確率インバージョンによる地中内部の可視化

使用随机反演可视化地下内部

• 批准号: 22K20601
• 负责人: Koch Michael・Conrad
• 金额: $ 1.83万
• 依托单位: Kyoto University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Research Activity Start-up
• 财政年份: 2022
• 资助国家: 日本
• 起止时间: 2022-08-31 至 2024-03-31
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-22K20601/
• 关键词: inverse problems; transdimensional; interface detection; RJMCMC; HMC; layer detection; 逆解析; 機能診断

The main aim of the research plan is the accurate visualization of subsurface through stochastic inversion. This involves identification of spatial parameters such as elastic modulus and geometry parameters that represent an interface where jump in spatial parameters occur. In the first phase of this project, a novel 1-D transdimensional MCMC/HMC algorithm has been developed that enables inversion in the cases where the number of spatial and geometry parameters are unknown. This is the first step in the development of the transdimensional-HMC for Interface Detection (t-HMCID) algorithm. Instead of detecting multiple voids, the algorithm is being tested on synthetic cone penetration (geotechnical) data for the detection of exact number of subsurface layers, their depths and their spatial properties. The results obtained are encouraging and shows superior/equivalent performance to competing algorithms. This algorithm will be extended to 2-D and 3-D and the insights obtained from the development of this algorithm will be applied to the development of the t-HMCID algorithm subsequently.The experimentation program was initiated with mixed results. An existing experimental apparatus was employed for the experiment. While the measurements of hydraulic head were accurate, a large error was observed in outflow rate measurements. A plan is being developed currently on possible modifications of the experimental device to yield robust results.

研究计划的主要目的是通过随机反演实现地下的精确可视化。这涉及空间参数的识别，如弹性模量和几何参数，这些参数表示空间参数发生跳跃的界面。在该项目的第一阶段，开发了一种新的一维跨维MCMC/HMC算法，可以在空间和几何参数数量未知的情况下进行反演。这是跨维hmc接口检测（t-HMCID）算法发展的第一步。该算法不是检测多个空洞，而是在合成锥突（岩土）数据上进行测试，以检测地下层的确切数量、深度和空间特性。得到的结果令人鼓舞，并显示出优于竞争算法的性能。该算法将扩展到二维和三维，从该算法的发展中获得的见解将应用于随后的t-HMCID算法的发展。实验项目开始时的结果好坏参半。实验用的是现有的实验装置。水头测量精度较高，但流量测量误差较大。目前正在制定一项计划，对实验装置进行可能的修改，以产生可靠的结果。

项目成果

Assessment of HMC Parameter Updates for Piping Zone Boundary Detection

用于管道区域边界检测的 HMC 参数更新评估

• 发表时间: 2022
• 期刊: Proceedings of the 8th International Symposium on Geotechnical Safety and Risk (ISGSR)
• 作者: Michael C. Koch;Kazunori Fujisawa and Akira Murakami
• 通讯作者: Kazunori Fujisawa and Akira Murakami

Hydraulic Conductivity Field Marginalization in HMC Based Estimation of Piping Zone Boundary

基于 HMC 的管道区域边界估计中的水力传导率场边缘化

• 发表时间: 2022
• 期刊: Proceedings of the 8th International Symposium on Geotechnical Safety and Risk (ISGSR)
• 作者: Kazunori Fujisawa;Michael C. Koch;Misato Osugi and Akira Murakami
• 通讯作者: Misato Osugi and Akira Murakami

Reversible Jump MCMC を用いた地層分布の推定

使用可逆跳跃 MCMC 估计地质分布

• 发表时间: 2022
• 作者: 松本 祐弥;Koch Michael;藤澤 和謙
• 通讯作者: 藤澤 和謙

Hamiltonian Monte Carloを用いた水みち形状と透水係数分布の同時推定

使用哈密顿蒙特卡罗同时估计水路形状和导水率分布

• 发表时间: 2022
• 作者: 柴田 達哉;Koch Michael;藤澤 和謙
• 通讯作者: 藤澤 和謙