GeoSeminar

TU students and faculty join Geosciences for a talk at GeoSeminar with Abatan, Ph.D., to learn more about machine learning algorithms.

The title: Data Science Application for Geosciences: Insights from Marcellus Shale Oil & Gas Asset Development and Allegheny Sandstones Paleo-fluvial Channel Analogue, NE United States.

Abstract:
Machine learning algorithms have become increasingly popular in geoscience applications, K-means clustering has emerged as a powerful tool for characterizing subsurface rock types using well log data. Geological exploration often involves the acquisition of vast amounts of well log data from drilling operations, which can be challenging to interpret manually. K-means clustering offers a data-driven approach to group similar rock types based on various log measurements, thereby simplifying the classification process. By partitioning the data into distinct clusters, geologists can gain valuable insights into subsurface lithology and better understand the geological composition of the Earth’s crust. This method not only streamlines the identification of rock types but also aids in making informed decisions for resource exploration and extraction, contributing to more efficient and sustainable practices in the field of geoscience.
In addition to rock type characterization, machine learning techniques such as Support Vector Machine (SVM) regression have found applications in predicting paleo-channel depth using cross-set thickness data. Paleo-channel depth prediction is crucial for understanding ancient river systems, which can provide insights into past environmental conditions and depositional processes. SVM regression, with its ability to model complex relationships in data, is well-suited for this task. By training on cross-set thickness data and associated geological attributes, SVM regression models can accurately estimate the depth of paleo-channels, shedding light on the history of river systems and aiding in geological reconstructions.

Brief Bio:
I work with BP America as a data scientist in the Innovation and Engineering (I&E) Group. I completed my Ph.D. in Sedimentology and Stratigraphy at West Virginia University in January 2020. My research focused on process-based analysis of fluvial stratigraphic record. I got my masters degree in Geosciences from the University of Tulsa in August 2013, where I focused on facies analysis of paleo-fluvial channels. I currently use statistics and machine learning to optimize production on BP assets in the Gulf of Mexico.