2023 Fiscal Year Research-status Report
| Project/Area Number |
22K20601
|
|---|
| Research Institution | Kyoto University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2022-08-31 – 2025-03-31
|
| Keywords | inverse problems / transdimensional / interface detection / RJMCMC / HMC / layer detection |
|---|
| Outline of Annual Research Achievements |
The research focuses on visualizing subsurface accurately via stochastic inversion, identifying parameters like elastic modulus and interface geometry. The 1-D transdimensional MCMC algorithm, developed initially, now handles correlated and uncorrelated data. Tested on synthetic and real geotechnical data, it precisely detects subsurface layers, depths, and spatial properties, outperforming competitors computationally. Its standout feature is capturing uncertainty in subsurface stratification. A manuscript detailing these findings is under journal review.
Experimental results were mixed despite using a new apparatus. While hydraulic head measurements were precise, outflow rate measurements still show significant errors, albeit less than phase 1. We're reassessing our experimental approach.
|
| Current Status of Research Progress |
Current Status of Research Progress
3: Progress in research has been slightly delayed.
Reason
The 1-D transdimensional algorithm developed in phase 1 of this project did not work well on real data (especially for correlated data). This algorithm was improved in the last year and is now able to quantify the uncertainty in number of layers, depth of layer interfaces and random field properties of each layer. This novel algorithm developed has given a valuable insight that accounting for correlations in observation data is of vital importance. This knowledge is now being used in the development of the 2-D transdimensional algorithm.
The new experimental apparatus helped solve some issues in the experimentation program. However, some complications still exist in measuring the outflow rate and we are working on methods to improve the fidelity of these measurements.
|
| Strategy for Future Research Activity |
After the successful development and validation of the 1-D transdimensional algorithm on different types of cone penetration test data, work is ongoing to develop the transdimensional HMCID algorithm for void detection.
During the development of the 1-D algorithm, the PI has learnt of a method to integrate cone penetration as well as geophysical data (such as electromagnetic data) to develop higher quality 2-D images of the subsurface by repeated use of 1-D transdimensional algorithms. To this end, I am currently collaborating with geophysicists at Geoscience Australia to integrate the 1-D transdimensional algorithm for cone penetration data (already developed in this project) with their 1-D transdimensional algorithm (open source program known as HiQGA) for electromagnetic data.
A few more trials will be made to obtain reliable measurement results. Learnings from past trials have helped us make improvements in obtaining reliable data. We will continue to modify the device till suitable observations are obtained.
|
| Causes of Carryover |
To achieve the goals of the research project in a more proper manner.
The funds will be used for conducting additional experiments, participating in academic conferences, submitting papers, etc.
|
