upon the initial conditions or the wake generators. The large-scale vorticity and the sizes of vortex in the circular and square cylinders are larger than those in the CS cylinder wake. The contributions to the Reynolds shear stresses from the large-scale turbulent structures account for 90-96% to the measured maximum Reynolds shear stresses for the three wakes. However, the small-scale structures make less contribution to the vorticity and Reynolds shear stresses.     

圆柱尾流雷诺应力与平均运动变形率的空间弛豫效应

针对圆柱尾流中沿流向存在的Karman涡街周期性涡旋结构,对湍流雷诺应力与平均运动变形率之间的空间弛豫效应进行了实验研究.在回流式水槽中,放入不同直径的圆柱模型,获得不同雷诺数下的圆柱尾流,利用二维高时间分辨率粒子图像测速(TRPIV)技术测量圆柱尾流二维瞬时速度空间分布图像的时间序列.经过数字图像处理,获得二维雷诺应...     

Experimental study on wake flow structures of screen cylinders using PIV

Experiments were conducted in a water flume using Particle Image Velocimetry (PIV) to study the evolution of the vortical structures in the wakes of four types of screen cylinders at a Reynolds number of about 3200. The results were compared with that of a bare cylinder. The screen cylinders were made of stainless steel screen meshes of various porosities (37%, 48%, 61% and 67%) rolled into cylindrical shapes. Smoke wire flow visualisations in a wind tunnel were also conducted in support of the PIV tests. Depending on the porosity of the screen mesh, two vortex formation mechanisms for the screen cylinder wakes were identified. One was associated with wake instability and the other was associated with shear-layer (Kelvin-Helmholtz) convective instability which involved merging through pairing and tripling of small-scale vortices within the shear layers. The former was responsible for the formation of large-scale vortices in the bare cylinder and the screen cylinder wakes with 37% and 48% porosities, while the latter was responsible for the screen cylinder wakes with 61% and 67% porosities. The results also showed that with increasing porosity, the vortex formation region was extended farther downstream and the Reynolds shear stress, the Turbulent Kinetic Energy (TKE) and vortex intensity were decreased constantly.     

Turbulence measurements around a body in a re-circulating water channel by a hot film probe

Measurements were made in the stern boundary layers and near wakes of an elliptic cylinder and a slender ship model. Turbulence intensities, Reynolds stresses, kinematic eddy viscosities and mixing lengths are presented. For the elliptic cylinder, furthermore, auto-correlation and power spectrum are obtained. It is shown that the separation from the cylinder increases the turbulence intensities, and the Kármán vortices enhance the turbulence power at the vortex frequency. All distributions of Reynolds stresses in the thick boundary layer and wake of the ship model show a secondary low peak at about half the thickness.     

Interference Model of Wake Development Downstream of a Self-Propelled Body

The laws of development of the turbulent wake behind self-propelled and unpropelled bodies are studied. It is established that the interfering flows generated by the frame and the propelling system conserve their individuality far downstream of the body. It is shown that a function representing the sum of the solutions for two self-similar drag wakes fairly closely approximates the measured velocity deficit and Reynolds stress fields in the wakes behind self-propelled bodies. The possibility of reliable extrapolation of the results of approximating the bench test data to higher Reynolds numbers and greater distances from the body is discussed.     

LES-based Study of the Roughness Effects on the Wake of a Circular Cylinder from Subcritical to Transcritical Reynolds Numbers

This paper investigates the effects of surface roughness on the flow past a circular cylinder at subcritical to transcritical Reynolds numbers. Large eddy simulations of the flow for sand grain roughness of size k/D = 0.02 are performed (D is the cylinder diameter). Results show that surface roughness triggers the transition to turbulence in the boundary layer at all Reynolds numbers, thus leading to an early separation caused by the increased momentum deficit, especially at transcritical Reynolds numbers. Even at subcritical Reynolds numbers, boundary layer instabilities are triggered in the roughness sublayer and eventually lead to the transition to turbulence. The early separation at transcritical Reynolds numbers leads to a wake topology similar to that of the subcritical regime, resulting in an increased drag coefficient and lower Strouhal number. Turbulent statistics in the wake are also affected by roughness; the Reynolds stresses are larger due to the increased turbulent kinetic energy production in the boundary layer and separated shear layers close to the cylinder shoulders.