基于长短期记忆网络时序数据趋势预测及应用

为了研究状态监测大数据对设备运行状态的估计和预测,提出了一种人工经验与主成分分析相结合的长短期记忆网络方法（AEPCA-LSTM）,利用运行过程中的监测时序数据对设备运行趋势进行预测。首先,通过基于人工经验的主要成分分析方法（AEPCA）从状态监测系统中提取与目标变量最相关的状态变量作为输入;其次,利用长短期记忆网络（LSTM）对目标变量趋势变化进行预测,并考虑运行过程中新数据样本的持续产生,对模型进行定期更新,以提高模型的动态适应性。最后,将所提出的方法应用于船舶副机系统的涡轮增压器转速预测中,结果表明该方法相对于传统的PCA-LSTM和LSTM,具有更小的预测平均误差0.18037,即展现了其在时序数据趋势预测的优势。     

Phasing with near rectilinear Halo orbits: Design and comparison

In the next future, space agencies are planning to return to the Moon. The objective is to assemble an orbiting space station, called Gateway, on a Near Rectilinear Halo Orbit around the Moon as a base for future Moon and deep space missions. Within this framework, multiple side missions will be planned to sustain the Gateway (Artemis mission). The proposed work is thought in framework of the preliminary design of future cargo missions, in particular on the design of an efficient phasing trajectory, under the Circular Restricted Three body problem hypotheses, to bring a cargo vehicle from the end of the Earth-Moon transfer to the beginning of the proximity operations such as rendezvous and docking with the space station. The work aims covering the lack of literature in phasing trajectories with the NRHO by proposing three different strategies to connect the Earth-Moon transfer trajectory with the proximity operations. The three strategies are classified based on the choice of the parking orbits or the choice of the manifolds. Two strategies use butterfly and Halo orbits to park the vehicle before transferring to the target orbit. The third strategy, instead, uses manifolds to allow a direct phasing. In the paper, the three innovative strategies are designed and compare in a specific scenario.     

基于神经网络的预膜雾化模型

为了实现对预膜雾化质量的快速预估，基于神经网络建立了以结构参数和工况参数为输入,SMD值为输出的预膜雾化模型。使用试验数据训练和测试网络模型,结果显示,模型在较宽的工作范围内具有较高的精度和较好的泛化能力。SMD预测值随输入参数的变化规律符合试验结论,可作为给定工况参数下预膜式雾化装置的结构参数优化依据。神经网络能够较好地学习预膜雾化过程中的隐含规律。随着供油压差的增大,预膜器逐渐出现阻雾化效应,导致其雾化效果比单喷嘴和直混式雾化均差。     

An intelligent approach for flight risk prediction under icing conditions

Flight risk prediction is significant in improving the flight crew’s situational awareness because it allows them to adopt appropriate operation strategies to prevent risk expansion caused by abnormal conditions, especially aircraft icing conditions. The flight risk space representing the nonlinear mapping relations between risk degree and the three-dimensional commanded vector(commanded airspeed, commanded bank angle, and commanded vertical velocity) is developed to provide the crew with practi...     

飞机飞行振动预计技术

介绍了飞机飞行振动的主要来源和特征,分析了传统预计方法的不足,针对飞机飞行振动与高度、马赫数、攻角等飞行参数呈非线性关系的特点,提出了基于BP神经网络的飞行振动预计新技术,建立了预计模型,通过8组建模样本训练网络,确定了预计模型中的各个参数值。预计结果和实测飞机飞行振动信号比较分析表明:该方法预计精度高,验证了BP神经网络预计飞机飞行振动的可行性。     

An air-rail inter-modal strategy for aircraft recovery

It is common that airlines encounter a disruption of a flight schedule, which is mainly caused by resource shortages. In case of a disruption, subject to scarce resources, most of airlines lose flexibility of performing aircraft recovery on the basis of business interest priorities and need to delay, swap, and cancel flights. This paper proposes an air-rail inter-modal strategy to incorporate a High-Speed Rail(HSR) transport mode into an aviation network for aircraft recovery. The air-rail inter...     

基于BP神经网络的导弹保障人员可靠性分析

文章分析了导弹武器系统的保障可靠性,选取了影响保障人员可靠性的 5个因素作为输入,建立了导弹保障可靠性的神经网络模型；运用 BP算法训练数据,得出各影响因素的重要度排序；并针对提高导弹保障人员可靠 性提出了一系列解决措施。     

基于动力学模型的旋翼动平衡故障仿真及诊断

旋翼桨叶质量不平衡造成旋翼动不平衡,从而引起直升机振动。针对旋翼动不平衡故障,建立直升机动力学模型,对桨叶质量不平衡进行故障仿真及分析,建立质量不平衡故障与调整配重的对应关系；进而提出一种 BP神经网络和遗传算法结合的旋翼调整方法,建立输入参数与桨叶配重之间的模型,将四片桨叶的挥舞角和机体横滚、俯仰 2个方向的加速度值及相位作为网络输入,通过学习训练,根据输入数据预测调整配重,从而减小直升机 振动,解决旋翼动不平衡问题。     

基于多层次特征的高分影像海岸带地物分类

针对高分辨率影像地物光谱丰富复杂的问题，文章提出了一种结合像元-对象-空间格局特征的水域、盐田等海岸带精细地物分类方法。首先基于像元和空间格局特征分别利用遗传算法优化随机森林参数进行分类；其次将两种分类结果采用D-S证据理论融合；然后基于不同地物斑块的空间上下文信息，采用粒子群算法优化支持向量机（Support Vector Machine，SVM）参数实现面向对象的二次分类。为验证方法的有效性，收集山东省莱州湾海岸带附近区域的“高分二号”卫星影像，开展了水域及水利设施用地、盐田、工矿用地、交通运输用地、住宅用地、沿海滩涂、海域与耕地8类地物的分类实验，总体精度达到97.21%，Kappa系数为0.964 2。实验结果表明，文章提出的高分影像海岸带地物分类方法兼顾像元、对象及空间格局特征优点，能够实现多层次细致分类，并同时兼具较短的训练时长与较高的分类精度。该研究成果可为中国海岸带生态环境监测与土地利用规划提供一定技术支撑。     

基于SAN构建高可用实时存储系统

介绍SAN网络存储技术及其拓扑结构,针对某飞控中心数据存储系统的实际需求,设计和实现一个SAN存储结构上的高可用实时可扩展集群存储方案,并详细描述通过软、硬件结合实现冗余数据路径、故障切换、实时存储和查询的方法。