数据挖掘在运载火箭智能测试中的应用

针对运载火箭复杂系统的故障检测难以建立准确的数学模型的问题,研究了基于数据驱动的数据挖掘异常检测算法,对多种数据挖掘算法在运载火箭发动机异常检测的应用进行了研究和分析,提出了基于混合概率密度统计的多策略异常检测评价算法。该算法基于非监督学习的算法挖掘火箭发动机不同参数间的正常关联模型,火箭发动机早期的异常数据会引起正常关联模型的破坏,引入混合概率密度统计的多策略异常检测评价机制,可以有效屏蔽参数测量故障对系统故障检测的影响,从而更加准确给出系统异常程度。使用发动机历史试车数据作为样本进行特征模型的训练,使用一元、多元和混合概率密度模型对存在异常的发动机试车数据进行了实时异常检测的实验验证。实验结果表明,相比传统基于阈值和规则的异常检测算法,基于概率密度统计的多策略异常检测算法不仅可给出系统的正常和异常的状态,还可计算各参数和整个系统的异常值,为运载火箭进一步的故障诊断提供更加灵活的参考。     

基于支持向量机和卡尔曼滤波器的航空发动机故障诊断研究

针对航空发动机上可用传感器测量参数偏少情况下的健康参数估计问题,提出1种先分类后估计的方法。将传感器测量参数输入异常监测模块,对发动机工作状态进行监测,若监测结果为无故障则直接给出无部件故障的诊断结论;否则将测量参数输入最小二乘支持向量机(LSSVM),对部件故障进行分类,卡尔曼滤波器根据分类结果只对故障部件的健康参数进行估计。仿真结果表明:该方法可以减少需要估计的健康参数,提高估计精度。     

复杂曲面零件散乱点云特征点提取

复杂曲面零件的几何特征提取对加工质量的检测及逆向重构具有重要意义。提出了一种复杂曲面零件散乱点云特征点提取方法。首先,提出了基于高斯权重的邻域主成分分析的方法,通过估计每一点邻域的局部变化程度对点云模型进行初始标记;然后,采用基于标记的自动识别法实现特征点的提取;最后,通过特征点聚类的方式去除了特征点集中的异常点,完善了提取的点云特征。该方法直接操作于散乱点云,无需任何的拓扑连接信息。试验结果表明:该方法简单、有效,不需要过多地人为调节参数,特征点提取完整。     

Ionosphere-magnetosphere coupling

The large-scale electrical coupling between the ionosphere and magnetosphere is reviewed, particularly with respect to behavior on time scales of hours or more. The following circuit elements are included: (1) the magnetopause boundary layer, which serves as the generator for the magnetospheric-convection circuit; (2) magnetic field lines, usually good conductors but sometimes subject to anomalous resistivity; (3) the ionosphere, which can conduct current across magnetic field lines; (4) the magnetospheric particle distributions, including tail current and partial-ring currents. Magnetic merging and a viscous interaction are considered as possible generating mechanisms, but merging seems the most likely alternative. Several mechanisms have been proposed for causing large potential drops along magnetic field lines in the upper ionosphere, and many isolated measurements of parallel electric fields have been reported, but the global pattern and significance of these electric fields are unknown. Ionospheric conductivities are now thoroughly measured, but are highly variable. Simple self-consistent theoretical models of the magnetospheric-convection system imply that the magnetospheric particles should shield the inner magnetosphere and low-latitude ionosphere from most of the time-average convection electric field.     

基于联邦滤波的多源融合导航算法

在复杂多变环境下,单一导航源的定位性能和鲁棒性会受到一定的影响。针对汽车、小型飞行器在城市、峡谷、卫星信号缺失或被遮挡以及导航信息源繁多等情况,研究了基于联邦滤波的多源融合导航算法。该算法综合利用了各种不同的信息源,经过多传感器的高度集成、多信息源的数据融合,生成时空基准统一且具有抗干扰、连续、可靠的PNT服务信息。设计的联邦滤波器采用两级结构,在子滤波器中进行局部估计后,在主滤波器中进行最优合成。此外,每个子滤波器加入了故障诊断算法,且结合自适应滤波理论进行信息因子的自适应分配,有效提高了故障检测能力。最后,通过实验验证了不同信息源组合的有效性,表明所设计的基于联邦滤波的多源融合算法可以提供稳定、可靠以及高精度的多源融合定位服务,具有一定的研究意义和实际价值。     

Model-based robust estimation and fault detection for MEMS-INS/GPS integrated navigation systems

In micro-electro-mechanical system based inertial navigation system（MEMS-INS）/global position system（GPS） integrated navigation systems, there exist unknown disturbances and abnormal measurements. In order to obtain high estimation accuracy and enhance detection sensitivity to faults in measurements, this paper deals with the problem of model-based robust estimation（RE） and fault detection（FD）. A filter gain matrix and a post-filter are designed to obtain a RE and FD algorithm with current measurements, which is different from most of the existing priori filters using measurements in one-step delay. With the designed filter gain matrix, the H-infinity norm of the transfer function from noise inputs to estimation error outputs is limited within a certain range; with the designed post-filter, the residual signal is robust to disturbances but sensitive to faults. Therefore, the algorithm can guarantee small estimation errors in the presence of disturbances and have high sensitivity to faults. The proposed method is evaluated in an integrated navigation system, and the simulation results show that it is more effective in position estimation and fault signal detection than priori RE and FD algorithms.     

改进故障隔离的容错联邦滤波

为了解决重置模式下联邦滤波器中子系统故障对导航系统污染的问题,提出利用故障检测函数构建时变量测噪声的容错联邦滤波结构。通过将故障子滤波器等价为量测噪声趋于无穷大的正常系统,来取代传统的故障隔离方法;推导出了子滤波器对应的最优估计值,用以消除子滤波器估计次优性对故障检测的影响;采用动态信息分配系数,以减少故障信息对全局估计的影响。采用惯性/天文/景象/地形(INS/CNS/SMNS/TERCOM)的组合导航系统进行了仿真验证,结果表明该容错联邦滤波方法在子系统发生故障时的估计性能优于故障隔离方法。因此,所提方法具有提高故障子滤波器精度、保证无故障子滤波器鲁棒性以及全局估计精度的优势,具有较高的实用价值。     

A method for satellite time series anomaly detection based on fast-DTW and improved-KNN

In satellite anomaly detection, there are some problems such as unbalanced sample distribution, fewer fault samples, and unobvious anomaly characteristics. These problems cause the extisted anomaly detection methods are difficult to train accurate classification model, and the accuracy of anomaly detection is hard to improve. At the same time, the monitoring data of satellite has high dimension and is difficult to extract effective features. Based on the DTW over-sampling method, this paper realizes the over-sampling of fault samples in satellite time series, and constructs a distributed and balanced time series data set. The Fast-DTW method is applied to calculate the distance between different time series, which can improve the speed of similarity calculation. KNN (K-Nearest Neighbor) method is applied for classification and the best classification result is obtained by search the optimal hyper-parameters k. The results show that the proposed method has high anomaly detection accuracy and consumes short calculation time.     

Improving EGT sensing data anomaly detection of aircraft auxiliary power unit

The reliability of the on-wing aircraft Auxiliary Power Unit (APU) decides the cost and the comfort of flight to a large degree. The most important function of APU is to help start main engines by providing compressed air. Especially on the condition of sudden shutdown in the air, APU can offer additional thrust for landing. Therefore, its condition monitoring has drawn much attention from the academic and industrial field. Among the on-wing sensing data which can reflect its condition, Exhaust Gas Temperature (EGT) is one of the most important parameters. To ensure the reliability of EGT, one kind of data-driven anomaly detection framework for EGT sensing data is proposed based on the Gaussian Process Regression and Kernel Principal Component Analysis. The situations of one-dimensional and two-dimensional input data for EGT anomaly detection are considered, respectively. The cross-validation experiments are carried out by utilizing the real condition data of APU, which are provided by China Southern Airlines Company Limited Shenyang Maintenance Base. The anomalous stuck condition of EGT sensing data is also detected. Experimental results show that the proposed EGT sensing data anomaly detection method can achieve better performance of false positive ratio, false negative ratio and accuracy.     

Super resolution time delay estimation via MODE-WRELAX

We consider estimating time delays and amplitudes (real- or complex-valued) from the superposition of very closely spaced signals with known shapes. The well-known high resolution MODE (Method Of Direction Estimation) algorithm, which was originally proposed for angle estimation in array processing, is modified and used with our efficient time delay estimation method WRELAX (Weighted Fourier transform and RELAXation based) algorithm to deal with this problem. The proposed method is referred to as MODE-WRELAX. MODE-WRELAX provides better accuracy than MODE and higher resolution than WRELAX. Moreover, it can be used for both complex- and real-valued signals (including those with highly oscillatory correlation functions). Numerical results show that the MODE-WRELAX estimates can approach the corresponding Cramer-Rao bound. Efficient implementation of the algorithm is also discussed     

Stability of Parameter Estimates for a Gaussian Process

Maximum-likelihood estimates for the levels of the mean value function and the covariance function of a Gaussian random process are investigated. The stability of these estimates is examined as the actual covariance function of the process deviates from the form assumed in the estimators. It is found that the time-bandwidth product for stationary processes represents an upper bound on the number of estimator terms that can be safely used when estimating with uncertainty about the process covariance function. This result is consistent with other interpretations of the time-bandwidth product and tempers the conclusion that, in principle, an infinite number of estimator terms can be used to obtain a perfect estimate of the covariance level. In practice, the estimate of the level can never be perfect, and the accuracy of the estimate depends on the observation interval. Finally, conditions are established to ensure asymptotic stability of the estimates and physical interpretations are presented.     

基于多维航迹特征的异常行为检测方法

在信息融合领域,利用数据挖掘中的异常检测技术,可以基于目标的多维航迹特征来挖掘目标的异常行为。现有轨迹异常检测方法主要检测目标的位置异常,没有充分利用目标的属性、类型、位置、速度和航向等多维特征,在挖掘目标的异常行为时具有局限性。通过定义多因素定向Hausdorff距离和构造多维度局部异常因子,提出了一种基于多维航迹特征的异常行为检测方法,通过对多维航迹数据的异常检测,实现对目标异常行为的挖掘。在仿真军事场景和真实的民用场景上进行了实验分析,所提方法都能有效的检测出目标的异常行为。     

Impact of Geoid Improvement on Ocean Mass and Heat Transport Estimates

One long-standing difficulty in estimating the large-scale ocean circulation is the inability to observe absolute current velocities. Both conventional hydrographic measurements and altimetric measurements provide observations of currents relative to an unknown velocity at a reference depth in the case of hydrographic data, and relative to mean currents calculated over some averaging period in the case of altimetric data. Space gravity missions together with altimetric observations have the potential to overcome this difficulty by providing absolute estimates of the velocity of surface oceanic currents. The absolute surface velocity estimates will in turn provide the reference level velocities that are necessary to compute absolute velocities at any depth level from hydrographic data. Several studies have been carried out to quantify the improvements expected from ongoing and future space gravity missions. The results of these studies in terms of volume flux estimates (transport of water masses) and heat flux estimates (transport of heat by the ocean) are reviewed in this paper. The studies are based on ocean inverse modeling techniques that derive impact estimates solely from the geoid error budgets of forthcoming space gravity missions. Despite some differences in the assumptions made, the inverse modeling calculations all point to significant improvements in estimates of oceanic fluxes. These improvements, measured in terms of reductions of uncertainties, are expected to be as large as a factor of 2. New developments in autonomous ocean observing systems will complement the developments expected from space gravity missions. The synergies of in situ and satellite observing systems are considered in the conclusion of this paper. This revised version was published online in August 2006 with corrections to the Cover Date.     

On Stabilizing Role of Parallel Inhomogeneous Flow on Low-frequency Space Fluctuations

A fundamental reality throughout the space plasma is the existence of magnetic field-aligned flows. It is usually believed that the spatial transverse shear in the parallel flow destabilizes many low frequency oscillations and this may be the origin of low frequency oscillations in the ionosphere (V. V. Gavrishchaka et al., 1998, Phys. Rev. Lett., 80, 728 and Phys. Rev. Lett.: 2000, 85, 4285). Here we show that this notion of destabilizing influence of the shear in the parallel flow can be changed altogether if one takes the effect of the flow curvature (second spatial derivative) into account. The transverse curvature in the parallel flow can overcome the destabilizing influence of the shear and can render the low frequency modes stable. It is shown that unlike flow shear the effect of flow curvature is sign- and mode-dependent. This revised version was published online in August 2006 with corrections to the Cover Date.     

一种变信噪比条件下的恒虚警野值检测方法

野值检测又称异常值检测,是模式识别、机器智能和知识发现等领域经常面临的一个问题。当出现环境失配,数据信噪比（SNR）发生变化时,测试样本和训练样本所含噪声会有不同方差,以往的野值检测方法在虚警控制方面将会失效。针对这一问题,提出一种基于归一化残差（NR）的野值检测方法。该方法首先根据所需虚警概率和噪声方差变化情况确定野值检测门限,其次基于训练样本计算待考查模式的NR值,再比较NR值与检测门限的相对大小,从而判断待考查模式是否为野值。这一方法所依赖的检测门限对所需虚警率和噪声方差变化具有适应能力,因此可以在变信噪比条件下实现恒虚警（CFAR）野值检测。仿真实验验证了所提方法在虚警控制和野值检测方面的优越性能。     

一种非规则采样航空时序数据异常检测方法

航天器遥测数据的实时异常检测对于航天任务具有重要意义。以往方法大都考虑规则采样且缺失率较低的时序数据,然而航空时序数据具有维度大、噪声多、缺失率高、采样间隔不规则等特点,因此异常检测任务较为困难。针对非规则采样且具有缺失值的多维航空时序数据提出非规则采样多维时序数据异常检测（IMAD）算法。首先,采用带有可训练迟滞项的门控循环单元（GRU-D）对缺失值和非规则采样的时序数据进行建模;然后,采用变分自编码器建立随机性模型,学习正常时序数据的分布,从而对噪声数据具有鲁棒性;最后,利用基于极值理论的自适应阈值确定法确定合适阈值进行异常检测。结果显示,在两个真实航空时序数据集上,IMAD具有超出当前最新异常检测算法的性能;多个实验表明,IMAD在缺失率、参数以及数据集变化时,能够维持较好的异常检测效果,具有较强的鲁棒性。     

基于先验知识及其定量评估的自适应杂波抑制研究

 为了改善训练样本数受限的非均匀杂波环境中的系统检测性能,研究了基于先验知识及其定量评估的自适应杂波抑制算法。提出了使用经真实杂波信息白化后的先验杂波协方差矩阵与单位矩阵之差的谱范数,来定量评估杂波先验知识的准确程度,并给出了真实杂波协方差矩阵未知时的矩阵谱范数估计方法。结合先验知识定量评估结果,获得了具有先验知识约束时的杂波协方差矩阵最大似然估计方法。分别基于多脉冲相参雷达以及空时自适应雷达进行了杂波建模,在此基础之上分析了算法性能。仿真结果证实了该算法优于使用样本协方差矩阵及先验杂波信息形成杂波抑制权值的性能。     

An experimental method to obtain the hard alpha anomaly distribution for titanium alloy aeroengine disks

A probability-based damage tolerance methodology has been proposed to improve the recognition of material anomalies, especially hard alpha (TiN) anomalies for aeroengine rotor disks. A key input to this method is hard alpha anomaly distribution, which reflects the occurrence rate and size of anomalies present in the finished part material of titanium rotors. Since anomalies rarely occur naturally, an experimental method is proposed to obtain the anomaly distribution for titanium alloy aeroengine disks to reflect and equivalently replace the manufacturing development in titanium industry. In general, the anomaly distribution information can be divided into two parts: the Probability of Detection (POD) curve and the detected anomaly distribution, which contains the size and frequency data of the detected anomalies. The distribution can be established based on several appropriate assumptions and derivation steps with different confidence levels of POD curves and detected anomaly distributions. In this case, the distribution can be obtained in a relatively short time as a key input to the probability-based damage tolerance methodology. Then the Probability of Failure (POF) can be calculated, and the value is found to vary with different confidence levels. On this basis, the conservative estimated POF can be obtained in conjunction with confidence levels.