A critical assessment of fault growth models using Machine Learning tools
This research project is in competition for funding with one or more projects available across the NERC GW4 Doctoral Training Partnership (DTP). Usually the projects which receive the best applicants will be awarded the funding. Find out more information about the DTP and how to apply.
Application deadline: 7 January 2019
Start date: October 2019
DTP research theme: Dynamic Earth
Lithological heterogeneities in subsurface units are capable of distributing stresses in distinct ways, leading to the nucleation of faults and fractures in specific stratigraphic intervals. While easily documented at outcrop, the structural complexity induced by heterogeneous rocks can only be assessed in the subsurface by using attribute-based technology on seismic data.
With that in mind, this project will develop novel Machine Learning tools which will be implemented in 1st-generation prediction tools embedded in Ikon Science software. The final aim is to use the new Machine Learning tools to predict the geometry and evolution of fault and fracture patterns around subsurface structures (e.g. salt diapirs, footwall blocks, volcanic domes), and thus challenge the current models for fault growth and propagation around such structures.
Such an approach is key to the assessment of structural deformation in hydrocarbon reservoirs, seal units overlying these same reservoirs, for the optimal management of nuclear repositories, and to ‘de-risk’ Carbon Capture and Storage (CCS) sites.
Project aims and methods
Research at Ikon Science has been concentrated on the use of advanced seismic attributes and image processing techniques to estimate subsurface structural deformation. Structure tensors and curvature attributes are specifically used to develop interpolation algorithms to populate the modelling grid with well log data (Naeini and Hale, 2015). Building upon this early work, the PhD project proposed here will develop Machine Learning tools and apply them to the quantitative analysis of high-quality seismic data from SE Brazil, Norway, New Zealand, and the southern North Sea, all of which comprise hydrocarbon prospects and associated CCS sites.
The evolution of faults systems around large tectonic features, in the four study areas, will be assessed in a first stage using Machine Learning tools and in particular Deep Learning techniques. In a second stage, we will tie borehole information with seismic attribute data following Naeini and Hale (2015) methodology, so that we add lithological information to the seismic data and understand what volumes of rock are the most stressed around complex heterogeneous successions.
The aims of this project are to:
- improve the existing algorithms and develop new Machine Learning Algorithms for subsequent use in Ikon’s RokDoc platform
- test the software against (and make reliable observations from) seismic data, challenging the current models for fault growth in sedimentary basins
- assess leakage and deformation styles around interpreted fault types; 4) develop new algorithms useful in the selection of hydrocarbon drilling targets and CCS sites.
You will drive the project and will concentrate on specific areas (and structures) of your immediate interest. Hence, you will have total freedom to choose the research direction that is more appropriate to your skills.
You will have to demonstrate a high level of computing and mathematics skills plus some knowledge on how sedimentary basins deform in time and space. Knowledge of Python computing language is an advantage. All in all, we intend to capture a student with an existing degree in computing and/or mathematics and ideally some degree of experience in basin analysis.
CASE or Collaborative partner
You will spend a minimum of 3 months at the CASE or Collaborative Partner, Ikon Science, in London. Dr Ehsan Naeini of Ikon Science will also visit Cardiff periodically and we maintain contact via a monthly Skype session with him and with the whole supervisory team (the PhD student, Cardiff and Exeter supervisors).
kon Science has the possibility to train the student on a range of technical features and knowledge that will greatly benefit this project. The second supervisor (Dr. Padraig Cormoran) has vast experience in Machine Learning and mathematical modelling of geographic and earth surface processes. He will regularly train the student in a supervisory basis. Cardiff University also delivers courses on Python and other programming languages to PhD students, whereas Dr Tiago M. Alves (first supervisor) is expert in seismic interpretation and basin analysis. Ikon Science has a variety of licensed and in-house built software that can be used by the student.
References and background reading
- Alves, T.M., Kurtev, K., Moore, G.F., Strasser, M. (2014). Assessing the internal character, reservoir potential and seal competence of mass-transport deposits using seismic texture: a geophysical and petrophysical approach. American Association of Petroleum Geologists 98, 793-824.
- Marfurt, K.J. and Alves, T.M. (2015). Pitfalls and limitations in seismic attribute interpretation of tectonic features. Interpretation 3(1), SB5-SB15.
- Naeini, E.Z. and Hale, D. (2015). Image- and horizon-guided interpolation. Geophysics 80, 47-56.
Ze, T. and Alves, T.M. (2016). The role of gravitational collapse in controlling the evolution of crestal fault systems (Espírito Santo Basin, SE Brazil). Journal of Structural Geology 92, 79-98.
Prof. Stephen Hesselbo, Camborne School of Mines, University of Exeter