基于卷积神经网络的图像识别综述

随着工业化进程的迅猛发展，产生了大量的图像信息，传统的图像识别技术难以处理如此庞大的图像数据以及满足速度和精度上的要求，大数据及深度学习技术应运而生，基于卷积神经网络的图像识别方法成为目前图像识别的主流算法。文中首先介绍了传统图像识别技术及存在的问题，引入了卷积神经网络的深度学习方法，重点说明了卷积网络中间层的结构和特点，然后介绍图像识别中经典的卷积神经网络模型及相互间的区别，最后简要综述卷积神经网络在图像识别中的应用，指出了有监督的卷积网络学习缺点及无监督学习的研究方向。     

基于改进的SqueezeNet直升机滚动轴承故障诊断

针对现有基于卷积神经网络的故障诊断方法存在模型参数量和计算量大的问题，提出一种改进的SqueezeNet模型应用于直升机滚动轴承故障诊断。该模型借鉴VGG16模型的思想，在经典的SqueezeNet基础上，采用3个3×3卷积核代替1个7×7卷积核，实现了在相同感知野条件下增加网络容量、增强非线性、减少网络参数量，采用卷积层、池化层和Fire模块、池化层两大结构交替的方式组成模型特征提取层，在保障特征提取能力的情况下，进一步减少了网络参数量。通过轴承数据开展模型故障诊断实验，结果表明该模型诊断精度达到99.65%，与传统卷积神经网络及经典的SqueezeNet模型相比诊断精度相当，而计算量与参数量最大缩减约6倍和36倍。     

航空电子设备冷却用环路热管冷凝器热沉分析

为确定环路热管用于机载电子设备散热的适用性,讨论了应用于航空电子设备冷却的环路热管冷凝器可用热沉,提出将机翼蒙皮作为环路热管冷凝器的热沉.通过机翼蒙皮的传热分析,计算了飞行任务包线下机翼蒙皮内表面温度,以确定环路热管冷凝器热沉温度范围.理论计算表明,利用环路热管将机载电子设备的热量耗散于飞机蒙皮是一种可行的热管理技术.      

三轴稳定式卫星的热特性研究

热特性研究是卫星热控制系统、红外辐射特性研究、红外隐身设计的基础和重要组成部分。建立卫星运行轨道计算模型,获得不同时刻卫星三维坐标及轨道高度。根据太阳、卫星、地球三者位置关系,建立三轴稳定式卫星空间热流计算模型。综合考虑空间热流、向外辐射、内部热载荷等因素,对卫星进行传热分析计算,获得各时刻卫星各面温度分布。分析了太阳吸收系数和卫星内部热载荷对表面温度分布的影响。研究结果表明:卫星在地球阴影区各面温度明显降低;除了散热面,太阳吸收系数对卫星表面温度影响显著;可以根据散热面大致地判断卫星运行状况。     

过热蒸汽射流冷却叶片热耦合数值模拟

在对射流冷却平板模型和射流冷却凹板模型数值模拟的基础上,设计了射流冲击冷却式蒸汽冷却叶片,并对其进行了热耦合数值模拟.通过改变平板和凹板的进口雷诺数,射流孔与靶板间距和凹板曲率,对靶板的冷却传热进行了对比研究,得到传热量与各个参数的变化规律.然后根据上面得到的变化规律设计出射流冲击冷却式蒸汽冷却叶片,应用ANSYS CFX 10.0数值程序,对冲击冷却的流动和传热过程进行了三维热耦合数值模拟,得到叶片表面温度和对流换热系数分布.      

航空发动机热端部件磷光测温技术研究进展

详细介绍了磷光测温技术的物理机制、测量方法、常用材料及制备工艺，并从磷光材料及其制备工艺、温度测试方法及系统、发动机环境下信号传输方案三个层面梳理了磷光测温技术在航空发动机热端部件表面温度测试领域的发展历程和研究动态。通过对航空发动机热端部件磷光测温技术研究进展的全面分析，充分验证了磷光测温技术与待测面发射率无关，受复杂燃气组分吸收散射影响小，可测量半透明介质内部温度场的独特优势，凸显了磷光测温技术在航空发动机极端环境下热端部件瞬态温度场测试中的应用潜力，有望实现更高的测温范围和测试精度，支撑新一代航空发动机的精细化设计。     

扰流柱布局形式对层板隔腔内流动与换热特性影响的数值模拟

正随着航空发动机推重比要求的不断提高,发动机热端部件热负荷急剧上升,尤其是燃烧室火焰筒需要有效的热防护。因此对先进冷却技术的研究需要不断进行,层板冷却是一种较为先进的冷却技术,国外的层板冷却技术已经进入实际应用阶段。Favaretto等人通过数值模拟和传热优化分析研究了层板结构不同的冲击孔、气膜孔、扰流柱直径和不同扰流柱高度的影响。Funazaki通过数值模拟及实验研究了放大层板冷却结构的内部对流换热系数。研究发现数值模型计算结果与实验值符合较好。另外,平均对流换热系数     

基于机器学习的单孔绝热气膜冷却效率分布预测研究

为了研究利用机器学习技术对绝热冷却效率进行直接预测的方法及特点,搭建了基于上采样卷积神经网络的机器学习模型,生成了用于训练以及验证的数值模拟数据集,使用了监督式学习的方法对模型进行了训练。训练使用反向传播算法和基于随机梯度下降的Adam优化算法,输入模型的参数包括吹风比、主流湍流强度、喷射角、孔形状、孔尺寸,模型输出为绝热冷却效率云图。模型的预测结果显示,基于上采样卷积神经网络的模型在回归预测问题上性能表现良好（测试集像素点绝对误差在0.05左右）,同时,给出了此类卷积网络的训练意见。研究表明,对于云图式的回归预测目标,卷积神经网络预测结果可靠,灵活性高,有良好的工程应用价值。     

间隙射流对端壁冷却和传热特性影响的研究

采用数值求解三维Reynolds-averaged Navier-Stokes(RANS)方程和k-ω湍流模型,研究了间隙射流对燃气轮机叶片端壁冷却和传热特性的影响。通过数值结果与实验数据的比较,验证了数值方法的正确性。在此基础上,研究了间隙射流质量流量比、间隙射流角度对端壁流动结构、气膜冷却性能以及传热特性的影响规律。结果表明,受到端壁二次流结构的限制,冷却气体主要集中在叶片前部吸力面侧。当间隙射流质量流量比小于1%时,会发生主流入侵现象,从而削弱前缘马蹄涡,并且会增加通道喉部区域的热负荷;随着质量流量比的增加,端壁气膜覆盖面积增大,而当射流质量流量比大于1%时,主流入侵现象消失,间隙射流将增强前缘马蹄涡,提高端壁前部的传热,并且减少端壁前部热负荷。随着间隙射流角度的增加,射流引起的分离涡增强,导致端壁前部的传热增强,而端壁气膜有效度降低,端壁热负荷增加。特别是在质量流量比为1.5%时,射流角度从30°增大到90°时,端壁平均气膜有效度减小53.4%。     

基于深度学习的航空发动机不平衡故障部位识别

针对基于机匣测点的航空发动机不平衡故障部位识别问题,提出了基于深度卷积神经网络的航空发动机不平衡故障部位诊断方法。针对某典型双转子航空发动机,建立整机耦合动力学模型,并利用数值积分算法实现不平衡故障数值仿真;在从发动机压气机端到涡轮端的高、低压转子上选择4个不平衡故障部位作为诊断对象,通过仿真分析得到发动机典型转速下的转子不同部位不平衡故障的仿真样本;计算4个机匣测点信号的规范化频谱,通过对大量仿真数据的处理得到反映不同不平衡故障部位的故障样本集;利用仿真得到的大量不平衡故障样本,训练深度卷积神经网络,利用深度卷积神经网络的优良特征学习能力实现航空发动机不平衡故障的不同部位进行识别,数值试验结果表明该方法对航空发动机不平衡故障部位的识别准确率达到95%。     

Softmax分类器深度学习图像分类方法应用综述

基于深度学习的人工智能图像分类方法研究是当前计算机视觉领域的研究热点。面向深度学习中的Softmax图像分类方法，首先回顾了图像分类技术的发展历程，接着介绍了图像识别技术中的分类器，并解释了Softmax回归函数的分类实现原理。基于Softmax回归分类器的应用，详细阐述了多种图像分类技术，具体包括浅层神经网络、深度置信网络、深度自编码器和卷积神经网络。同时，对比介绍了各种级联模型的具体结构、训练方法、实际应用、分类效果以及优缺点。最后，从Softmax回归分类器、深度学习网络模型和高维数据分类三个方面对基于Softmax回归分类器的深度学习模型在图像分类方面的发展与应用前景进行了展望。     

A point cloud deep neural network metamodel method for aerodynamic prediction

Aiming to reduce the high expense of 3-Dimensional (3D) aerodynamics numerical simulations and overcome the limitations of the traditional parametric learning methods, a point cloud deep learning non-parametric metamodel method is proposed in this paper. The 3D geometric data, corresponding to the object boundaries, are chosen as point clouds and a deep learning neural network metamodel fed by the point clouds is further established based on the PointNet architecture. This network can learn an end-to-end mapping between spatial positions of the object surface and CFD numerical quantities. With the proposed aerodynamic metamodel approach, the point clouds are constructed by collecting the coordinates of grid vertices on the object surface in a CFD domain, which can maintain the boundary smoothness and allow the network to detect small changes between geometries. Moreover, the point clouds are easily accessible from 3D sensors. The point cloud deep learning neural network, which employs re-sampling technique, the spatial transformer network and the fully connected layer, is developed to predict the aerodynamic characteristics of 3D geometry. The effectiveness of the proposed metamodel method is further verified by aerodynamic prediction and robust shape optimization of the ONERA M6 wing. The results show that the proposed method can achieve more satisfactory agreement with the experimental measurements compared to the parametric-learning-based deep neural network.     

基于深度学习的无人机航拍目标检测研究综述

目标检测是提高无人机（UAV）感知能力的关键技术之一,其研究对于无人机的应用有着重要意义。与基于手工特征的传统方法相比,基于卷积神经网络的深度学习方法具有强大的特征学习和表达能力,成为目前目标检测任务的主流算法。近年来,目标检测技术已经在自然场景图像上取得了一系列突破性进展,在无人机领域的研究也逐渐成为热点。首先系统阐述了基于深度学习的目标检测算法的研究进展,并总结了相关算法的优缺点。对常见的航空影像数据集进行了梳理并介绍了迁移学习的方法;从无人机影像背景复杂、目标较小、视场大、目标具有旋转性的特点出发,对无人机目标检测在近期的研究进行了归纳和分析。最后讨论了存在的问题和未来可能的发展方向。     

宽弦空心风扇转子叶片叶身结构设计参数分析

对空心叶片模型型腔和加强筋结构设计参数进行了分析;考虑到制造工艺可行性,采用ANSYS有限元软件对空心叶片模型进行了建模及应力分析。分析结果表明,加强筋数量增加,或加强筋与蒙皮厚度比增大,或加强筋扩散连接区域与非连接区域长度比增大,对蒙皮最大应力值影响不大,但对加强筋最大应力值影响较大。     

EXPERIMENTAL STUDY ON FILM COOLING EFFECTIVENESSAND HEAT TRANSFER COEFFICIENT BY NAPHTHALENE SUBLIMATION

ＮｏｍｅｎｃｌａｔｕｒｅＡ＝ｈｅａｔｏｒｍａｓｓｔｒａｎｓｆｅｒａｒｅａｃｐ＝ｓｐｅｃｉｆｉｃｈｅａｔａｔｃｏｎｓｔａｎｔｐｒｅｓｓｕｒｅＤ＝ｄｅｎｓｉｔｙｈ＝ｌｏｃａｌｃｏｎｖｅｃｔｉｏｎｈｅａｔｔｒａｎｓｆｅｒｃｏｅｆｉｃｉｅｎｔｈｄ＝ｌｏｃａｌ...     

基于混合神经网络模型的磨粒电荷检测方法研究

由于传统的滑油磨粒在线监测方法无法获取电荷分布位置信息,难以准确测量荷电颗粒数目及其携带的电荷量。为此,本文提出一种基于静电层析成像（Electrostatic tomography,EST）技术和深度学习算法的荷电颗粒检测方法。对EST传感器测量数据采用BP神经网络算法重建出测量截面上电荷的分布图像,采用卷积神经网络（Convolutional neural networks,CNN）算法分析重建图像以识别荷电颗粒数目,将识别的颗粒数目和传感器测量数据组合成输入向量,通过1个多层前馈网络确定带电颗粒数目、感应电荷值与颗粒电荷量值之间的映射关系,得到准确的各颗粒的电荷量值。实验结果表明：混合神经网络模型对数据样本的测量误差为9%,可满足滑油监测对于准确性的要求。     

基于多特征图像增强深度卷积神经网络的航天用电子元器件分类算法

为了实现航天用电子元器件的全自动及非接触识别,并减少由照明系统造成的图像亮度不均、偏色等问题对检测结果的影响,通过结合局部、区域和总体三个层次特征提升物体检测精度,提出了一种基于多特征图像增强深度卷积神经网络(MFIE-DCNN)的航天用电子元器件分类算法。MFIE-DCNN算法包含多特征学习和深度学习,其学习过程类似于人类视觉系统,能够对形状、方向和颜色特征进行深度挖掘,突出元器件边界信息,抑制背景杂波干扰。实验结果表明,该算法能够区分电路板板载元器件的种类,检测准确度优于传统算法。对比基于稀疏自动编码器的深度神经网络,检测结果提高了近20%。     

Analytical redundancy of variable cycle engine based on variable-weights neural network

In this paper, variable-weights neural network is proposed to construct variable cycle engine’s analytical redundancy, when all control variables and environmental variables are changing simultaneously, also accompanied with the whole engine’s degradation. In another word, variable-weights neural network is proposed to solve a multi-variable, strongly nonlinear, dynamic and time-varying problem. By making weights a function of input, variable-weights neural network’s nonlinear expressive capability is increased dramatically at the same time of decreasing the number of parameters. Results demonstrate that although variable-weights neural network and other algorithms excel in different analytical redundancy tasks, due to the fact that variable-weights neural network’s calculation time is less than one fifth of other algorithms, the calculation efficiency of variable-weights neural network is five times more than other algorithms. Variable-weights neural network not only provides critical variable-weights thought that could be applied in almost all machine learning methods, but also blazes a new way to apply deep learning methods to aeroengines.     

Regression model for civil aero-engine gas path parameter deviation based on deep domain-adaptation with Res-BP neural network

The variations in gas path parameter deviations can fully reflect the healthy state of aero-engine gas path components and units; therefore, airlines usually take them as key parameters for monitoring the aero-engine gas path performance state and conducting fault diagnosis. In the past, the airlines could not obtain deviations autonomously. At present, a data-driven method based on an aero-engine dataset with a large sample size can be utilized to obtain the deviations. However, it is still difficult to utilize aero-engine datasets with small sample sizes to establish regression models for deviations based on deep neural networks. To obtain monitoring autonomy of each aero-engine model, it is crucial to transfer and reuse the relevant knowledge of deviation modelling learned from different aero-engine models. This paper adopts the Residual-Back Propagation Neural Network (Res-BPNN) to deeply extract high-level features and stacks multi-layer Multi-Kernel Maximum Mean Discrepancy (MK-MMD) adaptation layers to map the extracted high-level features to the Reproduce Kernel Hilbert Space (RKHS) for discrepancy measurement. To further reduce the distribution discrepancy of each aero-engine model, the method of maximizing domain-confusion loss based on an adversarial mechanism is introduced to make the features learned from different domains as close as possible, and then the learned features can be confused. Through the above methods, domain-invariant features can be extracted, and the optimal adaptation effect can be achieved. Finally, the effectiveness of the proposed method is verified by using cruise data from different civil aero-engine models and compared with other transfer learning algorithms.     

Measurements of heat transfer and pressure in a trailing edge cavity of a turbine blade

 An experimental investigation is conducted to obtain the heat transfer and pressure drop data for an integral trailing edge cavity test section that simulates a novel turbine blade's internal cooling passage with bleed holes. Local heat transfer is measured on both the suction and pressure sides by a transient liquid crystal technique, while pressures at six positions are recorded by pressure calibrators. Moreover, flow characteristic and its effect on heat transfer are analyzed for conditions with or without bleed flow. The experimental results show that, in the cases with bleed flow, local heat transfer on the pressure side exceeds that on the suction side in the first and second channels. In the cases without bleed flow, in the first and third channels, local heat transfer on the suction side weakens whilst it increases significantly on the pressure side. For the second channel, non-bleed condition leads to a more balanced heat transfer distribution between the upstream and downstream channel. Besides, after the bleed holes are blocked, heat transfer in the first bend region on the suction side declines sharply, while the opposite phenomenon occurs for the second bend region on the pressure side. In both bleed and non-bleed cases, the total pressure of six measurement positions decreases continuously along the channel at the same Reynolds number and it promotes for higher Reynolds number. Among all the measurement points, under the same flow rate condition, the highest speed occurs at Position 5, which also shows the maximum difference between the total and static pressures. When the bleed holes are blocked, the total pressure at each measurement position appears to increase.