Evolving Traditional Subsurface Interpretation with ArtificialIntelligence”
The potential impact of Artificial Intelligence (AI) within E&P organizations is only just starting to be realized. With increasing compute power, improved data processing speeds and advances in technologies (such as deep learning neural networks), it is likely that AI will continue to be a disruptive technology that will change E&P organizations for many years to come. Although AI can be used within many workflows, some of the most significant progress has been within seismic interpretation. In particular, the detection of faults. The identification of seismic faults is a good example of a task that is challenging and time consuming for geoscientists, yet, with the right network architecture and training, is achievable in a short time to a level of completeness and accuracy that is far outside the reach of most interpreters.
In this talk we will present a series of case studies demonstrating workflows that integrate human experience with AI prowess to deliver a step change in the seismic interpretation process enabling improved interpretation in both time taken and accuracy, allowing previously unidentified elements (to date considered sub-seismic) to be identified and understood.
During the presentation we will review comparisons of traditional vs AI fault identification techniques and discuss the bnefits of analysing the seismic date in a truly 3D sense. Revealing the structural elements in 3D can be greatly beneficial to an interpreter, helping them visualize features that would otherwise remain unseen. Best practice for structural interpretation involves interpreting the structures perpendicular to the strike of the faults. However, nature does not always allow for this, with most seismic volumes containing faults of varying orientations. The advent of 3D AI solutions allows the interpreter to work in a traditional 2D environment (in-line and crossline interpretation), whilst having a strong appreciation for the 3D shapes of faults, often an issue which remains with traditional interpretation techniques.Hugo Garcia.jpg
To achieve this technological advance, we have developed bespoke deep learning networks that have been trained using thousands of examples of different fault signatures from different geological environments across varying data quality. The pretrained networks can be applied directly onto unseen data cubes or can be fine-tuned with interpretations from new data sets, capturing additional knowledge from existing interpretations.
Critically, any interpreter changes to AI predicted results are captured by the deep learning network, allowing the network to learn from the interpreter’s experience and knowled and subsequently applied on future data sets.
Hugo has over 10 years industry experience working as a Senior Geoscientist for Geoteric in both Aberdeen and Houston. Hugo is currently leading Geoterics Houston technical team utilizing machine learning and artificial intelligence to deliver detailed interpretations across structurally diverse settings. Hugo gained his MSc degree in Geological resources from Universidade de Aveiro in Portugal where he stayed as a researcher until joining Geoteric in 2011.