Session 1B — Lectures by Fellows/Associates

Umesh Waghmare, JNCASR, Bengaluru

Machine learning for automatic interpretation of subsurface geologic features from 3D surface seismic data View Presentation

Seismic method is one of the most suited geophysical methods, which provides quite accurate information on subsurface structures and properties from surface measurement. This has been widely used for exploration of hydrocarbons and coal seams, identification of mineralized prospects, understanding geo-tectonics, comprehending earthquake processes, and assessment of ground water contamination. A phenomenal growth of processing/modeling of voluminous data has been possible due to availability of high performance computing system to generate improved images of subsurface. However, human analysts struggle in interpreting such volume of data, when the subsurface is geologically complex. Is it possible to automate the process of interpretation? To find the answer, we have adopted the concept of AI/ML, which is being employed in almost all fields of Science, Technology and Medicines for quick analysis and decision making. We have computed a new attribute, called meta-attribute, by fusing a number of other seismic attributes that are associated with a specific geologic feature. We shall demonstrate the application for automatic delimitation of subsurface geologic features such as fault network, gas plumes, intrusive, magmatic sills & plumbing, fluid migration, mass transport deposit etc. for quick and advanced interpretation of 3D seismic data with much reduced intervention by a human analyst.