基于强化学习的航空器机场智能静态路径规划

随着人工智能迅速发展以及“智慧机场”的提出,研究人工智能在机场如何有效地辅助机场管制人员,驾驶员指挥航空器在地面滑行具有重要意义。本文提出一种基于强化学习的滑行路径规划方法,构建航空器机场地面强化学习移动模型,并以海口美兰机场为案例采用 Python 内置工具包 Tkinter 进行场面仿真;在此基础上,考虑机场航空器滑行规则,采用 Off-Policy 中 Q-Learning 算法求解贝尔曼方程,实现航空器在 Model-based 环境中进行静态路径规划。结果表明：本文所提方法能够实现停机位到跑道出口智能静态路径规划     

基于小波包分析与多核学习的滚动轴承故障诊断

为了更准确地诊断滚动轴承故障,提出了一种基于小波包分析与多核学习的滚动轴承故障诊断方法.该方法首先对振动信号进行3层小波包分解,将振动信号分解为不同频带的信号,提取各频带的相对能量特征,构建特征向量;然后采用多核学习算法从训练样本集中学习核函数与分类器;最后使用训练出的分类器识别滚动轴承故障类型.为了验证方法的有效性,进行了滚动轴承故障诊断实验,实验结果表明该方法的故障诊断准确率达到98.25%,与传统的基于小波包与支持向量机的滚动轴承故障诊断方法相比,其故障诊断准确率更高,同时由于避免了核函数的选择问题,该方法更便于实际应用.      

月球取样返回器回收半实物仿真系统研究

月球取样返回器回收半实物仿真系统是通过模拟返回器取压孔附近压力环境,将回收程控装置硬件实物接入仿真系统并构成实时仿真回路的半实物仿真平台。文章介绍了月球取样返回器回收半实物仿真系统的研制意义,对系统总体结构和设计方案、网络环境、运行流程和新的设计特点进行概括,在面向对象的基础上建立了分阶段的月球取样返回器降落伞回收系统整个工作过程的动力学模型,利用空投试验测量数据对模型的仿真精度进行验证分析,并对半实物仿真系统的实时性和压力模拟精度进行了分析。结果表明,半实物仿真系统的各项技术参数均达到设计指标,可为中国探月工程三期中月球取样返回器回收分系统的可靠性评估提供重要的技术支持。     

用于深层月壤采样返回的软质取心袋的设计与测试验证

月壤钻探采样技术尤其无人自主采样返回技术是我国探月工程的关键技术之一。文章提出了一种适于深层月壤钻探采样的取心机构,能够实现对月壤样品的原位包裹,以保持样品的层理特性信息。该机构选用Kevlar作为包裹样品用的软质取心袋材料,采用满足要求的编织工艺织造出致密薄壁的软袋,并对软袋进行了翻转抽拉力、力学性能及涂层测试分析等测试试验,验证其可满足深层月壤钻探采样任务的需求。     

Roadmap to a human Mars mission

We propose a new roadmap for the preparation of the first human mission to Mars. This proposal is based on the work of ISECG and several recent recommendations on human Mars mission architectures. A table is proposed to compare the possible benefits of different preparatory missions. Particular attention is paid to the possibility of qualifying important systems thanks to a heavy Mars sample return mission. It is shown that this mission is mandatory for the qualification of Mars aerocapture at scale-1, EDL systems at scale 1 and Mars ascent. Moreover, it is a good opportunity to test many other systems, such as the heavy launcher and the transportation systems for the trips beyond LEO. These tests were not mentioned in the last ISECG report. This strategy is facilitated in the case of the simplified Mars mission scenarios that have recently been presented because it is suggested that relatively small vehicles with small crew sizes are used in order to optimize the payload mass fraction of the landing vehicles and to avoid the LEO assembly. An important finding of the study is that a human mission to the surface of the Moon is not required for the qualification of the systems of a human mission to Mars. Since affordability is a key criterion, two important missions are proposed in the roadmap. The first is a heavy Mars sample return mission and the second is a manned mission to a high Earth orbit or eventually to the vicinity of the Moon. It is shown that both missions are complementary and sufficient to qualify all the critical systems of the Mars mission.     

利用神经元拓展正则极端学习机预测时间序列

为实现对于时间序列预测数据的准确预测,提出一种神经元拓展正则极端学习机(NERELM,Neuron-Expanding Regularized Extreme Learning Machine),并研究了其在时间序列预测中的应用.NERELM根据结构风险最小化原理权衡经验风险与结构风险,以逐次拓展隐层神经元的方式自动确定最佳的网络结构,以避免传统神经网络训练过程中需人为确定网络结构的弊端.应用于时间序列的仿真结果表明:NERELM可有效实现对于RELM最佳网络结构的自动确定,具有预测精度高与计算速度快的优点.      

LEVEL DECISION-MAKING FOR AERO-ENGINE MODULES BASED ON INCREMENTAL KNOWLEDGE LEARNING

The maintenance of an aero-engine usually includes three levels, and the maintenance cost and period greatly differ depending on the different maintenance levels. To plan a reasonable maintenance budget program, airlines would like to predict the maintenance level of aero-engine before repairing in terms of performance parame- ters, which can provide more economic benefits. The maintenance level decision rules are mined using the histori- cal maintenance data of a civil aero-engine based on the rough set theory, and a variety of possible models of upda- ting rules produced by newly increased maintenance cases added to the historical maintenance case database are in- vestigated by the means of incremental machine learning. The continuously updated rules can provide reasonable guidance suggestions for engineers and decision support for planning a maintenance budget program before repai- ring. The results of an example show that the decision rules become more typical and robust, and they are more accurate to predict the maintenance level of an aero-engine module as the maintenance data increase, which illus- trates the feasibility of the represented method.     

基于预测扭矩前馈的涡轴发动机内模控制方法

基于内模原理设计了涡轴发动机功率涡轮转速控制器.针对主旋翼扭矩变化对功率涡轮转速的干扰,提出了一种基于极端学习机的扭矩预测方法.极端学习机训练基于动态仿真数据,其输入通过相关分析获得.基于内模原理的功率涡轮转速控制器采用极点配置的设计方法,将输入信号的内模直接加入控制器,实现鲁棒跟踪.扭矩前馈采用比例微分(PD)控制策略,实现对发动机负载变化干扰的有效补偿.数字仿真结果表明:极端学习机扭矩预测精度高,扭矩相对误差小于1.5‰,与不加前馈控制相比,所提出的控制方法减小了机动飞行过程中功率涡轮转速的超调或下垂30%以上.      

样本重复使用失效响应曲线分析结构可靠度方法

用基于样本重复使用失效响应曲线逼近法,为充分利用迭代过程中的样本点,提出了样本重复使用的思想,并由此发展出一种结构可靠度的计算方法.利用可靠度指标作为收敛条件,从迭代收敛后的所有样本点中选取靠近失效界线处的样本点构造失效曲线响应函数,最后通过蒙特卡洛抽样的方法计算可靠度.多个算例表明:该响应函数避免了对传统意义上的功能函数响应面的模拟,降低了模型维数,又由于适当地 引入了带交叉项、具有曲线旋转功能的响应函数,因此能够很好地逼近失效曲线,有效地提高可靠度计算精度.