arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2311.05060astro-ph.EPastro-ph.IM
收藏

基于水平集方法的小天体全球重力反演

Small bodies global gravity inversion via the level-set method

Alfonso Caldiero, Sébastien Le Maistre

AI总结:

本研究提出一种基于水平集方法的小天体全球重力反演方法,通过分段常数密度假设、迭代正则化优化及多初始值聚类,从引力势数据推断内部结构并评估不确定性。

摘要:

我们提出了一种从小天体引力势测量值(例如由航天器无线电跟踪提供的测量值)推断其内部大尺度非均匀性的方法。本文通过将解限制为分段常数密度分布来缓解重力反演的非唯一性问题,这类分布由分散在背景介质中的多个均匀密度区域(质量异常体)构成。每个异常体的边界被隐式定义为标量场(称为水平集函数)的0等值面,因此通过修改该场即可改变异常体的形状和位置。与密度分布相关的引力势通过线积分多面体方法计算,得到其球谐展开的系数。随后通过带Tikhonov正则化的迭代最小二乘方法调整密度分布,在每次迭代中估计水平集函数的校正量、每个异常体的密度差以及背景密度,以最小化预测重力系数与实测重力系数之间的残差。考虑到该问题的非凸性以及假设缺乏先验知识(除了天体的形状之外),估计过程会针对多个随机初始分布重复执行,并根据与输入测量值无关的全局属性对得到的解进行聚类。这提供了与数据一致的候选内部模型族,每个模型族中局部密度值的分布范围被用于评估与估计分布相关的不确定性。

英文摘要:

We propose an approach to infer large-scale heterogeneities within a small celestial body from measurements of its gravitational potential, provided for instance by spacecraft radio-tracking. The non-uniqueness of the gravity inversion is here mitigated by limiting the solutions to piecewise-constant density distributions, thus composed of multiple regions of uniform density (mass anomalies) dispersed in a background medium. The boundary of each anomaly is defined implicitly as the 0-level surface of a scalar field (called the level-set function), so that by modifying this field the shape and location of the anomaly are varied. The gravitational potential associated with a density distribution is here computed via a line-integral polyhedron method, yielding the coefficients of its spherical harmonics expansion. The density distribution is then adjusted via an iterative least-squares approach with Tikhonov regularization, estimating at every iteration corrections to the level-set function, the density contrast of each anomaly, and the background density, in order to minimize the residuals between the predicted gravity coefficients and those measured. Given the non-convexity of the problem and the lack of prior knowledge assumed (save for the shape of the body), the estimation process is repeated for several random initial distributions, and the resulting solutions are clustered based on global properties independent of the input measurements. This provides families of candidate interior models in agreement with the data, and the spread of the local density values across each family is used to assess the uncertainties associated with the estimated distributions.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑