From Prestack to Rock Type: A Neural Network Facies Inversion
By Kamal Hami-Eddine
One of the big challenges for hydrocarbon recovery is combining geological information about lithology and geophysical data acquired through reflection seismic. As these data can take different forms (litho-logs, cuttings, and for seismic, post and prestack attributes) and can have different resolutions, the manual integration of all the information contained in them requires long analysis and is sometimes impossible to solve.
We therefore propose a new methodology for predicting lithology interpreted at wells using 3D seismic attributes (post and prestack). This technique aims at finding patterns in seismic attributes that will predict lithology kind, distribution and uncertainty, using a probabilistic approach.
This demonstration illustrates the principles of the methodology, and shows the workflow-based approach embedded in SeisEarth to successfully estimate rock type distributions from prestack data.