基于滑动时窗策略自适应优化支持向量机的航空发动机性能参数在线预测

针对传统航空发动机性能参数时间序列预测方法存在的不足,提出了基于滑动时窗策略自适应优化支持向量机(Support Vector Machine,SVM)在线预测模型。该方法解决了训练样本动态适应性差的特点和老旧数据信息影响预测模型精度的问题。在该方法中,滑动时窗策略实时更新时窗数据训练样本,最终误差预报准则(Final Prediction Error,FPE)自适应地确定嵌入维数,遗传算法(Genetic Algorithm,GA)则实时自适应优化SVM建模参数。应用航空发动机排气温度偏差值(Delta Exhaust Gas Temperature,DEGT)数据进行实例验证,结果表明基于滑动时窗策略的自适应GA优化的SVM (GASVM)在线预测模型比传统的GASVM预测模型预测精度有显著提高。进一步分析了预测模型不同时窗宽度对短期预测精度的影响,展示了1步～10步预测的效果,结果表明在线预测模型在不同时窗宽度下短中期(5步以内)预测效果良好且稳定。文中提出的在线预测模型可用于航空发动机性能参数的预测,实现对航空发动机未来性能变化的预警。     

基于支持向量机方法的发动机性能趋势预测

为了提高对航空发动机性能趋势预测的精度，提出利用支持向量机方法来预测表征发动机整体性能的参数一性能综合指数。建立了基于支持向量回归的一步及多步预测模型，利用该模型对性能正常衰退及性能异常发动机的综合指数分别进行预测，并与自回归(AR)模型的预测值进行比较。结果表明，基于支持向量机的预测模型比AR模型的预测精度更高，其四步预测精度由80．56％提高到88．51％。因此该模型尤其适合中、长期预测。     

一种新的导航星选取算法研究

在导航星表的建立过程中，由于恒星的数量太多，往往要进行筛选，通常这种选择复杂费时，而结果往往并不是最优的。本文引入了动态星等阈值分布函数，将传统星等阈值过滤算法中的静态阈值用动态星等阈值代替，建立了一种新的动态星等阈值过滤选择模式。而基于统计学习理论的支持向量机方法为求解高维非线性动态星等阈值分布函数提供了新的途径。本文讨论了这种基于支持向量机的导航星自动选择算法——回归选取算法。实验表明，用该算法所选取的导航星表，导航星数量少、分布均匀性好。同时它还能适应多种任务的导航星选取要求，具有很强的通用性。     

单脉冲雷达导引头抗压制式干扰时域仿真

为研究单脉冲主动雷达导引头在自卫干扰(SSJ)和非SSJ条件下的抗干扰性能,建立了单脉冲角跟踪系统对SSJ和非SSJ压制式噪声干扰源进行跟踪的时域模型。仿真结果表明,在SSJ条件下,角跟踪回路的抗压制式噪声干扰性能较佳,跟踪情况与目标视线角速度有关;在非SSJ条件下,当导引头接收天线处远距支援干扰(SOJ)强于目标回波信号时,SOJ会引入较大的天线指向偏差。所建模型可用于雷达导引头抗干扰性能的仿真研究。     

基于引力球结构支持向量机多类算法的涡轮泵故障诊断

在涡轮泵故障诊断中，多类故障诊断是经常出现的问题。为提高多类故障诊断速度，在球结构支持向量机的基础上，提出一种引力球结构的支持向量机多类算法，该算法充分考虑样本分布疏松程度，经过试验优化分析得到最佳分类引力公式。用该算法和其他常用算法对涡轮泵仿真故障进行分类比较，结果表明基于引力球结构的支持向量机故障诊断算法学习速度快，诊断效果好。     

Hybrid calibration method for six-component force/torque transducers of wind tunnel balance based on support vector machines

A hybrid calibration approach based on support vector machines (SVM) is proposed to characterize nonlinear cross coupling of multi-dimensional transducer. It is difficult to identify these unknown nonlinearities and crosstalk just with a single conventional calibration approach. In this paper, a hybrid model comprising calibration matrix and SVM model for calibrating linearity and nonlinearity respectively is built up. The calibration matrix is determined by linear artificial neural network (ANN), and the SVM is used to compensate for the nonlinear cross coupling among each dimension. A simulation of the calibration of a multi-dimensional sensor is conducted by the SVM hybrid calibration method, which is then utilized to calibrate a six-component force/torque transducer of wind tunnel balance. From the calibrating results, it can be indicated that the SVM hybrid calibration method has improved the calibration accuracy significantly without increasing data samples, compared with calibration matrix. Moreover, with the calibration matrix, the hybrid model can provide a basis for the design of transducers.     

Silkworms culture as a source of protein for humans in space

This paper focuses on the problem about a configuration with complete nutrition for humans in a Controlled Ecological Life Support System (CELSS) applied in the spacebases. The possibility of feeding silkworms to provide edible animal protein with high quality for taikonauts during long-term spaceflights and lunar-based missions was investigated from several aspects, including the nutrition structure of silkworms, feeding method, processing methods, feeding equipment, growing conditions and the influences on the space environmental condition changes caused by the silkworms. The originally inedible silk is also regarded as a protein source. A possible process of edible silk protein was brought forward in this paper. After being processed, the silk can be converted to edible protein for humans. The conclusion provides a promising approach to solving the protein supply problem for the taikonauts living in space during an extended exploration period.     

Low light intensity effects on the growth,photosynthetic characteristics,antioxidant capacity,yield and quality of wheat (Triticum aestivum L.) at different growth stages in BLSS

Minimizing energy consumption and maximizing crop productivity are major challenges to growing plants in Bioregenerative Life Support System (BLSS) for future long-term space mission. As a primary source of energy, light is one of the most important environmental factors for plant growth. The purpose of this study is to investigate the effects of low light intensity at different stages on growth, pigment composition, photosynthetic efficiency, biological production and antioxidant defence systems of wheat (Triticum aestivum L.) cultivars during ontogenesis. Experiments were divided into 3 intensity-controlled stages according to growth period (a total of 65 days): seedling stage (first 20 days), heading and flowering stage (middle 30 days) and grain filling stage (last 15 days). Initial light condition of the control was 420 μmol m⁻² s⁻¹ and the light intensity increased with the growth of wheat plants. The light intensities of group I and II at the first stage and the last stage were adjusted to the half level of the control respectively. For group III, the first and the last stage were both adjusted to half level of the control. During the middle 30 days, all treatments were kept the same intensity. The results indicated that low-light treatment at seedling stage, biomass, nutritional contents, components of inedible biomass and healthy index (including peroxidase (POD) activity, malondialdehyde (MDA) and proline content) of wheat plants have no significant difference to the control. Furthermore, unit kilojoule yield of group I reached 0.591 × 10⁻³ g/kJ and induced the highest energy efficiency. However, low-light treatment at grain filling stage affected the final production significantly.     

Developing a dust storm detection method combining Support Vector Machine and satellite data in typical dust regions of Asia

Enhancing the dust storm detection is a key part for the environmental protection, human healthy and economic development. The goal of this paper is to propose a new Support Vector Machine (SVM)-based method to automatically detect dust storms using remote sensing data. Existing methods dealing with this problem are usually threshold-based that are of great complexity and uncertainty. In this paper we propose a simple and reliable method combining SVM with MODIS L1 data and explore the optimal band combinations used as the feature vectors of SVM. The developed method was evaluated by MODIS and OMI data qualitatively and quantitatively on three study sites located in the Arabian Desert, Gobi Desert and Taklimakan Desert, and it was also compared to three other traditional methods based on their accuracy, complexity, reliability and sensitivity to thresholds. The detection results demonstrated that the combination of (Band7 − Band3)/(Band7 + Band3) ((B7 − B3)/(B7 + B3)), Band20 − Band31 (B20 − B31), and Band31/Band32 (B31/B32) can detect the dust storms more precisely than other individual bands or their combination. The comparison among those cases indicated that the proposed automatic method exhibited an advantage of minimizing the uncertainty and complexity, which were the limits of defining thresholds based on the threshold-based methods. The conclusions can provide references for studies that focus on statistical-based dust storm detection.     

Improving orbit prediction accuracy through supervised machine learning

Due to the lack of information such as the space environment condition and resident space objects’ (RSOs’) body characteristics, current orbit predictions that are solely grounded on physics-based models may fail to achieve required accuracy for collision avoidance and have led to satellite collisions already. This paper presents a methodology to predict RSOs’ trajectories with higher accuracy than that of the current methods. Inspired by the machine learning (ML) theory through which the models are learned based on large amounts of observed data and the prediction is conducted without explicitly modeling space objects and space environment, the proposed ML approach integrates physics-based orbit prediction algorithms with a learning-based process that focuses on reducing the prediction errors. Using a simulation-based space catalog environment as the test bed, the paper demonstrates three types of generalization capability for the proposed ML approach: (1) the ML model can be used to improve the same RSO’s orbit information that is not available during the learning process but shares the same time interval as the training data; (2) the ML model can be used to improve predictions of the same RSO at future epochs; and (3) the ML model based on a RSO can be applied to other RSOs that share some common features.