Seismic and petrophysical logs.

Seismic Interpretation is an area of Oil & Gas, which deals with 2- or 3-dimensional acoustic images of the subsurface. These images are collected by exploding dynamite shells or sending vibrations into the ground. These generate acoustic waves, which travel through the subsurface, refract and reflect from the variety of layers. The recording of these waves allows a 2D/3D image to be generated. Such images allow Geoscientists to locate hydrocarbon reservoirs and determine their properties without having to drill multiple wells. Providing advanced tools that would enhance the resolution of and detect such regions is completed with the help of ML.

Another area of geophysical imaging involves recording the rock properties around a well using a variety of electromagnetic and acoustic methods (spanning from gamma-ray to magnetic and seismic imaging). Identifying petrophysical properties and rock types based on such geophysical data is a tedious and time consuming process and can be automated to a high degree with various ML methods.

Once the facies are identified, nearby wells can be merged together in 2D or 3D profiles, referred to as sections. there is a significant amount of uncertainty in how these should be interpolated / extrapolated. We combine the historical data on surrounding geology, interpreted nearby sesmic sections and log data to create data-driven section interpolations with a variety of ML techniques.

Identifying and locating microseismic anomalies (such earthquakes or hydraulic fractures during unconventional O&G operations) is another area of ML application. Earthquakes cause billions of dollars worth of damage and countless human casualties. Contamination of groundwater with hydraulic fluids causes long-lasting drinking water problems and is taken very seriously by O&G companies and by the communities affected by O&G activity. These problems are similar in their fundamental sense, with other types of geophysical data being used for earthquake detection as well. Seismic modeling (modeling the behaviour of foundations when encountering earthquake shocks) is another area that is being transformed by ML, as the current approaches used in the industry simplify the problem greatly and do not fit in the complex foundation systems used in modern foundation design.

Medical imaging, such as MRI/X-rays, although seemingly a different industry poses the same problem as seismic interpretation: identifying a certain region in 3D/2D image (generated via electromagnetic waves). Automation of analysis of such scans allows the data to be processed both faster and more accurate.