Name: Ramy Abdallah
Type: User
Company: Unversity of Aberdeen
Bio: Data Geoscientist | DTS, DAS, and VSP Processing and Interpretation | PhD Candidate in Geology Reducing uncertainty of subsurface interpretations
Twitter: RamySE_Abdallah
Location: Aberdeen Scotland
Blog: https://jovian.com/ramysaleem
Ramy Abdallah's Projects
Using interpreted geological cross-sections to generate synthetic seismic sections.
Datasets are essential for all scientists to conduct research, especially coders geoscientists. Research is an inquiry-based process that includes recognising a question, gathering data, analysing and evaluating results, drawing conclusions, and sharing the knowledge gained. The ability to conduct research mainly depends on the datasets. There are massive open-source data available online. However, it is often challenging for students and researchers to navigate the datasets, collect the data and download it. Because mainly data discoverability is poor, documentation is sometimes lacking, and licences can be confusing. I hope with these two projects to add toward the solution of these problems.
Full Stack Data Science Masters - Practice
Generally, we digitise legacy data in geosciences, such as interpreted geological or seismic sections. Digitisation is creating a computerised representation of a printed analogue.
Exploring 118 wells of 1 MM+ rows and 29 columns of wireline petrophysical data using the Pandas library. Analysed & Visualised wireline logs petrophysical dataset using - Pandas, Numpy, Matplotlib, Plotly & seaborn libraries Discovered insights of wireline logs quality & interpretation (missing data and imbalance class
The application of AI in structural interpretation workflows is still ambiguous. This study aims to use ML techniques to classify images of folds and fold-thrust structures. Here we show that convolutional neural networks as supervised deep learning techniques provide excellent algorithms to discriminate between geological image datasets.
This model is a simple method to classify, quantify and illustrate the uncertainty in subsurface interpretation using python open source libraries. This model takes a practical and coding focused approach to visualise the uncertainty in subsurface interpretations by calculating zones and levels of uncertainty—the Five uncertainty zones created by measuring the distance from outcrops, galleries and boreholes.
Analysing Unstructured Geosciences Data for a Changing World.
In this study we try to predict loan state of our customers as a classification problem.
To identify lithologies, geoscientists use subsurface data such as wireline logs and petrophysical data. However, this process is often tedious, repetitive, and time-consuming. This project aims to use machine learning techniques to predict lithology from petrophysical logs, which are direct indicators of lithology.
This project will explore, analyse and visualise publicly available wells datasets from the United States offshore data centre, the USGS boreholes website - Bureau of Safety and Environmental Enforcement (BSEE) https://www.data.bsee.gov/Main/Default.aspx with a particular focus on the Gulf of Mexico (GOM) wells. This project will study sandstones quality as a reservoir, the production history of the operators on the Gulf of Mexico and a well summary report to highlight any possible problem. The reservoir quality analysis will examine relationships between average values of porosity, permeability, depth, temperature, pressure, thickness, age, and play type for data files from 2009 until 2019.The porosity plotted and shown in a wide range of plots as a function of permeability and burial depth. Also, the median (P50) porosity will be plotted against depth to examine the porosity trend. Moreover, this project will investigate the companies oil and gas production in the gulf of Mexico for the last five years. Lastly, the analysis will include an investigation of well summary reports of five wells. The project will include web scrapping to collect online well summary reports to generate a word cloud. The project results can be useful for specifying realistic distributions of parameters for both exploration risk evaluation and/or reservoir modelling by machine learning algorithms in the next project.
Identify seismic anomalies with image training
Our project revolutionizes seismic image interpretation with advanced deep learning, using Convolutional Neural Networks (CNNs) to automate workflows and improve accuracy in subsurface exploration. Trained on synthetic seismic data from coal mines, our algorithm excels in identifying faults, folds, and flat layers.
Geological Inference Through Displacement-Distance Profile Guided by Machine Learning.
This study offers critical automatic workflows to examine the displacement–distance profiles and insert the mechanical stratigraphy.