Deep Learning Enhanced Subsurface Imaging
Deep Neural Networks (DNN) can be used for the discovery of oil and gas, the gathering of geothermal information and underground CO2 storage. Subsurface imaging and characterization are typically nonlinear, ill-posed, inverse problems facing many challenges, such as scarcity of data, high uncertainties, and nonuniqueness of solution.
In recent years, there has been a surge of research efforts leveraging Machine Learning (ML) and Artificial Intelligence to overcome these challenges. He first explained Physics-Driven Deep Neural Networks and how that works for subsurface characterization while drilling.
The second example he gave featured an enhanced full waveform inversion scheme, which used machine learning low frequency seismic signals. His third example was about deep learning assisted joint inversion of multi-geophysics data for the monitoring of underground CO2 storage sites.
Jiefu Chen received his Ph.D. in electrical engineering from Duke University in 2010. His research interests include inverse problems and ML for scientific computing.