SFB 1313 Milestone Presentation by Dongwon Lee

March 25, 2021 /

Doctoral researcher at the Institute of Applied Mechanics (CE)| University of Stuttgart
Thursday, 25 March 2021 | 5 pm CET

Dongwon Lee, SFB 1313 doctoral researcher at the Institute of Applied Mechanics (CE) (research project B05), will give his milestone presentation "Hydromechanics of fracture and fracture networks: Experimental characterization" on 25 March 2021 at 5 pm CET.

Date: Thursday, 25 March 2021
Time: 5 pm CET
Title: "Hydromechanics of fracture and fracture networks: Experimental characterization"
Place: online presentation >>> If you are interested in participating in the lecture, please contact simon.emmert@iws.uni-stuttgart.de

Abstract

Micro X-Ray Computed Tomography (μ-XRCT) has developed significantly in recent years, both in terms of hardware but also in terms of efficient image analysis. Image segmentation became one of the most prominent tools in a laboratory environment to visualize the 3D pore structure of porous media in a non-destructive way.  However, as with various other visualization techniques, the analyzed data obtained from segmentation suffer from some drawbacks, such as the inherent noise and imaging artefacts. In order to overcome such issues, the effort to improve the potential of XRCT data post-processing is ongoing. These efforts include the improvement of filtering techniques to degrade the signal-to-noise ratio, and the corresponding segmentation techniques to extract the features of interest.

In our study related to filtering techniques, we compare three existing non-linear filtering methods, which are: non-local means (NLM), anisotropic diffusion (AD) and adaptive manifold non-local means (AMNLM) methods. We show for representative fractured rock and salt precipitation samples, that AMNLM filter outperforms the others.  Additionally, segmentation of fracture data often reaches the limitations of a XRCT device related to resolution due to the large aspect ratios, i.e. length-to-width ratio of single fractures. Therefore, we adopt five different workflows and compare the results with each other. The comparison shows that one of the adopted workflows which includes the U-net model, which is one of Convolutional Neural Network model, is superior compared to the other workflows in terms of segmentation quality and computational efficiency.

Furthermore, our on-going hydro-mechanical experimental investigation of single fracture are discussed. This study was conducted by an artificially created fracture in an aluminum core. The fluid-saturated fracture with rough fracture surfaces was stimulated with harmonic flow-rates. We were able to observe a non-linear pressure response due to fracture deformations including volume changes. Compared to natural fractures, artificially created fractures differ in some characteristic properties. This includes differences in global storage capacity and local storativity, effective permeability and (local) contact stiffness. We therefore hydraulically fractured a low-permeability/porosity Bentheim sandstone core by in order to acquire a single natural-like hydraulic fracture. On the basis of the developed fracturing workflow, we will perform further stimulation experiments in the future.

SFB 1313 Research Project B05 Hydromechanics of fractures and fracture networks: A combined numerical multi-scale and experimental investigation

To the top of the page