基于模糊神经网络的导弹最优寻的末制导律

提出了一种适用于一类末制导段采用推力矢量控制的新型导弹的模糊神经网络最优寻的末制导律.在制导律设计时不仅要求导弹能量最省,脱靶量最小,同时考虑了推力矢量控制的非线性特点,并且为改善该神经网络系统的学习效果,在学习算法中还引入模糊学习规则.数字仿真表明所提出的模糊神经网络制导律对于机动目标具有较好的攻击能力.     

故障诊断神经网络系统的专家知识表达方法

通过介绍其它故障诊断技术存在的问题，研究了故障诊断神经网络系统，提出了两种故障诊断神经网络的专家知识表达方法，并将其应用于飞行器液压动力系统的故障诊断，仿真结果表明：提出的方法是正确可行的，并且专家知识表达方法对诊断结果有很大影响，该方法可推广应用于其它动力系统的故障诊断     

基于深度神经网络的航天器姿态控制系统故障诊断与容错控制研究

针对传统航天姿控系统故障诊断与容错控制诊断精度及控制分配效率较低的问题,提出了一种基于深度神经网络的航天器姿态控制系统故障诊断与容错控制方法。以控制力矩陀螺为执行机构的航天器发生执行机构故障工况时,所提出的方法可保证鲁棒的姿态控制。首先,利用三个异构深度神经网络实现传统容错控制器的故障诊断、姿态控制和力矩分配等功能,建立了全神经网络的智能自适应容错控制器架构。然后,对三个神经网络的网络层数、神经元数目和激活函数等参数进行优化调整,对比分析了神经网络参数对控制器性能的影响。最后,对所提出的新型控制器在控制力矩陀螺发生故障时的控制精度和鲁棒性进行了仿真验证。仿真结果表明,对于具有冗余控制力矩陀螺的航天器,提出的方法不仅能在单一陀螺故障下实现高精度的容错控制,也能在发生多陀螺故障时保证一定的姿态稳定控制。     

基于SCSO-BP神经网络的卫星姿态控制系统故障预测

近年来,随着人工智能的迅速发展,基于人工神经网络的卫星姿态控制系统故障预测方法得到了越来越多的重视。在反向传播(Back Propagation ,BP)神经网络中,权重和偏置是重要的可调节参数,与神经网络的预测性能密切相关。BP神经网络的初始权重和偏置为随机生成,设置不当容易导致网络在训练过程中陷入局部极值,进而影响预测性能。为了提高BP神经网络的预测性能,提出了一种将沙猫群优化(Sand Cat Swarm Optimization ,SCSO)算法与BP神经网络相结合的预测方法。在训练过程中,首先通过SCSO 算法对BP神经网络权重和偏置进行预训练,在此基础上,利用精调后的BP神经网络对卫星姿态控制系统周期渐变故障数据的未来趋势进行预测。实验结果表明,与原始BP神经网络预测方法相比,SCSO-BP预测方法能够有效减小预测误差,具有更好的预测精度。     

Plants + soil/wetland microbes: Food crop systems that also clean air and water

The limitations that will govern bioregenerative life support applications in space, especially volume and weight, make multi-purpose systems advantageous. This paper outlines two systems which utilize plants and associated microbial communities of root or growth medium to both produce food crops and clean air and water. Underlying these approaches are the large numbers and metabolic diversity of microbes associated with roots and found in either soil or other suitable growth media. Biogeochemical cycles have microbial links and the ability of microbes to metabolize virtually all trace gases, whether of technogenic or biogenic origin, has long been established. Wetland plants and the rootzone microbes of wetland soils/media also been extensively researched for their ability to purify wastewaters of a great number of potential water pollutants, from nutrients like N and P, to heavy metals and a range of complex industrial pollutants. There is a growing body of research on the ability of higher plants to purify air and water. Associated benefits of these approaches is that by utilizing natural ecological processes, the cleansing of air and water can be done with little or no energy inputs. Soil and rootzone microorganisms respond to changing pollutant types by an increase of the types of organisms with the capacity to use these compounds. Thus living systems have an adaptive capacity as long as the starting populations are sufficiently diverse. Tightly sealed environments, from office buildings to spacecraft, can have hundreds or even thousands of potential air pollutants, depending on the materials and equipment enclosed. Human waste products carry a plethora of microbes which are readily used in the process of converting its organic load to forms that can be utilized by green plants. Having endogenous means of responding to changing air and water quality conditions represents safety factors as these systems operate without the need for human intervention. We review this research and the ability of systems using these mechanisms to also produce food or other useful crops. Concerns about possible pathogens in soils and wastewater are discussed along with some methods to prevent contact, disease transmission and to pre-screen and decrease risks. The psychological benefits of having systems utilizing green plants are becoming more widely recognized. Some recent applications extending the benefits of plants and microbes to solve new environmental problems are presented. For space applications, we discuss the use of in situ space resources and ways of making these systems compact and light-weight.     

MIDACO on MINLP space applications

A numerical study on two challenging mixed-integer non-linear programming (MINLP) space applications and their optimization with MIDACO, a recently developed general purpose optimization software, is presented. These applications are the optimal control of the ascent of a multiple-stage space launch vehicle and the space mission trajectory design from Earth to Jupiter using multiple gravity assists. Additionally, an NLP aerospace application, the optimal control of an F8 aircraft manoeuvre, is discussed and solved. In order to enhance the optimization performance of MIDACO a hybridization technique, coupling MIDACO with an SQP algorithm, is presented for two of these three applications. The numerical results show, that the applications can be solved to their best known solution (or even new best solution) in a reasonable time by the considered approach. Since using the concept of MINLP is still a novelty in the field of (aero)space engineering, the demonstrated capabilities are seen as very promising.     

基于自适应神经网络的飞行器主动容错控制

针对飞行器非线性系统执行器故障,利用RBF神经网络和自适应控制律,提出了基于自适应神经网络的故障重构和容错控制方法。设计了自适应神经网络观测器,利用神经网络逼近故障,引入调节因子,设计自适应律以在线调整神经网络权重向量和中心向量。构造自适应神经网络控制器,结合神经网络设计补偿控制输入。利用Lyapunov稳定性定理证明了所提方法可以实现系统渐近稳定。仿真实验结果验证了所提的方法对故障系统具有良好的观测性能、控制精度和响应速度。     

机械臂的改进区间二型模糊神经网络控制

针对具有不确定性模型参数的双关节机械臂系统,提出基于改进区间二型模糊神经网络逼近器的自适应反演控制算法.相比于一型模糊系统,区间二型模糊系统由于自身的区间前件和隶属函数,更有效地处理高度非线性系统.然而现有的二型模糊寻找上下输出的交叉点过程中KM迭代算法计算量大、耗时高,使得传统的二型模糊系统不适用于实际控制应用.利用...     

基于神经网络的自由漂浮空间机械臂 自适应鲁棒控制

针对自由漂浮空间机械臂动力学模型难以精确获得,且无法表达为关于未知参数的线性形式问题,提出基于自适应神经网络的鲁棒控制方法.对于不确定性空间机械臂系统模型中存在的未知不确定部分,利用神经网络的万能逼近特性,设计神经网络控制器来补偿未知模型,避免传统控制中的保守上界估计;采用泰勒线性化技术将神经网络隐含层中的高斯函数线性化,设计包括网络权值、高斯中心及宽度在内的网络全参数自适应学习律,实现在线实时调整,提高控制精度;设计鲁棒自适应控制器来抑制外界扰动,并补偿逼近误差,提高系统鲁棒性;基于Lyapunov理论证明闭环系统的一致最终有界（UUB）.仿真试验表明所提控制方法能够获得较好控制效果,对空间机械臂控制具有一定工程应用价值.     

基于模型诊断技术的神经网络实现方法

针对基于模型的故障诊断流程中故障检测和故障识别两个关键问题,提出了一种基于神经网络的实现方法.首先利用BP神经网络进行参数估计,并结合系统模型进行故障检测;然后采用ART2神经网络进行数据聚类,并基于聚类结果进行系统故障识别;最后,设计实现了基于BP/ART2神经网络的故障诊断系统.基于BP神经网络的参数估计方法可以准确地估计诊断对象在不同状态下的参数,为故障检测提供有效依据;基于ART2神经网络的数据聚类不仅可以识别对象的已知故障类型,还可以识别出未知故障,对先验信息较少的系统进行故障识别更具有效性.通过永磁直流电机故障诊断案例的应用,证明方法能具有一定的工程实用性.     

空间晶体生长炉综合控制方法

空间晶体生长炉控制对象复杂, 具有纯滞后、强耦合、非线性等不利于控制的特性. 为实现炉温的精确控制, 提出了一个空间晶体生长炉综合控制方法并通过软件来实现. 将相关系数法和增广递推最小二乘法相结合, 辨识系统模型, 使用Hooke-Jeeves模式搜索法进行PID参数的自整定, 采用PID加前馈控制的算法对系统进行仿真, 使用同样的算法和控制参数对空间晶体生长炉进行实时控制, 并在控制的过程中实时调整控制参数, 以达到最佳的控制结果. 实验结果表明, 当三个温区的温度均达到700oC时, 温度稳定度小于±0.3oC, 满足系统设计要求. 同时介绍了实现综合控制软件的基本原理、功能、特点及在实际中的应用. 实际工程的应用结果表明, 该软件可满足空间晶体生长炉对控制系统的要求.      

基于自校正模糊神经控制的无刷直流传动系统

提出了一种自校正模糊神经网络控制器(SCFNNC)来实现无刷直流电动机起动、调速、制动等各运行阶段的性能指标.该SCFNNC是采用调整系统增益参数的方法完成较完善的控制规则的.重点研究了系统自校正增益参数的确定方法,模糊控制器的设计,人工神经网络实现模糊控制规则的方法等.自校正增益参数是根据系统对超调量、转速稳态误差、动态速降的期望值来确定的.设计模糊控制器时是根据系统的性能指标,确定出合适的模糊控制规则表,用于训练神经网络.为使系统的性能达到最佳,采用了自校正模糊神经控制、开关控制和比例控制相结合的复合控制方法,通过数学仿真证实配备SCFNNC的系统具有优良的动、静态特性,及较强的鲁棒性.     

神经网络理论在微分对策中的应用

阐述了微分对策理论在实际应用中存在的问题,并从动态冲突的实际过程出发,将神经网络理论与分层式智能控制理论相结合,来模型化微分对策问题.将微分对策的双向优化问题转化成基于神经网络理论的目标机动性辨识和单向优化控制问题,将微分对策的定性与定量结合起来,为有效加入人的经验提供了一种技术途径.并采用自适应分布式学习速率,大大提高了神经网络的学习速度和精度.     

Forecasting space weather: Can new econometric methods improve accuracy?

Space weather forecasts are currently used in areas ranging from navigation and communication to electric power system operations. The relevant forecast horizons can range from as little as 24 h to several days. This paper analyzes the predictability of two major space weather measures using new time series methods, many of them derived from econometrics. The data sets are the A_p geomagnetic index and the solar radio flux at 10.7 cm. The methods tested include nonlinear regressions, neural networks, frequency domain algorithms, GARCH models (which utilize the residual variance), state transition models, and models that combine elements of several techniques. While combined models are complex, they can be programmed using modern statistical software. The data frequency is daily, and forecasting experiments are run over horizons ranging from 1 to 7 days. Two major conclusions stand out. First, the frequency domain method forecasts the A_p index more accurately than any time domain model, including both regressions and neural networks. This finding is very robust, and holds for all forecast horizons. Combining the frequency domain method with other techniques yields a further small improvement in accuracy. Second, the neural network forecasts the solar flux more accurately than any other method, although at short horizons (2 days or less) the regression and net yield similar results. The neural net does best when it includes measures of the long-term component in the data.     

Development of autonomous control in a closed microbial bioreactor.

Space-based life support systems which include ecological components will rely on sophisticated hardware and software to monitor and control key system parameters. Autonomous closed artificial ecosystems are useful for research in numerous fields. We are developing a bioreactor designed to study both microbe-environment interactions and autonomous control systems. Currently we are investigating N-cycling and N-mass balance in closed microbial systems. The design features of the system involve real-time monitoring of physical parameters (e.g. temperature, light), growth solution composition (e.g. pH, NOx, CO2), cell density and the status of important hardware components. Control of key system parameters is achieved by incorporation of artificial intelligence software tools that permit autonomous decision-making by the instrument. These developments provide a valuable research tool for terrestrial microbial ecology, as well as a testbed for implementation of artificial intelligence concepts. Autonomous instrumentation will be necessary for robust operation of space-based life support systems, and for use on robotic spacecraft. Sample data acquired from the system, important features of software components, and potential applications for terrestrial and space research will be presented.     

空间微重力组合体及其控制

空间微重力组合体是绳系卫星系统的升级版,有更强大的功能和应用前景。可是在无控状态下,新旧两款系统的留位（系绳伸出长度不变）和回收（系绳回卷收缩）运动都不稳定,必须置入有效的控制算法才能使其稳定运行。从绳系卫星的简化动力学模型入手,抓住微重力和动态运动中的惯性力相互关系,推导出基础型和扩展型的距离速率控制算法,简明有效地抑制这些系统的不稳定运动和改进空间组合体在轨交会对接控制技术。该算法也可应用于有分布质量和弹性变形的连续系绳的动力学模型以及其他更复杂的绳系卫星（家族）系统的控制中。     

复杂深空通信网络的LTP传递时延建模及验证

针对简化场景下建立的LTP（Licklider Transmission Protocol）传递时延模型无法直接应用于复杂场景的问题，提出了一种复杂深空通信网络的LTP传递时延理论模型。已有研究多关注于一到两跳的简化场景或 bundle 保管开启状态下多跳场景数据传输模型的建立，然而其结果在复杂深空通信网络中，对于在 bundle 保管关闭状态下的深空通信并不适用。本文对LTP在复杂深空通信网络中的传输机制进行了分析，基于此分析对LTP在复杂深空通信网络中的传递时延进行了理论建模，搭建了仿真验证平台，仿真分析了模型的正确性及优势。结果表明，该模型计算结果与实验结果较为吻合，相比基于简化场景建立的理论模型，本文所提出的模型更加适用于复杂深空通信网络。      

基于神经网络的重力卸载系统控制器设计

吊挂系统是地面模拟空间机械臂重力卸载试验的重要方法之一.针对传统PID控制方式动作响应慢、鲁棒性差等缺点,提出了一种基于径向基函数(RBF)神经网络的智能控制方式.该方式有很强的非线性拟合能力,且学习规则简单,可映射任意复杂的非线性关系,便于计算机实现.利用该特性,设计了一种重力卸载精度较PID控制方式更高的控制器.该...     

空间目标一维距离像重构的线性规划方法

高分辨一维距离像是空间目标故障诊断与识别的重要手段。针对典型的空间目标 ,文章采用稀疏成份分析法研究了一维距离像的超分辨重构和特征提取问题。将处理实信号的基寻踪方法推广到复信号 ,并利用约束最小二乘解提出了一种次优的距离像重构算法 ,实现了空间目标散射中心的有效分离。数值结果验证了其合理性和有效性。     

多体卫星高稳定度智能控制方案研究

卫星上大型挠性天线的加减速运动会对星体产生较大的扰动,影响了星体姿态的指向精度和稳定度.运用Lyapunov稳定性理论,设计了变结构和神经网络控制器,并在星体前馈中引入扰动补偿力矩,从而保证星体姿态角速度在不确定性干扰下能以指数形式收敛到某一给定的有界范围内,仿真证明了姿态的稳定度满足给定的指标要求.