基于支持向量的迭代修正质心文本分类算法

针对质心分类算法容易产生归纳偏置或模型失配问题的不足,提出一种基于支持向量的迭代修正质心分类算法.该方法仅使用由支持向量机(SVMs,Support Vector Machines)选出的支持向量来构造质心向量,然后利用训练集误分样本来迭代修正初始质心向量.与其他分类算法相比,该算法取得较好的宏平均F₁和微平均F₁,在8个常用文本分类数据集上的实验验证了该算法的有效性,特别是在不均衡文本语料上.     

基于GPR模型的自适应平方根容积卡尔曼滤波算法

与传统算法一样,动态系统的参数化模型(含噪声统计特性)未知或不够准确易导致容积卡尔曼滤波(CKF)效果严重下降,甚至滤波结果发散.为此,利用高斯过程回归(GPR)方法对训练数据进行学习,得到动态系统的状态转移GPR模型和量测GPR模型以及噪声统计特性,用以替代或增强原有动态系统模型,并将其融入到平方根容积卡尔曼滤波(SRCKF)中,分别提出了无模型高斯过程SRCKF (MFGP-SRCKF)和模型增强高斯过程SRCKF (MEGP-SRCKF)两种算法.仿真结果表明:这两种新的自适应滤波算法提高了动态系统模型精度,且实时自适应调整噪声的协方差,克服了传统算法滤波性能易受系统模型限制的问题;与MFGP-SRCKF相比,在给定一个不够准确的参数化模型,且有限的训练数据未能遍布估计状态空间的情况下,MEGP-SRCKF具备更高的滤波精度.     

基于数理统计的航材备件预测方法研究

备件预测是航空公司备件订货和供货方零部件生产计划的基础,本文提出以数理统计方法来判断附件寿命分布形式,对附件寿命和故障率进行估算,并据此给出航材备件需求量及其保障率预测模型,并运用实例进行了验证。     

用于发动机涡轮盘叶片加工的夹具设计

介绍了用于加工发动机涡轮盘叶片夹具的设计准则,并给出了采用矢量三角函数和VC软件计算夹具极限偏差的公式及其判别依据,采用该准则设计的夹具用于涡轮盘叶片加工,产品满足设计要求,并通过了发动机热试车考核。     

机器人铣削加工动态变形误差预测与在线感知

航天舱段零件的机器人铣削加工中,强激励铣削力作用下的动态加工变形是影响加工精度的重要因素,复杂工况下动态误差的离线理论预测难以准确反映实际变形误差。为此,考虑加工过程中复杂工况下不确定性因素对动态变形误差的影响,本文提出了融合在线测量位姿、力数据的动态误差预测方法。首先,建立动态变形误差理论预测模型,引入加工误差预测偏差项;其次,结合有限实验数据对预测偏差的影响因素进行主成分分析(PCA),提取信息贡献率大的主成分作为特征向量,利用支持向量机(SVM)建立预测偏差关于特征向量的动态误差预测回归模型;最后,构建基于机器人关节位置和切削力测量数据的加工动态误差预测模型,实现动态误差在线感知,为加工精度控制提供数据基础。     

Unstable unsteady aerodynamic modeling based on least squares support vector machines with general excitation

Common, unsteady aerodynamic modeling methods usually use wind tunnel test data from forced vibration tests to predict stable hysteresis loop. However, these methods ignore the initial unstable process of entering the hysteresis loop that exists in the actual maneuvering process of the aircraft. Here, an excitation input suitable for nonlinear system identification is introduced to model unsteady aerodynamic forces with any motion in the amplitude and frequency ranges based on the Least Squares Support Vector Machines (LS-SVMs). In the selection of the input form, avoiding the use of reduced frequency as a parameter makes the model more universal. After model training is completed, the method is applied to predict the lift coefficient, drag coefficient and pitching moment coefficient of the RAE2822 airfoil, in sine and sweep motions under the conditions of plunging and pitching at Mach number 0.8. The predicted results of the initial unstable process and the final stable process are in close agreement with the Computational Fluid Dynamics (CFD) data, demonstrating the feasibility of the model for nonlinear unsteady aerodynamics modeling and the effectiveness of the input design approach.     

Soybean cultivation for Bioregenerative Life Support Systems (BLSSs): The effect of hydroponic system and nitrogen source

Soybean [Glycine max (L.) Merr.] is one of the plant species selected within the European Space Agency (ESA) Micro-Ecological Life Support System Alternative (MELiSSA) project for hydroponic cultivation in Biological Life Support Systems (BLSSs), because of the high nutritional value of seeds. Root symbiosis of soybean with Bradirhizobium japonicum contributes to plant nutrition in soil, providing ammonium through the bacterial fixation of atmospheric nitrogen. The aim of this study was to evaluate the effects of two hydroponic systems, Nutrient Film Technique (NFT) and cultivation on rockwool, and two nitrogen sources in the nutrient solution, nitrate (as Ca(NO₃)₂ and KNO₃) and urea (CO(NH₂)₂), on root symbiosis, plant growth and seeds production of soybean. Plants of cultivar ‘OT8914’, inoculated with B. japonicum strain BUS-2, were grown in a growth chamber, under controlled environmental conditions.     

Characteristics of the soil-like substrates produced with a novel technique combining aerobic fermentation and earthworm treatment

The soil-like substrate (SLS) technique is key for improving the closure of bioregenerative life support system (BLSS) by recycling the inedible biomass of higher plants. In this study, a novel SLS technique (NSLST) was proposed: aerobic fermentations at 35 °C for 1 day, then 60 °C for 6 days, finally 30 °C for 3 days, followed by earthworm treatment for 70 days. Comparing with the original SLS technique (OSLST), its process cycle was 13 days shorter, and the dry weight loss rate (81.1%) was improved by 24.77%. The cellulose and lignin degradation rates were 96.6% and 94.6%. The concentrations of available N, P and K in mature SLS were respectively 776.1 mg/L, 348.0 mg/L and 7943.0 mg/L. Low CH₄ and NH₃ production was observed, but no accumulation. According to the seed germination test, the SLSs were feasible for plant growth. This investigation will provide a preliminary foundation for BLSS design.     

Selection and hydroponic growth of potato cultivars for bioregenerative life support systems

As part of the ESA-funded MELiSSA program, Ghent University and the Université catholique de Louvain investigated the suitability, growth and development of four potato cultivars in hydroponic culture under controlled conditions with the aim to incorporate such cultivation system in an Environmental Control and Life Support System (ECLSS). Potato plants can fulfill three major functions in an ECLSS in space missions: (a) fixation of CO₂ and production of O₂, (b) production of tubers for human nutrition and (c) production of clean water after condensation of the water vapor released from the plants by transpiration. Four cultivars (Annabelle, Bintje, Desiree and Innovator) were selected and grown hydroponically in nutrient film technique (NFT) gullies in a growth chamber under controlled conditions. The plant growth parameters, tuber harvest parameters and results of tuber nutritional analysis of the four cultivars were compared. The four potato cultivars grew well and all produced tubers. The growth period lasted 127 days for all cultivars except for Desiree which needed 145 days. Annabelle (1.45 kg/m²) and Bintje (1.355 kg/m²) were the best performing of the four cultivars. They also produced two times more tubers than Desiree and Innovator. Innovator produced the biggest tubers (20.95 g/tuber) and Desiree the smallest (7.67 g/tuber). The size of Annabelle and Bintje potatoes were intermediate. Bintje plants produced the highest total biomass in term of DW. The highest non-edible biomass was produced by Desiree, which showed both the highest shoot and root DW. The manual length and width measurements were also used to predict the total tuber mass. The energy values of the tubers remained in the range of the 2010 USDA and Souci-Fachmann-Kraut food composition databases. The amount of Ca determined was slightly reduced compared to the USDA value, but close to the Souci-Fachmann-Kraut value. The concentration of Cu, Zn and P were high compared to both databases.     

深空探测航天器巡航段自主导航方法研究

对深空探测航天器自主导航方法进行了研究。为了应对深空探测中航天器轨道动力学模型的误差，在分光计测量航天器相对于太阳径向速度基础上，引入了小行星的视线矢量测量。通过最小二乘法计算出由小行星视线矢量所得到的位置信息，采用改进的信息融合方法修正扩展卡尔曼滤波中不精确的动力学模型造成的状态估计误差。同时计算了模型的能观度，对模型的可观性进行了分析。最后对算法进行了仿真分析，仿真结果表明，该算法对动力学模型的依赖性明显低于其他算法，在相同模型精度下，可获得更好的滤波精度。