基于改进深度学习模型的焊缝缺陷检测算法

针对传统深度学习模型在进行焊缝缺陷检测时对小缺陷目标检测效果不理想问题,提出基于改进深度学习Faster RCNN模型的焊缝缺陷检测算法,算法通过多层特征网络提取多尺度特征图并共同作用于模型后续环节,以充分利用模型中的低层特征,增加细节信息;改进模型的区域生成网络,加入多种滑动窗口,从而优化了模型锚点的长宽比设置,提高检测能力。实验表明,改进Faster RCNN模型取得最优的缺陷检测结果,对于小缺陷目标仍取得较好的检测精度,从而验证了算法的有效性。     

发动机流量测控设备校准系统设计与实现

为了验证发动机流量测控设备的准确性和可靠性,本文建立一套流量测控设备的校准系统,实现对发动机流量测控设备进行校准。完成了发动机流量测控设备校准系统的不确定度分析,设计了校准系统中的关键硬件系统,开发了校准系统软件系统,从而实现对发动机流量测控设备的自动控制和校准,该校准系统具有配置灵活和操作简单的特点。     

流动显示的现状和发展

本文以第四届国际流动显示会议为主要内容,结合其它资料介绍流动显示的现状和发展。     

Capillary flow rate limitation in asymmetry open channel

This paper focuses on the stability of capillary forced flow.In space,open capillary channels are widely used as the liquid and gas separation devices to manage liquid positioning and transportation.Surface collapse happens when the flow rate exceeds the critical value,leading to a failure of propellant management.Knowledge of flow rate limitation is of great significance in design and optimization of propellant management devices(PMDs).However,the capillary flow rate limitation in an asymmetry channel has not been studied yet in the literature.In this paper,by introducing an equivalent angle to convert the asymmetry corner to a symmetry one,the one-dimensional theoretical model is developed.The flow rate limitation can then be investigated as a function of the channel geometry as well as liquid property based on the model.Comparisons between the asymmetry and symmetry channels bring forth the characteristics of the two kinds of channels,and demonstrate good accordance between the new advanced model and the existing one in the literature.This theoretical model can provide valuable reference for PMD designers.     

基于截尾概率-非概率混合模型的可靠性优化算法

针对工程中截尾概率变量与非概率变量同时存在的情况,给出一种新的截尾概率与非概率混合可靠性模型。在该混合可靠性模型基础上,按照可靠性指标(RIA)法给出双层嵌套可靠性优化模型,并采用改进搜索策略后的ST-Powell优化算法在外层搜索设计变量的最优值,内层采用能保证收敛的改进的有限步长迭代法求解混合可靠性指标。数值算例表明,改进搜索策略后的ST-Powell优化算法的全局寻优性得到显著提升;改进搜索策略后的ST-Powell优化算法与改进的有限步长迭代法相结合求解双层嵌套混合可靠性优化模型的正确性得到验证,且对于非线性程度较高的极限状态函数同样能够得到满足截尾概率与非概率混合可靠性模型指标要求的最优解,并对工程结构算例具有很好的适应性。     

SHS法制备硼化物陶瓷粉体的表征分析

采用自蔓延高温还原合成方法制备TiB2,TiB2-Al2O3和ZrB2-Al2O3陶瓷粉末.利用XRD,XPS,SEM以及TEM等分析测试手段对合成粉末进行表征和分析.结果表明,TiB2,ZrB2和Al2O3分别以主晶相的形式存在于所合成的各自粉体之中.相比之下,TiB2单相陶瓷粉末颗粒分布较宽.由于自蔓延高温合成(SHS)温度很高,部分颗粒形成团聚,宏观上使颗粒的平均粒径变大(＞5μm);而TiB2-Al2O3和ZrB2-Al2O3复合陶瓷粉末,因合成过程中Al2O3的形成,使得颗粒粒度分布明显变窄,分布均匀,颗粒尺寸也相应减小.分析认为这主要与复合粉末合成过程中,不同颗粒间形成良好结合的界面有关.     

内冷涡轮叶栅三维气热耦合数值模拟

为了研究涡轮动叶的内部冷却技术,对采用绝热边界的实心叶片、采用气热耦合的实心叶片和采用气热耦合的空冷叶片进行了研究.发现叶片导热对叶片温度场的影响相当显著;具有带肋蛇形通道和涡流矩阵肋片的内冷结构能使叶片温度大幅下降,但叶片表面温度分布不均匀性增大,换热系数沿叶片表面呈不规则分布.      

The role of AVDR in linear cascade testing

Linear cascade testing plays an important role in the research and development of turbomachinery and is widely used over the world.The ideal cascade model of a turbomachinery blade row is two-dimensional.In actual linear cascade testing, the flow through the test section converges due to the development of the boundary layer and secondary flow along the sidewall surfaces of the test section.Axial velocity density ratio(AVDR) is adopted to account for the deviation of the tested cascade flow from the ideal 2D model.Among numerous published cascade works, the influence of AVDR on cascade performance is seen to be complicated with many affecting factors, such as those related to cascade/blade geometry and flow conditions.Also, controlling AVDR is limited by the facility capability.Furthermore, real blade-to-blade flow in turbomachines is usually associated with AVDR greater than unity due to limited span of blades between the hub and shroud such that cascade testing without reducing AVDR could be favored sometimes.All these facets add complexity and diversification to the matter.The current paper reviews previous studies and results on AVDR.Consolidated understanding on the role of AVDR and recommendations on how to deal with it in linear cascade testing are provided.      

自动钻铆技术在波音737尾段项目中的应用

通过自动钻铆技术在波音737尾段中的应用,介绍了自动钻铆技术的特点及工艺.     

A Nonlinear Sub-grid Scale Model for Compressible Turbulent Flow

本文采用直接过滤的Navier-Stokes方程组作为可压缩湍流大涡模拟控制方程组。方程组中因过滤产生的高阶相关项用Taylor级数展开近似,但仅保留级数的一阶导数项。这样产生的误差相当于丢失了模型的部分耗散作用,本文用一种动力学非线性亚格子模型来补偿丢失的耗散影响。本文根据Caylay-Hamilton定理导出了一个非线性亚格子模型,模型中包含的系数由动力学模式确定。与传统的动力学Smagorinsky模式相比,这种动力学非线性模型(DNM)稳定性更好,且不需要在统计均匀方向进行统计平均来计算模型系数,因此减少了计算开销。本文用这种非线性亚格子模型对绕双椭球高超声速湍流进行了模拟,并将所得结果与实验值、计算值及理论值进行了对比。结果表明,本文模型可以有效地模拟可压缩湍流流场。