Seismic data processing is the backbone of the seismic exploration industry. In the era of AI and deep learning, the methods of seismic data processing have undergone a tremendous revolution. This course will help the audience (potentially students and seismic processing practitioners in the oil and gas industry) stay up-to-date with the latest advances in deep-learning-based seismic data processing.
Duration
Two half days
Intended Audience
Students, young professionals, seismic data processing
Prerequisites (Knowledge/Experience/Education Required)
Basic knowledge of machine learning, deep learning, and supervised/unsupervised learning methods.
Course Summary
Deep learning has achieved numerous successes in all aspects of seismic exploration, particularly in labor-intensive seismic processing tasks such as first-arrival picking and NMO velocity picking. In this course, we will provide an overview of the most state-of-the-art applications of deep learning in common tasks spanning the entire seismic data processing workflow and guide the audience through a deep dive into the implementation details of mainstream deep learning approaches for advanced seismic data processing.
The following processing tasks will be covered:
- Automatic first arrival picking for traveltime tomography;
- Seismic data denoising for random, regular, and erratic noise;
- Seismic data reconstruction from regular, irregular, and aliased grids;
- Multiples attenuation;
- NMO velocity picking;
- Seismic resolution enhancement. The intention of this course is to equip the audience with the basic principles of deep-learning-based seismic data processing and guide them in developing more
Learner Outcomes
Learners will be able to:
- Be able to mathematically define the deep learning problems of common seismic processing tasks
- Be able to write simple codes to reproduce the SOTA deep learning approaches for seismic processing
- Be able to get access to the mainstream open datasets for training deep-learning models
- Be able to design an advanced deep learning approach for solving a seismic processing problem
