- Home
- Conferences
- Conference Proceedings
- Conferences
82nd EAGE Annual Conference & Exhibition - Workshop Programme
- Conference date: October 18-21, 2021
- Location: Amsterdam, Netherlands / Online
- Published: 18 October 2021
41 - 45 of 45 results
-
-
Deployment of Machine Learning Solutions to Production Seismic Processing
By R. HeggeSummaryOnce machine learning solutions have proven themselves in a research environment, they generally need additional work before they can be used in production.
-
-
-
Growing up: On Productizing Sub-Surface Machine Learning Workflows
By J. LimbeckSummaryIn this presentation we cover the main pillars of the strategy we are following in Shell that allows more rapid movement from data science proof of concepts to the deployment stage where value for the business is generated. Where applicable, the concepts are illustrated with examples.
-
-
-
Deriving High Fidelity Velocity Model using Acoustic Full Waveform Inversion
More LessSummaryIn this abstract, we present the results of dynamic matching full waveform inversion (DMFWI) applied to various data sets from different geological settings. DMFWI is an acoustic algorithm and concentrates on inverting the kinematic difference while minimizing the impact of amplitude. With proper constraints, this algorithm is effective in inverting for highly accurate velocity models.
-
-
-
Incorporating Probabilistic Petrophysical Information into Viscoelastic Full Waveform Inversion
More LessSummaryFull waveform inversion (FWI) augmented with petrophysical information enhances wavefield-based velocity model building. In regions of shallow low-velocity anomalies, estimating accurate and lithologically feasible viscoelastic models is crucial for accurate seismic processing and imaging. However, inverting for velocity and attenuation models simultaneously is a challenging task because of the severe interparameter crosstalk between these parameters. We derive viscoelastic models of subsurface properties using viscoelastic FWI and explicitly incorporate petrophysical penalties to guide models toward realistic lithology, i.e., to models consistent with the seismic data as well as with the petrophysical context in the area of study. This methodology mitigates artifacts created by interparameter crosstalk, and prevents geologically implausible earth models. We define this penalty using multiple probability density functions (PDFs) derived from petrophysical information, such as well logs. In order to formulate the penalty term, we build each PDF considering spatial distribution patterns in the petrophysical data. With a realistic synthetic example, we demonstrate that the combined inversion objective function establishes a robust foundation for viscoelastic FWI by explicitly guiding models toward plausible solutions in the specific geological context of the exploration problem.
-
-
-
Three-dimensional Elastic Model Building from Ambient Noise Seismic: a Case Study from Southern Oman
Authors M. Danilouchkine, A. Adwani, R. Plessix, F. Ten Kroode, Q. Al-Siyabi, F. Ernst and O. Al DroushiSummaryIn many areas in the Middle East the presence of large vertical variations of lithologies gives rise to strong multiples and mode conversions of primary P-wave reflections and leads to difficulties in imaging the subsurface. An accurate model of the subsurface is highly desirable
to circumvent this problem. In this paper we present a successful attempt to model building, based on elastic Full Waveform Inversion (eFWI) and ambient noise seismic interferometry (ANSI). An unorthodox deployment of the latter technology on continuously acquired land data made it possible to extract the ground roll and to subsequently invert it in the frequency range up to 1.2 Hz. The study was conducted using the real 3d seismic dataset, acquired in the Southern Oman. The obtained low-resolution elastic model extends to a depth of 4 km, captures the major velocity contrasts in the shallow and deep subsurface, and conforms well with interpreted subsurface interfaces.
-