Optimal seismic acquisition design, reconstruction and denoising

• Author / Creator

  Guo, Yi

• Seismic acquisition constitutes a significant economic commitment, accounting for up to 80% of the overall cost of seismic exploration. This cost is intrinsically linked to the quantity of deployed sensors and sources, each carrying its own set of expenses related to acquisition, deployment, and maintenance. With increasing demands to shrink seismic acquisition expenses, minimize ecological impacts, and adhere to health, safety, and environmental (HSE) guidelines, innovative methodologies like compressive sensing (CS) seismic acquisition have emerged. These techniques aim at reducing the number of sensors while still fulfilling the aforementioned objectives. However, practical field constraints can hamper the random sampling strategies typically employed by CS.
  Addressing the challenges of elevated operational costs, diminished efficiency, and low sampling density in seismic field acquisition, this dissertation concentrates on the formulation of optimal, cost-effective seismic acquisition layouts to address the challenge of maintaining data quality while utilizing fewer sensors, as compared to conventional high-density surveys. Within this context, the thesis elaborates on two central paradigms — CS seismic acquisition and optimal sparse sensing seismic acquisition — each applicable under different conditions of prior information availability and field operational constraints.
  Chapter 2 discusses the CS theory and its application in seismic acquisition, considering data-free survey design. In the realm of data-driven methods, this work embarks on a nuanced examination of diverse strategies for optimization, encapsulated in Chapters 3 and 4. An over-complete pre-trained basis library is also incorporated for data-driven scenarios, facilitating computationally efficient and straightforward data reconstructions. Chapter 5 involves using a data-free objective function for seismic acquisition design. The optimal survey design strategies range from optimal sparse sensing employing QR-column pivoting to more advanced methodologies rooted in reinforcement learning (RL) and deep reinforcement learning (DRL). Noteworthy contributions include the development of a novel theoretical framework for synchronous spatiotemporal compression through design in CS, resulting in a theoretical acquisition cost reduction exceeding 50%. Additionally, an RL-based optimal acquisition design algorithm is introduced, achieving a theoretical reduction in acquisition costs exceeding 65% via the optimization of field sampling points. Particular emphasis is placed on RL and DRL techniques, which are operationalized through the formulation of a Markov decision process (MDP) model for sensor placement decisions.
  Empirical validations across diverse application scenarios — including ocean bottom node (OBN) survey, simultaneous source acquisition, time-lapse studies for carbon dioxide storage monitoring, and vibroseis route design — indicate the high efficacy of the proposed methodologies in significantly reducing acquisition costs. The innovative approaches proposed herein hold transformative potential not only for the realm of seismic acquisition but also extend to other domains of geophysical exploration.

• Subjects / Keywords

  • Seismic acquisition design
  • reconstruction
  • denoising
  • compressive sensing
  • reinforcement learning

• Graduation date

  Spring 2024

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/r3-55p8-fm79

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.