基于机器学习的单脉冲搜索候选体识别对FAST观测CRAFTS数据的应用研究

单脉冲搜索作为脉冲星探测的有力工具,在探测旋转射电暂现源以及快速射电暴中扮演着重要角色。为了从海量的射电巡天数据中快速筛选出最有价值的单脉冲搜索候选体,候选体识别已经从早期启发式阈值判断发展到基于机器学习自动识别。对于FAST观测,研究了基于机器学习的单脉冲搜索候选体识别应用到CRAFTS(the commensal radio astronomy FAST survey)超宽带脉冲星数据的性能表现。在评估过程中,使用单脉冲事件组识别(SPEGID)和单脉冲搜索器(SPS)两类自动识别方法,通过7种不同机器学习分类器对CRAFTS基准数据集产生的单脉冲搜索候选体进行自动识别;作为对比,也使用了启发式阈值判断的方法(RRATtrap和Clusterrank)。结果表明,SPEGID具有最好的性能表现(最高的F1-score值95.1%、次高的召回率95.4%、最低的假阳性率4.7%),SPS具有最快的筛选速度(平均每小时筛选4010个候选体)。通过对比分析结果,探讨了如何基于FAST观测数据开展高效的单脉冲搜索候选体识别。

Pulsar Candidate Classification Using a Computer Vision Method from a Combination of Convolution and Attention

Artificial intelligence methods are indispensable to identifying pulsars from large amounts of candidates.We develop a new pulsar identification system that utilizes the CoAtNet to score two-dimensional features of candidates,implements a multilayer perceptron to score one-dimensional features,and relies on logistic regression to judge the corresponding scores.In the data preprocessing stage,we perform two feature fusions separately,one for one-dimensional features and the other for two-dimensional features,which are used as inputs for the multilayer perceptron and the CoAtNet respectively.The newly developed system achieves 98.77%recall,1.07%false positive rate(FPR)and 98.85%accuracy in our GPPS test set.

Cleaning Radio Frequency Interference in Pulsar-Folded Data Based on the Conditional Random Fields with an Adaptive Prior

Radio astronomy observations are frequently impacted by radio frequency interference(RFI).We propose a novel method,named 2σCRF,for cleaning RFI in the folded data of pulsar observations,utilizing a Bayesian-based model called conditional random fields(CRFs).This algorithm minimizes the“energy”of every pixel given an initial label.The standard deviations(i.e.,rms values)of the folded pulsar data are utilized as pixels for all subintegrations and channels.Non-RFI data without obvious interference is treated as“background noise,”while RFI-affected data have different classes due to their exceptional rms values.This initial labeling can be automated and is adaptive to the actual data.The CRF algorithm optimizes the label category for each pixel of the image with the prior initial labels.We demonstrate the efficacy of the proposed method on pulsar folded data obtained from Five-hundred-meter Aperture Spherical radio Telescope observations.It can effectively recognize and tag various categories of RFIs,including broadband or narrowband,constant or instantaneous,and even weak RFIs that are unrecognizable in some pixels but picked out based on their neighborhoods.The results are comparable to those obtained via manual labeling but without the need for human intervention,saving time and effort.

Constraints on Dark Energy from the CSST Galaxy Clusters

We study the potential of the galaxy cluster sample expected from the Chinese Space Station Telescope(CSST)survey to constrain dark energy properties.By modeling the distribution of observed cluster mass for a given true mass to be log-normal and adopting a selection threshold in the observed mass M_(200m)≥0.836×10^(14)h^(-1)M_(⊙),we find about 4.1×10^(5)clusters in the redshift range 0≤z≤1.5 can be detected by the CSST.We construct the Fisher matrix for the cluster number counts from CSST,and forecast constraints on dark energy parameters for models with constant(w_(0)CDM)and time dependent(w_(0)w_(a)CDM)equation of state.In the self-calibration scheme,the dark energy equation of state parameter w_(0)of the w_(0)CDM model can be constrained toΔw_(0)=0.036.If w_(a)is added as a free parameter,we obtainΔw_(0)=0.077 andΔw_(a)=0.39 for the w_(0)w_(a)CDM model,with a Figure of Merit for(w_(0),w_(a))of 68.99.Should we have perfect knowledge of the observable-mass scaling relation("known SR"scheme),we would obtainΔw_(0)=0.012 for the w_(0)CDM model,andΔw_(0)=0.062 andΔw_(a)=0.24 for the w_(0)w_(a)CDM model.The dark energy Figure of Merit of(w_(0),w_(a))increases to 343.25.This indicates again the importance of calibrating the observable-mass scaling relation for optically selected galaxy clusters.By extending the_(max)imum redshift of the clusters from z_(max)~1.5 to Z_(max)~2,the dark energy Figure of Merit for(w_(0),w_(a))increases to 89.72(self-calibration scheme)and 610.97("known SR"scheme),improved by a factor of~1.30 and~1.78,respectively.We find that the impact of clusters’redshift uncertainty on the dark energy constraints is negligible as long as the redshift error of clusters is smaller than 0.01,achievable by CSST.We also find that the bias in logarithm mass must be calibrated to be 0.30 or better to avoid significant dark energy parameter bias.

An RFI Mitigation Pipeline for CRAFTS Multi-beam Data Based on Signal Cross-Correlation Function and SumThreshold Algorithm

The increasing radio frequency interference(RFI)is a well-recognized problem in radio astronomy research.Pulsars and Fast Radio Bursts(FRBs)are high-priority science targets of the ongoing Commercial Radio Astronomy FAST Survey(CRAFTS).To improve the quality of RFI removal in searches of pulsars and FRBs based on CRAFTS multi-beam data,we here propose an intuitive but powerful RFI mitigation pipeline(CCF-ST).The“CCF-ST”is a spatial filter constructed by signal cross-correlation function(CCF)and Sum-Threshold(ST)algorithm.The RFI marking result is saved in a“mask”file,a binary format for RFI masks in PRESTO.Three known pulsars,PSR B0525-21,PSR B0621-04,and PSR J0943+2252 from CRAFTS L-band 19 beams data are used for evaluation of the performance of CCF-ST in comparison with other methods,such as PRESTO’s“rfifind”,ArPLS-ST and ArPLS-SF.The result shows that CCF-ST can reduce effective data loss rate and improves the detected signal-to-noise ratio of the pulsations by~26%and~18%respectively compared with PRESTO’s“rfifind”and ArPLS-ST.The CCF-ST also has the advantage of low computational cost,e.g.,reducing the time consumption by~40%and memory consumption by~90%compared with ArPLS-SF.We expect that the new RFI mitigation and analysis toolkit(CCF-ST)demonstrated in this paper can be applied to CRAFTS and other multi-beam telescope observations to improve the data quality and efficiency of pulsar and FRB searches.

Fractions of Compact Object Binaries in Star Clusters:Theoretical Predictions

The binary population in field stars and star clusters contributes to the formation of gravitational wave(GW)sources.However,the fraction of compact-object binaries(CBs),which is an important feature parameter of binary populations,is still difficult to measure and very uncertain.This paper predicts the fractions of important CBs and semi-compact object binaries(SCBs) making use of an advanced stellar population synthesis technique.A comparison with the result of N-body simulation is also presented.It is found that most CBs are formed within about 500 Myr after the starburst.The fractions of CBs and SCBs are demonstrated to correlate with stellar metallicity.The higher the metallicity becomes,the smaller the fraction of black hole binaries(BHBs),neutron star binaries(NSBs) and SCBs.This suggests that the GW sources of BHBs and NSBs are more likely to form in metal-poor environments.However,the fraction of black hole-neutron star binaries is shown to be larger for metalrich populations on average.

月球着陆器过渡轨道的数值仿真

针对月球背面着陆器过渡轨道的选择问题,选择南极艾德肯盆地SPA作为登陆位置,设计了月球正面远月点100km高程,背面SPA为近月点10km高程的椭圆型过渡轨道,利用动力学方法和月球高阶重力场模型GRGM660PRIM,对该轨道进行了仿真分析.结果表明,着陆器的在轨时间接近7d,其中近月点在SPA中心平坦区域的时间超过3d.较长的在轨时间有利于后续着陆决策操作的准备,说明该设计的初始轨道根数具有一定的实际意义,可为着陆器的成功登陆提供一定程度的参考.

基于VPython的运动仿真在大学物理教学中的应用

针对物理可视化教学中,如何快速有效地对复杂物理现象和规律进行三维仿真,提出利用开源免费的三维图形库VPython进行快速建模与仿真。以小球半圆弧轨道运动为例,对其进行建模、分析与模拟,验证了VPython仿真分析的高效性。将其应用于物理教学,有助于学生理解和掌握物理规律,在一定程度上有助于VPython的应用与推广,为教师及学生VPython的应用提供有益的参考。

利用基于互信息的不平衡分类方法识别稀有光谱（英文）

从海量恒星光谱中发现稀有光谱是天文学研究的重要课题之一。与一般光谱相比,稀有光谱数量较少,因此,传统分类方法无法正常工作。究其原因是这些方法不仅在分类决策时并未对稀有光谱予以更多关注,而且只关注分类的准确率。鉴于此,在总结当前分类方法的基础上,深入分析互信息与决策树之间的关系,提出基于互信息的代价缺失决策树。SDSS DR8中K型、F型、G型以及M型恒星光谱上的比较实验表明,与传统分类方法相比,所提方法能够较好地完成稀有光谱识别的任务。

伽马射线暴X射线余辉中星风—星际介质转变的行为

伽马射线暴(GRB)暴周环境是研究伽马暴物理的重要手段。其余辉的斜率“由陡变平”的行为可以用伽马暴暴周环境由星风向星际介质转变来解释,但至今还缺乏系统的样本搜寻。基于外激波模型,对2004年~2018年Swift卫星观测的X射线余辉进行系统的搜寻,寻找具有星风—星际介质转变特征的GRB,发现了6个GRB具有星风—星际介质转变的特征。对其进行模型拟合后,得到了一组最佳拟合结果相应的物理参数。结果表明:①星风—星际介质转变时间一般在暴后800~2000 s,转变半径R t在1016~1017 cm,密度跳跃参数在4~10,电子谱指数p在2.1~2.6;②它们的各向同性动能E k,iso在1053~1054 erg,与一般长暴的典型值相比要小一些;③具有星风—星际介质转变的GRBs在喷流张角比较大,大部份大于10°。

影响长江口深水航道骤淤的非常态天气过程Ⅱ:台风要素敏感性分析及典型台风路径

为进一步研究台风中心气压、台风距离长江口距离(简称距离)、最大风速半径和移动风速4个典型参数对长江口航道骤淤的影响水平,采用正交试验方法,设计了64组典型试验工况,选择藤田-高桥圆形风场经验公式架构了台风场,然后采用SWAN模型模拟不同台风工况的波况,并计算牛皮礁站的波能。据波能的极差分析和方差分析,得出台风要素的敏感性次序。研究表明,长江口过境台风中,距离最敏感,而最大风速半径为较不敏感因素。选取主要的敏感性参数(距离、中心气压)为代表参数,概化移动风速和最大风速半径。在合理选取能够引起骤淤的临界波能的标准下,给出较易引发骤淤的典型包络线范围,为骤淤的预报提供了新的参考指标。

影响长江口深水航道骤淤的非常态天气过程Ⅰ:台风的路径特征及数值验证

为归纳长江口深水航道台风期骤淤的发生规律及特征,分析了发生骤淤时刻的气象条件与对应的波浪条件。研究发现,牛皮礁站的波能与骤淤具有较好的相关性;从台风路径上分析,长江口东侧过境台风对航道的骤淤影响显著。结合历史台风路径,选取3个典型路径的台风,选择藤田-高桥圆形经验风场和CFSR(climate forecast system reanalysis)风场的混合风场复演了台风场,然后采用SWAN模型模拟了不同路径台风期间的波况,最后以牛皮礁站的浅水波能流为判别参数,分析不同路径台风对长江口深水航道骤淤的影响。研究表明长江口东侧过境的台风是较易产生较大波能并进一步诱发骤淤的典型台风路径,这一分析结果与2010年以来的骤淤实测台风路径结果吻合。

恒星低质量光谱的连续谱拟合方法

恒星的连续谱是由于黑体辐射导致的光辐射强度随波长(频率)连续光滑变化的光谱。每条观测到的光谱数据中都会包含连续谱、谱线和噪声。恒星的分类主要是依据光谱的谱线、连续谱的相对强度以及光谱的其他特征。恒星连续谱的分布以及谱线的轮廓是由恒星大气内的物理因素决定的,也可以根据连续谱及谱线对恒星大气的物理参数进行估计。因而处理光谱的主要问题就是提取连续谱,并且通过归一化进行谱线的提取。恒星连续谱提取的算法主要有多项式逼近、中值滤波、形态滤波以及小波滤波等,但是这些方法对于低质量光谱处理的鲁棒性不是很好,因此有必要研究一种新的算法对低质量光谱的连续谱进行提取。在仔细分析恒星低质量连续谱的基础上,提出一种基于蒙特卡罗方法的低质量恒星连续谱拟合方法。该方法对恒星光谱筛选过程中不在范围内的点利用蒙特卡罗均匀分布进行自动插值,让每一个波长都对应一个流量点,然后对这些流量点进行低阶多项式迭代拟合,从而得到连续谱。为了验证算法对不同信噪比的低质量光谱连续谱提取的鲁棒性,利用不同的信噪比在原始光谱中加入不同的高斯白噪声对低质量光谱进行模拟。结果表明蒙特卡罗算法对不同信噪比的低质量光谱的拟合具有较高的精度与较强的鲁棒性。

月幔温度演化的二维有限元模拟

为研究月球自吸积结束后温度的变化,利用有限元方法构建极坐标系二维温度演化模型。对控制方程进行无量纲化处理,并对动量方程采用Boussinesq近似。结果表明,早期高温月幔产生了多个大尺度的对流环,随后由于内部热量的急剧释放,温度下降,压强减小,早期形成的对流环变得不稳定,月幔的平均温度由1 735K最终下降至1 278K左右。数值模拟结果精度较好,表明本文模型及程序代码具有可行性,结合其他地球物理数据,如月表热流值,将模型扩展至三维,可用于研究月球吸积完成后的热演化。

大张角波纹喇叭的优化设计

针对平方公里阵(Square Kilometre Array,SKA)天线对高灵敏度的需求,利用基于旋转体时域有限差分法(BOR-FDTD)和自适应协方差矩阵进化策略(CMA-ES)的波纹喇叭优化设计技术,提出了以灵敏度为目标的大张角波纹喇叭优化设计方法。分别以天线口径效率和灵敏度为优化目标对工作于4.6~8.51 GHz的大张角波纹喇叭进行优化设计。计算结果表明,以灵敏度为优化目标所设计的波纹喇叭综合性能更优,其交叉极化和反射损耗均优于-20 d B,用于SKA天线的口径效率在85.1%以上,灵敏度优于7.68 m^2/K。

基于CMA-ES优化方法的波纹喇叭优化技术

介绍了旋转体时域有限差分法(BOR-FDTD),导出了电磁场迭代计算公式。给出了自适应协方差矩阵进化策略(CMA-ES)的基本原理和步骤。提出了一种基于BOR-FDTD和CMA-ES的波纹喇叭优化设计技术,并将该项技术用于平方公里阵(Square Kilometre Array,SKA)望远镜天线Band 4(2.8~5.18 GHz)波纹喇叭馈源的优化设计。计算结果表明,该馈源在工作频带内反射损耗基本在-20 d B以下,天线口径效率均优于86.5%,且口径效率随频率的变化较小。

低信噪比巡天数据中特殊恒星光谱的搜寻方法

特殊恒星是金属丰度异常的恒星,其中包含的信息对于研究宇宙起源、太阳系的演变以及生命的演化都有着重要的意义。因此,特殊恒星的搜寻是国内外巡天项目中的重要目标。恒星光谱中包含着恒星的化学成分、物理性质以及运动状态等丰富的信息,它是开展恒星研究的重要依据。恒星的识别、分类以及特殊恒星的发现主要依据的是恒星光谱数据。随着LAMOST和SDSS等国内外大规模数字巡天项目的深入展开,恒星光谱的数据量达到了前所未有的高度,如此大的数据量为特殊恒星的发现提供了强有力的支撑。因此如何利用这些数据快速准确地发现特殊、稀少甚至于未知类型的恒星光谱是天文学研究的重要问题。数据挖掘是结合模式识别、机器学习、统计分析及相关专家背景知识,从数据中提取出隐含的过去未知的有价值的潜在信息的技术,其在处理大数据方面有着天然的优势,越来越多的数据挖掘方法被应用到巡天数据处理及分析之中。目前针对特殊恒星搜寻的数据挖掘算法主要包含随机森林、聚类分析以及异常值检测等,但随着巡天深度的拓展,观测的目标越来越暗,进而观测光谱的信噪比也随之变低。低信噪比光谱中存在着大量的无用信息,直接利用相关算法对其进行分析处理得到的结果往往存在很大的偏差。因此,如何从大量低信噪比恒星光谱巡天数据中有效地搜寻出特殊的恒星光谱,是当前面临的一个重要问题。由于低信噪比恒星光谱本身的特点,对于从中搜寻特殊恒星光谱的工作开展较少。为了解决此问题,在仔细研究光谱数据处理方法的基础上,针对低信噪比巡天数据中特殊恒星光谱的搜寻,提出了一种以主成分分析(PCA)和基于密度峰值聚类为基础的方法。该方法首先选取O,B,A,F,G,K和M各种类型的高信噪比恒星光谱,进行波长统一和流量插值后,利用主成分分析得到特征光谱;然后利用方差贡献率最大的前几个特征光谱对低信噪比的恒星光谱进行重构得到高信噪比的光谱;最后利用重构之后的高信噪比光谱进行聚类,聚类分析中得到的离群数据即为所要搜寻的特殊恒星光谱。在聚类时,考虑到恒星光谱数据本身的特点,采用了一种基于密度峰值的聚类方法来进行聚类及离群点的挖掘。实验表明,该方法能够在低信噪比的恒星光谱巡天数据中准确地搜寻出数量相对较少的特殊恒星。同时,也可应用于诸如LAMOST、SDSS等各种银河系巡天的光谱数据分析与挖掘中。

Numerical Investigation of Run-ups on Cylinder in Steep Regular Wave

The run-up on offshore structures induced by the steep regular wave is a highly nonlinear flow with a free surface. This article focuses on the investigation of the steep regular wave run-up on a single vertical cylinder by solving the Navier-Stokes equations. A numerical wave tank is established based on the open-source package to simulate the wave scattering induced by a vertical cylinder. The VOF method is applied to capture the large deformation and breaking of the free surface. The numerical model is validated by experimental results. The relative wave run-ups on the front face and the back face along the centerline of a cylinder are analyzed. The changes of the relative run-ups with the wave steepness, the relative diameter and the relative depth are studied. It is found that the relative run-ups on the front face and the back face of the cylinder depend mainly on the wave steepness and the relative diameter, while the dependence on the relative depth is weak. The empirical formulae are proposed to calculate the relative run-ups in terms of the wave steepness of incident regular waves and the relative diameter of a cylinder.

基于模式控制的四脊喇叭馈源设计

四脊喇叭馈源具有较宽的工作频带,但随着频率变化波束宽度难于保持恒定,导致照射效率降低。为了提高天线的口径效率,提出了一种通过控制馈源口面模式设计四脊喇叭馈源的方法。首先以天线口径效率为目标,通过优化得出喇叭口面上的最优模式比例,然后通过改变四脊喇叭的结构参数来调控其口面的模比,使其逐渐逼近最优模比。利用该设计方法,设计了一个半照射角为40°的四脊喇叭馈源,在3∶1带宽内,方向图波束宽度恒定且具有旋转对称性,天线口径效率优于60%。

A GPU based single-pulse search pipeline(GSP)with database and its application to the Commensal Radio Astronomy FAST Survey(CRAFTS)

We developed a GPU based single-pulse search pipeline(GSP)with a candidate-archiving database.Largely based upon the infrastructure of the open source PulsaR Exploration and Search Toolkit(PRESTO),GSP implements GPU acceleration of the de-dispersion and integrates a candidate-archiving database.We applied GSP to the data streams from the Commensal Radio Astronomy FAST Survey(CRAFTS),which resulted in quasi-real-time processing.The integrated candidate database facilitates synergistic usage of multiple machine-learning tools and thus improves efficient identification of radio pulsars such as rotating radio transients(RRATs)and fast radio bursts(FRBs).We first tested GSP on pilot CRAFTS observations with the FAST Ultra-Wide Band(UWB)receiver.GSP detected all pulsars known from the the Parkes multibeam pulsar survey in the corresponding sky area covered by the FAST-UWB.GSP also discovered 13 new pulsars.We measured the computational efficiency of GSP to be~120 times faster than the original PRESTO and~60 times faster than an MPI-parallelized version of PRESTO.