Imaging and full waveform inversion using machine learning
This Master's project was designed for Alrik Frost who started his Master's program in Earth Sciences, UiB, in the fall semester 2023. The Master's project is given by the research group Geophysics.
Main content
Project description
Motivation (background):
Machine learning has been used successfully in both the natural and social sciences to process large amount of data and, for example, to determine and analyze complex patterns in these data. Since geophysics, and especially seismic exploration, also involves the processing of large amount of data it is natural to apply machine learning algorithms to seismics and, in particular, to the most challenging part of seismic processing: the generation of accurate velocity models of the subsurface.
Hypothesis (scientific problem):
Determining Earth’s subsurface properties is one of the most challenging tasks in geophysics, both theoretically and computationally. Typically, this is done using either imaging or full waveform inversion (FWI). The latter is more expensive but also more accurate. In practice imaging/FWI consists of optimizing the misfit between the observed and computed waveforms using gradient based methods. One particular type of machine learning method, stochastic gradient descent (SGD), consists of a variation of gradient based methods, and this method has already successfully been used in FWI. The goal of this thesis is to develop and test a new machine learning algorithm, using SGD as starting point, and apply this to imaging and/or FWI.
Test (work):
The default modeling method in imaging and FWI, both for acoustic and elastic media, has been finite difference modeling. An alternative modeling method, which is faster and specifically valid at higher frequencies, is based on the ray-Born approximation. FWI using ray-Born modeling method has successfully been done for a number of models and codes are available. It has not been used in conjunction with SGD however. Therefore, a first task in this project would be to implement SGD for ray-Born based FWI. This can be done in combination with other methods, such as shot encoding, to further speed up the inversion algorithm. A second step would then be to compare this against a more traditional inversion method (for example, one that uses l-BFGS). Finally, if there is time variations of SGD or other machine learning codes can be used in the inversion.
Proposed course plan during the master's degree (60 ECTS):
GEOV274, GEOV300, GEOV355, INF264, special pensum (5sp), INF265
