Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2010.02087

Machine learning on Crays to optimise petrophysical workflows in oil and gas exploration

Nick Brown·Anna Roubickova·Ioanna Lampaki·Lucy MacGregor·Michelle Ellis·Paola Vera de Newton
View original source ↗Read source document ↗Export metadata ↓

Abstract

The oil and gas industry is awash with sub-surface data, which is used to characterize the rock and fluid properties beneath the seabed. This in turn drives commercial decision making and exploration, but the industry currently relies upon highly manual workflows when processing data. A key question is whether this can be improved using machine learning to complement the activities of petrophysicists searching for hydrocarbons. In this paper we present work done, in collaboration with Rock Solid Images (RSI), using supervised machine learning on a Cray XC30 to train models that streamline the manual data interpretation process. With a general aim of decreasing the petrophysical interpretation time down from over 7 days to 7 minutes, in this paper we describe the use of mathematical models that have been trained using raw well log data, for completing each of the four stages of a petrophysical interpretation workflow, along with initial data cleaning. We explore how the predictions from these models compare against the interpretations of human petrophysicists, along with numerous options and techniques that were used to optimise the prediction of our models. The power provided by modern supercomputers such as Cray machines is crucial here, but some popular machine learning framework are unable to take full advantage of modern HPC machines. As such we will also explore the suitability of the machine learning tools we have used, and describe steps we took to work round their limitations. The result of this work is the ability, for the first time, to use machine learning for the entire petrophysical workflow. Whilst there are numerous challenges, limitations and caveats, we demonstrate that machine learning has an important role to play in the processing of sub-surface data.

Explore related subjects

physics.geo-phcs.LG

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Nick Brown, Anna Roubickova, Ioanna Lampaki, Lucy MacGregor, Michelle Ellis, Paola Vera de Newton. 2020-10-01. Machine learning on Crays to optimise petrophysical workflows in oil and gas exploration. https://doi.org/10.1002/cpe.5655

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Assessing foundational atomistic models for iron alloys under Earth's core conditions

We assess the capability of recently developed foundational atomistic models (FAMs) to simulate iron alloys under the extreme pressures and temperatures of Earth's core. Static equations of state for hexagonal close-packed (hcp) and body-centered cubic (bcc) iron, computed using 17 FAMs, are benchmarked against ab initio calculations. Two representative models, MatterSim and MACE, are further evaluated for their ability to reproduce phonon spectra, liquid structure, and melting relations of iron at core conditions. While both models capture several key properties, MACE substantially overestimates the stability of bcc iron and fails to correctly describe the stability of hcp iron. Their performance is also examined for binary liquids, superionic phases, and a seven-component Fe-Ni-Si-S-O-H-C liquid. Although these FAMs were not explicitly trained on data from core conditions, they can reproduce several structural and dynamical properties across a wide range of compositions. However, none of the tested models consistently reproduces all first-principles benchmarks. By analyzing the origins of these discrepancies, we identify several limitations of current FAMs, particularly the lack of an explicit treatment of thermal electronic excitations, which significantly affect phase stability and thermodynamic properties under core conditions. We further discuss directions for improving FAMs to enable predictive simulations of core-forming materials under extreme conditions.

physics.geo-ph↗
arXiv2026

Time-Resolved Surface-Fault Displacement During the 2026 Kumamoto Earthquake From Near-Fault Video

Video recordings can reveal how rapidly fault displacement develops at the Earth's surface, but camera motion and recording artifacts can obscure the ground signal. We analyze a secondary copy of security-camera footage that captured surface displacement during the 28 July 2026 Kumamoto earthquake; the native recording was unavailable and could not be recovered. Two independent image-tracking methods were used. Optical flow follows identifiable image features, whereas normalized cross-correlation (NCC) template matching follows fixed image patches by their similarity. Both measured target-region motion relative to spatially separated reference regions while correcting motion shared by the recording. Image displacement was calibrated to the magnitude of the field-measured offset vector: 1.05 m right-lateral and 0.90 m east-side-up, or 1.383 m in total. We characterize the principal rise by the time required for displacement to progress from 20% to 80% of the selected final level. Across prespecified endpoint choices, optical flow gives 0.866-0.901 s and NCC gives 0.910-0.928 s. These durations correspond to average rates of 0.920-0.958 and 0.895-0.912 m/s, respectively. Checks using independently published tracking windows reproduce the displacement scale, although exact timing is more sensitive in spatially restricted tests. The record also shows an early apparent peak and decline followed by renewed apparent horizontal displacement. Because that later motion may represent either continued ground displacement or the geometry of the secondary recording, neither the permanent endpoint nor physical overshoot can be determined. The most robust conclusion is that the central part of the surface displacement developed in approximately 0.9 s at an average rate near 0.9 m/s.

physics.geo-ph↗
arXiv2026

Shape matters: DEM investigation of geometry-controlled mechanical response in irregular rock fragments under static and dynamic loading

Mechanical characterization of subsurface rock formations typically requires standardized cylindrical core specimens, which are often unavailable from fractured or unconventional reservoir sequences. This study uses a Discrete Element Method (DEM) framework calibrated to Sulphur Mountain Formation siltstone to investigate how fragment geometry governs mechanical response in irregular rock particles, which are direct analogues of drill cuttings. Nineteen specimens (18 procedurally generated irregular geometries and one reference Brazilian disk), normalized to a 10 mm bounding sphere radius, were tested under quasi-static (Brazilian-type) and dynamic (Short Impact Load Cell) loading modes. Four mechanical outputs of failure force (Ff), apparent strength (σf), apparent stiffness (Ka), and stiffness (E) were regressed against seven retained shape descriptors. Results show that force-based quantities are strongly controlled by the surface-area-to-volume (SA/V) ratio under static loading (Adj. R2 = 0.69), while area-normalized quantities (σf and E) are geometrically insensitive. Dynamic loading amplifies geometric sensitivity for failure force and introduces an independent role for surface concavity depth. These findings establish a preliminary quantitative framework for interpreting mechanical measurements from irregular rock fragments when a standardized core is unavailable.

physics.geo-ph↗