PhD candidate
Research group | Njord
Main supervisor | François Renard
Co-supervisor | Benoît Cordonnier, Jessica Ann McBeck
Affiliation | Department of Geosciences, UiO
Contact | syadhisy.dhanapal@geo.uio.no
Short bio
I completed my BSc in Geophysics with a minor in Physics at the University of Washington in Seattle, USA. Shortly after, I started working in the oil and gas industry, specifically in seismic and non-seismic data acquisition, and seismic data processing.
My interest in programming led me to pursue an MSc in Computer Science at the University of York, which piqued my curiosity in applying novel machine learning methods in geoscience data. As a Doctoral Research Fellow with CompSci & Njord Center at UiO, I will be investigating machine learning methods in rock physics and geo-energy experiments.
Research interests and hobbies
With the increase in global temperatures attributed to anthropogenic CO2 emission, CO2 storage in the subsurface has become a key topic. I am interested in exploring the geomechanical impact of CO2 storage in various porous media. I am also looking forward to using and testing physics-guided and/or transformer-based machine learning architectures in x-ray microtomography datasets.
My hobbies include making personalized greeting cards, cooking, exploring new cities & cultures, and spending quality time outdoors with my family and pet poodle.
CompSci project
Machine Learning in Rock Physics & Geo-Energy Experiments
4D x-ray microtomography provides in-situ imaging of triaxial rock physics experiments. These experiments give valuable insights on the integrity of the investigated porous media when subjected to pressure and temperature settings akin to subsurface conditions where macroscale rock failures occur. Acquisition of x-ray microtomograms contain noise that needs to be removed prior to interpretation. Example of noise include random noise and ring artefacts. Conventional denoise techniques such as edge enhancement, non-local means filters and others have been widely used to produce images with higher signal-to-noise ratio (SNR). Post denoise, image segmentation techniques such as greyscale thresholding and local segmentation are applied to differentiate grains, pores, and fractures from the rock matrix. First objective of the PhD is as follows:
1. To investigate and produce novel machine-learning (ML) techniques to perform image denoising, followed by image segmentation in an automated manner to increase image quality for analysis and interpretation stage.
The ML techniques will then be tested and optimized for rock triaxial deformation study within the context of carbon dioxide (CO2) storage in basalt. Therefore, the second objective of the PhD is as follows:
2. To investigate the impact of CO2 storage in basalt with respect to local geomechanics failure.
Local geomechanics failure will be studied with respect to injection-induced fractures along the borehole and mineralization-induced fractures surrounding the amygdales within the basalt core. 4D x-ray microtomograms of the injection and mineralization will be acquired using the HADES triaxial compression tool at the European Synchrotron Radiation Facility in Grenoble, France.
In summary, using the ML-processed x-ray microtomograms, the PhD research will investigate the propagation and coalescence of the fractures and its role in inducing the failure of the basalt core.
Publications
CompSci publications
None yet.
Previous publications
None
