广义回归神经网络在日长变化预报中的应用

传统的日长变化预报多是基于线性模型,如最小二乘模型、自回归模型等,但是日长变化包含了复杂的非线性因素,线性模型预报的效果往往不甚理想.所以尝试使用一种非线性神经网络—广义回归神经网络(GRNN)模型进行日长变化预报,并将结果与使用BP (Back Propagation)神经网络模型和其它模型的预报结果进行比较.结果表明,GRNN用于日长变化预报是高效可行的.     

遗传算法优化的BP神经网络卫星钟差预报

针对BP (Back Propagation)神经网络模型预测卫星钟差中权值和阈值的最优化问题, 提出了基于遗传算法优化的BP神经网络卫星钟差短期预报模型, 给出了遗传算法优化BP神经网络的基本思想、具体方法和实施步骤. 为验证该优化模型的有效性和可行性, 利用北斗卫星导航系统(BeiDou navigation satellite system, BDS)卫星钟差数据进行钟差预报精度分析, 并将其与灰色模型(GM(1,1))和BP神经网络模型预报的结果比较分析. 结果表明: 该模型在短期钟差预报中具有较好的精度, 优于GM(1,1)模型和BP神经网络模型.     

基于极限学习机的极移中长期预报

传统的极移预报多是基于最小二乘外推和自回归等线性模型,但极移包含了复杂的非线性成分,线性模型的预报效果往往不甚理想。将一种新型神经网络极限学习机(Extreme Learning Machine, ELM)用于极移中长期预报。首先利用最小二乘外推模型对极移序列进行拟合,获得趋势项外推值,然后采用极限学习机对最小二乘拟合残差进行预报,最终的极移预报值为趋势项外推值与残差预报值之和。将极限学习机的预报结果同反向传播(Back Propagation, BP)神经网络与地球定向参数预报比较活动(Earth Orientation Parameters Prediction Comparison Campaign, EOP PCC)的预报结果进行对比,结果表明,极限学习机用于极移中长期预报是高效可行的。     

基于Takagi-Sugeno模糊神经网络模型的卫星钟差预报方法

卫星钟差预报精度的不断提升是精密导航的关键问题.为了进一步提高钟差的预报精度和更好地反映钟差的变化特性,提出一种基于Takagi-Sugeno模糊神经网络(Fuzzy Neural Network,FNN)的钟差预报方法.该方法首先根据钟差数据的特点对钟差进行预处理,然后以预处理后的数据建立一种高精度预报钟差的Takagi-Sugeno模糊神经网络算法.采用IGS(International Global Navigation Satellite System Service)不同采样间隔的精密钟差数据进行了短期预报试验,并与ARIMA(Auto-Regressive Integrated Moving Average)模型、GM(1,1)模型及QP(Quadratic Polynomial)模型进行了对比试验,分析结果表明:对不同类型原子钟,该方法用于钟差短期预报是可行的、有效的,其获得的卫星钟差预报结果明显优于常规方法.     

应用神经网络技术预报海表温度变化异常

介绍了神经网络技术在时间序列模拟和预报中的应用要点;应用神经网络技术对Nino3.4海区海表温度变化异常(SSTA)进行了实验预报,每次给出1～12个月预报值;将预报值与SSTA实测值以及其它模型的预报值进行了比较。结果表明,在1990年1月～2007年10月期间,SSTA预报值与实测值的均方误差为±0.76℃。将2004年10月～2007年10月期间该方法的预报结果与美国气候预测中心(CPC)发布的各种模型的预报结果(只给出第2～10月预报值)进行比较,结果表明这一期间该方法预报值相对实测值的均方误差为±0.86℃,精度略好于另外一个神经网络模型的预报精度(±0.98℃)。     

基于LS+AR模型的三种方式预报地球自转参数的比较

为研究最小二乘(least squares,LS)模型和自回归(autoregression,AR)模型的组合(LS+AR)方法用于地球自转参数(Earth rotation parameters,ERP)的预报时,不同的预报方式对预报结果的影响,我们采用递推、迭代和间隔这3种预报方式对ERP进行预报。结果表明,这3种方式对日长变化(length of day,LOD)所有跨度预报的精度相当,而递推方式在极移所有跨度的预报上表现出精度优势,间隔方式次之,迭代方式最差。在数据利用率和计算速度方面,递推和迭代方式的数据利用率高,但前者的计算量明显小于后者,而间隔方式的数据利用率低,但计算速度最快。     

样本输入方式对极端学习机预报日长变化的影响

针对极端学习机(Extreme Learning Machine,ELM)用于日长(Length-Of-Day,LOD)变化预报过程中,样本输入方式对预报结果的影响进行了研究。采用跨度、连续和迭代3种样本输入方式对日长变化进行预报。结果表明,不同的样本输入方式对预报结果有很大影响,样本按跨度输入的预报精度最低;样本采用连续输入方式在短期和中长期预报中预报精度较高,但计算速度较慢,较适合中长期预报;样本按迭代输入方式的短期预报精度稍优于连续输入方式,而中长期预报精度则不如连续输入方式,但具有较高的预报效率。这对于日长变化的实时快速预报有着较高的现实意义。     

顾及最小二乘拟合端点效应的日长变化预报

针对日长(Length Of Day,LOD)变化预报中最小二乘(Least Squares,LS)拟合存在端点效应的问题,采用时间序列分析方法对日长变化序列进行端点延拓,形成一个新序列,然后用新序列建立最小二乘模型,最后再结合最小二乘模型和自回归(Autoregressive,AR)模型对原始日长变化序列进行预报。实验结果表明,在日长变化序列两端增加统计延拓数据,能有效减小最小二乘拟合序列的端点畸变,从而提高日长变化的预报精度,尤其对中长期预报精度提高明显。     

跳步自回归模型在中长期日长变化预报中的应用

利用自回归模型进行日长变化中长期预报时,预报精度逐渐降低.跳步自回归模型在中长期的预报中具有良好的预报精度,且具有较好的预测稳定性.因此,尝试采用跳步自回归模型替代自回归模型进行日长变化预报.最后,利用国际地球自转参数与参考系服务(International Earth Rotation and Reference Systems Service,IERS)提供的EOP08 C04日长变化序列进行实验,并分析比较两种模型的预报结果.实验结果表明,跳步自回归模型用于改善自回归模型中长期预报精度是可行有效的.     

基于小波分析和神经网络的卫星钟差预报性能分析

为了有效地进行卫星钟差预报和更好地反映卫星钟差特性,提出了一种基于小波分析和神经网络的4阶段混合模型来实现卫星钟差的预报,并给出了基于小波分析和径向基函数(Radial Basis Function,RBF)神经网络进行卫星钟差预报的基本思想、预报模型和实施步骤.采用"滑动窗"划分数据,利用神经网络预测小波分解和去噪后的钟差序列各层系数,更精确地把握钟差序列复杂细致的变化规律,从而更好地逼近钟差序列.为验证该混合预报模型的可行性和有效性,利用GPS卫星钟差数据进行钟差预报精度分析,并将其与灰色系统模型和神经网络模型进行比较分析.仿真结果显示,该模型具有较好的预报精度,可为实时GPS动态精密单点定位提供较高精度的卫星钟差.     

A thermodynamic model of high temperature lava vaporization on Io

We modified the MAGMA chemical equilibrium code developed by Fegley and Cameron (1987, Earth Planet. Sci. Lett. 82, 207-222) and used it to model vaporization of high temperature silicate lavas on Io. The MAGMA code computes chemical equilibria in a melt, between melt and its equilibrium vapor, and in the gas phase. The good agreement of MAGMA code results with experimental data and with other computer codes is demonstrated. The temperature-dependent pressure and composition of vapor in equilibrium with lava is calculated from 1700 to 2400 K for 109 different silicate lavas in the ONaKFeSiMgCaAlTi system. Results for five lavas (tholeiitic basalt, alkali basalt, Barberton komatiite, dunite, and a molten type B1 Ca, Al-rich inclusion) are discussed in detail. The effects of continuous fractional vaporization on chemistry of these lavas and their equilibrium vapor are presented. The predicted abundances (relative to Na) of K, Fe, Si, Al, Ca, and Ti in the vapor equilibrated with lavas at 1900 K are lower than published upper limits for Io's atmosphere (which do not include Mg). We predict evaporative loss of alkalis, Fe, and Si during volcanic eruptions. Sodium is more volatile than K, and the Na/K ratio in the gas is decreased by fractional vaporization. This process can match Io's atmospheric Na/K ratio of 10±3 reported by Brown (2001, Icarus 151, 190-195). Silicon monoxide is an abundant species in the vapor above lavas. Spectroscopic searches are recommended for SiO at IR and mm wavelengths. Reactions of metallic vapors with S- and Cl-bearing volcanic gases may form other unusual gases including MgCl₂, MgS, MgCl, FeCl₂, FeS, FeCl, and SiS.     

A Robotics Perspective On Human Spaceflight

There has traditionally been a dichotomy in the space community regarding the efficacy of human versus robotic exploration of space. I argue that no such dichotomy is necessary, and that there is a natural and synergistic division of labour between man and machine, and that this division of labour will evolve in symbiotic fashion. The present state-of-the-art robotics technology is insufficient to replace the human in space, but is sufficient to act as a useful, even necessary, tool in aiding the astronaut in the conduct of useful work. I further argue that as robotics technology advances, the human will be further relieved to perform tasks best suited to human decision-making and flexibility that is unlikely in the near-term to be matched by autonomous or teleoperated machines. This revised version was published online in July 2006 with corrections to the Cover Date.     

CO J=2-1 and J=3-2 Line Observations of Molecular Clouds toward the Directions of 59 EGOs in the Northern Sky

In order to investigate the differences between the molecular clouds which are associated with the massive star forming regions and those which are not, we have performed the single-dish simultaneous observations of ¹²CO J=2-1 and J=3-2 lines toward a sample of 59 Spitzer Extended Green Objects (EGOs) as the massive star forming regions in the northern sky. Combining our results with the data of the ¹²CO J=1-0 observations toward the same sample EGOs in the literature, we have made the statistical comparisons on the intensities and linewidths of multiple ¹²CO lines between the molecular clouds associated with EGOs (EGO molecular clouds, in brief) and other non-EGO molecular clouds. On this basis, we have discussed the effects of the gas temperature, density, and velocity field distributions on the statistical characteristics of the two kinds of molecular clouds. It is found that both the EGO molecular clouds and non-EGO molecular clouds have similar mass ranges, hence we conclude that for the formation of massive stars, the key-important factor is probably not the total mass of a giant molecular cloud (GMC), but the volume filling factor of the molecular clumps in the GMC (or the compression extent of the molecular gas in the cloud).     

Orbital change experiments with a Mars general circulation model

We use a Mars general circulation model to examine the effect of orbital changes on the planet’s general circulation and climate system. Experiments are performed for obliquities ranging from 0° to 60° for two different longitudes of perihelion. Each experiment simulates a full Mars year assuming a fixed atmospheric dust distribution and fixed amount of CO₂ in the atmosphere/cap system. We find that global mean surface temperatures and pressures decline with increasing obliquity due to the increasing extent of the winter polar caps. The seasonal CO₂ cycle and intensity of the solstice circulation amplify considerably with increasing obliquity such that global dust storms are likely at both solstices. The most significant feature of the high obliquity solstice circulations is the development of an intense low-level jet associated with the return branch of the Hadley circulation.Model surface stresses are used to map regions of preferred dust lifting, which are defined in terms of an annual deflation potential. For the present obliquity, the model-predicted regions of high deflation potential are in good agreement with Cantor et al.’s (2001, J. Geophys. Res.106, 23653-23688) observations, which gives us some confidence in the model’s ability to predict where lifting might occur when Mars’ orbit parameters are different than they are today. In general we find that the dust lifting potential increases sharply with obliquity and is greatest at times of high obliquity when perihelion coincides with northern summer solstice. Over an obliquity cycle, the model global annual deflation potential ranges from several tenths of a millimeter at 0° obliquity to almost 15 mm at 60° obliquity. Much higher values are possible when the atmosphere is very dusty.We find a strong correlation between the deflation potential and surface thermal inertia: regions of high deflation potential correspond to regions of high thermal inertia (high rock abundance), and regions of low deflation potential correspond to regions of low thermal inertia (high dust/sand abundance). Furthermore, while the regions of preferred lifting (high deflation potential) expand somewhat with increasing obliquity and dust loading, the central parts of Tharsis, Arabia, and Elysium show no tendency for significant lifting at any obliquity or longitude of perihelion. These regions may therefore be very old and represent net long-term sinks for atmospheric dust. It is the topography of the planet, through its influence on surface pressure and wind systems, which ultimately determines where dust accumulates.Finally, as was found by Fenton and Richardson (2001, J. Geophys. Res.106, 32885-32909), we find no tendency for the development of east-southeasterly winds at the Pathfinder site for any of our orbital change experiments. This suggests that the ancient wind regime discussed by Greeley et al. (2000, J. Geophys. Res.105, 1829-1840) was produced by other factors, such as polar wander.     

Third Order Effect of Rotation on Stellar Oscillations of a B Star

We aim at investigating the effect of rotation up to the third order in the angular velocity of a star on the p and g modes, based on the formalism developed by Soufi et al. Our ultimate goal is the study of oscillations of β Cephei stars which are often rapidly rotating stars. Our results show that the third-order perturbation formalism presented by Soufi et al. should be corrected for some missing terms and some misprints in the equations. As a first step in our study of β Cephei stars, we quantify by numerical calculations the effect of rotation on the oscillation frequencies of a uniformly rotating zero-age main-sequence star with 12 M<,??>. For an equatorial velocity of 100km s-1, it is found that the second-and third-order corrections for (l, m)=(2, 2), for instance, are of the order of 0.01% of the frequency for radial order n=6 and reaches up to 0.5% for n=14.     

Construction of the Database for Pulsating Variable Stars

A database for pulsating variable stars is constructed to favor the study of variable stars in China. The database includes about 230,000 variable stars in the Galactic bulge, LMC and SMC observed in an about 10 yr period by the MACHO(MAssive Compact Halo Objects) and OGLE(Optical Gravitational Lensing Experiment) projects. The software used for the construction is LAMP, i.e., Linux+Apache+MySQL+PHP. A web page is provided for searching the photometric data and light curves in the database through the right ascension and declination of an object. Because of the flexibility of this database, more up-to-date data of variable stars can be incorporated into the database conveniently.     

Born–Infeld type phantom model in the ω–ω′ plane

In this paper, we investigate the dynamics of Born–Infeld (B–I) phantom model in the ω–ω′ plane, which is defined by the equation of state parameter for the dark energy and its derivative with respect to N (the logarithm of the scale factor a). We find the scalar field equation of motion in ω–ω′ plane, and show mathematically the property of attractor solutions which correspond to ω _φ ∼−1, Ω _φ =1, which avoid the “Big rip” problem and meets the current observations well.      

Shear and vorticity in a combined Einstein-Cartan-Brans-Dicke inflationary lambda-Universe

A combined BCDE (Brans-Dicke and Einstein-Cartan) theory with lambda-term is developed through Raychaudhuri’s equation, for inflationary scenario. It involves a variable cosmological constant, which decreases with time, jointly with energy density, cosmic pressure, shear, vorticity, and Hubble’s parameter, while the scale factor, total spin and scalar field increase exponentially. The post-inflationary fluid resembles a perfect one, though total spin grows, but the angular speed does not (Astrophys. Space Sci. 312: 275, 2007d).