Guillaume Caumon

Concurrent number cruncher: a GPU implementation of a general sparse linear solver

A wide class of numerical methods needs to solve a linear system, where the matrix pattern of non-zero coefficients can be arbitrary. These problems can greatly benefit from highly multithreaded computational power and large memory bandwidth available on graphics processor units (GPUs), especially since dedicated general purpose APIs such as close-to-metal (CTM) (AMD–ATI) and compute unified device architecture (CUDA) (NVIDIA) have appeared. CUDA even provides a BLAS implementation, but only for dense matrices (CuBLAS). Other existing linear solvers for the GPU are also limited by their internal matrix representation. This paper describes how to combine recent GPU programming techniques and new GPU dedicated APIs with high performance computing strategies (namely block compressed row storage (BCRS), register blocking and vectorization), to implement a sparse general-purpose linear solver. Our implementation of the Jacobi-preconditioned conjugate gradient algorithm outperforms by up to a factor of 6.0 × leading-edge CPU counterparts, making it attractive for applications which are content with single precision.

Curvature Attribute from Surface-Restoration as Predictor Variable in Kupferschiefer Copper Potentials

This work explains a procedure to predict Cu potentials in the ore-Kupferschiefer using structural surface-restoration and logistic regression (LR) analysis. The predictor in the assessments are established from the restored horizon that contains the ore-series. Applying flexural-slip to unfold/unfault the 3D model of the Fore-Sudetic Monocline, we obtained curvature for each restored time. We found that curvature represents one of the main structural features related to the Cu mineralization. Maximum curvature corresponds to high internal deformation in the restored layers, evidencing faulting and damaged areas in the 3D model. Thus, curvature may highlight fault systems that drove fluid circulation from the basement and host the early mineralization stages. In the Cu potential modeling, curvature, distance to the Fore-Sudetic Block and depth of restored Zechstein at Cretaceous time are used as predictors and proven Cu-potential areas as targets. Then, we applied LR analysis establishing the separating function between mineralized and non-mineralized locations. The LR models show positive correspondence between predicted probabilities of Cu-potentials and curvature estimated on the surface depicting the mineralized layer. Nevertheless, predicted probabilities are particularly higher using curvatures obtained from Late Paleozoic and Late Triassic restorations.

Three-Dimensional Implicit Stratigraphic Model Building From Remote Sensing Data on Tetrahedral Meshes: Theory and Application to a Regional Model of La Popa Basin, NE Mexico

Remote sensing data provide significant information to constrain the geometry of geological structures at depth. However, the use of intraformational geomorphologic features such as flatirons and incised valleys often calls for tedious user interaction during 3-D model building. We propose a new method to generate 3-D models of stratigraphic formations, based primarily on remote sensing images and digital elevation models. This method is based on interpretations of the main relief markers and interpolation of a stratigraphic property on a tetahedral mesh covering the domain of study. The tetrahedral mesh provides a convenient way to integrate available data during the interpolation while accounting for discontinuities such as faults. Interpretive expert input may be provided through constrained interactive editing on arbitrary cross-sections, and additional surface or subsurface data may also be integrated in the modeling. We demonstrate this global workflow on a structurally complex basin in the Sierra Madre Oriental, Northeastern Mexico.

Method for Stochastic Inverse Modeling of Fault Geometry and Connectivity Using Flow Data

This paper focuses on fault-related uncertainties in the subsurface, which can significantly affect the numerical simulation of physical processes. Our goal is to use dynamic data and process-based simulation to update structural uncertainty in a Bayesian inverse approach. We propose a stochastic fault model where the number and features of faults are made variable. In particular, this model samples uncertainties about connectivity between the faults. The stochastic three dimensional fault model is integrated within a stochastic inversion scheme in order to reduce uncertainties about fault characteristics and fault zone layout, by minimizing the mismatch between observed and simulated data.The stochastic fault model uses a priori information such as fault orientation, location, size and sinuosity, to sample both geometrical and topological uncertainties with realistic fault descriptions. Each fault object is parameterized by the random vector used to simulate fault features. Then, during inversion, the random vector of the current model is stochastically perturbed, producing a new parameter vector used as input by the stochastic fault model to produce a new model. Even if the topology varies from one model to another, the algorithm produces correlated models so that their flow responses evolve quite smoothly.The methodology is applicable in general and illustrated on a synthetic two-phase flow example. A first set of models is generated to sample the prior uncertainty space. Then models minimizing reference water-saturation data misfit are used as seeds to generate continuous Monte Carlo Markov Chains (MCMC) of models with discrete states. Posterior models reduce uncertainties about fault position, while the topology varies from one model to another. A second example highlights the interest of the parameterization when interpreted data are available, by perturbing geological scenarios and falsifying those that do not match two-phase flow observations.

Automatic surface remeshing of 3D structural models at specified resolution

We propose a method to remesh the surfaces of 3D sealed geological structural models for subsequent volumetric meshing. The input of the method is a set of triangulated surfaces that are in contact along given lines and at given points. The output is a set of surfaces meshed with triangles as equilateral as possible. The method relies on a global Centroidal Voronoi optimization to place the vertices of the final surfaces combined with combinatorial considerations to either recover or simplify the surfaces, lines and points of the input model. When the final resolution is sufficient, the input contact lines and points are also contact lines and points of the final model. However, when dealing with models with complex contacts, resolution may be insufficient and instead of a refinement strategy that may lead to too many points, we propose to locally merge some features of the input model. This ability to simplify the input model is particularly interesting when the model is to be volumetrically meshed. The method is demonstrated on twelve structural models, including seven models built with an implicit modeling method, and one folded layer model affected by a discrete fracture network. HighlightsA surface remeshing method for 3D sealed geological structural models.This method is fully automatic, handles contacts properly and coherently.It generates a mesh at specified resolution and with high quality elements.Results on twelve models are evaluated and shown.

A methodology for pseudo-genetic stochastic modeling of discrete fracture networks

Stochastic simulation of fracture systems is an interesting approach to build a set of dense and complex networks. However, discrete fracture models made of planar fractures generally fail to reproduce the complexity of natural networks, both in terms of geometry and connectivity. In this study a pseudo-genetic method is developed to generate stochastic fracture models that are consistent with patterns observed on outcrops and fracture growth principles. The main idea is to simulate evolving fracture networks through geometric proxies by iteratively growing 3D fractures. The algorithm defines heuristic rules in order to mimic the mechanics of fracture initiation, propagation, interaction and termination. The growth process enhances the production of linking structure and impacts the connectivity of fracture networks. A sensitivity study is performed on synthetic examples. The method produces unbiased fracture dip and strike statistics and qualitatively reproduces the fracture density map. The fracture length distribution law is underestimated because of the early stop in fracture growth after intersection. Highlights? We propose a stochastic fracture simulation method mimicking mechanical behavior. ? Non-stationary information can be integrated in the simulation. ? Resulting fractures are better connected than in classical planar networks.

Stochastic structural modelling in sparse data situations

This paper introduces a stochastic structural modelling method that honours interpretations of both faults and stratigraphic horizons on maps and cross-sections in conjunction with prior information, such as fault orientation and statistical size–displacement relationships. The generated stochastic models sample not only geometric uncertainty but also topological uncertainty about the fault network. Faults are simulated sequentially; at each step, fault traces are randomly chosen to constrain a fault surface in order to obtain consistent fault geometry and displacement profile. For each simulated fault network, stratigraphic modelling is performed to honour interpreted horizons using an implicit approach. Geometrical uncertainty on stratigraphic horizons can then be simulated by adding a correlated random noise to the stratigraphic scalar field. This strategy automatically maintains the continuity between faults and horizons. The method is applied to a Middle East field where stochastic structural models are generated from interpreted two-dimensional (2D) seismic lines, first by representing only stratigraphic uncertainty and then by adding uncertainty about the fault network. These two scenarios are compared in terms of gross rock volume (GRV) uncertainty and show a significant increase in GRV uncertainty when fault uncertainties are considered. This underlines the key role of faults in resource estimation uncertainties and advocates a more systematic fault uncertainty consideration in subsurface studies, especially in settings in which the data are sparse.

Elements for measuring the complexity of 3D structural models

The reliable modeling of three-dimensional complex geological structures can have a major impact on forecasting and managing natural resources and on predicting seismic and geomechanical hazards. However, the qualification of a model as structurally complex is often qualitative and subjective making the comparison of the capabilities and performances of various geomodeling methods or software difficult. In this paper, we consider the notion of structural complexity from a geometrical point of view and argue that it can be characterized using general metrics computed on three-dimensional sealed structural models. We propose global and local measures of the connectivity and of the geometry of the model components and show how they permit to classify nine 3D synthetic structural models. Depending on the complexity elements favored, the classification varies. The models we introduce could be used as benchmark models for geomodeling algorithms. HighlightsWe describe complicated geometrical configurations in structural models.Nine benchmark 3D structural models for geomodeling methods are proposed.We propose four measures to compare geometry and connectivity of structural models.Several classifications of the proposed benchmark models are given.

Visualization of grids conforming to geological structures: a topological approach

Flexible grids are used in many Geoscience applications because they can accurately adapt to the great diversity of shapes encountered in nature. These grids raise a number difficult challenges, in particular for fast volume visualization. We propose a generic incremental slicing algorithm for versatile visualization of unstructured grids, these being constituted of arbitrary convex cells. The tradeoff between the complexity of the grid and the efficiency of the method is addressed by special-purpose data structures and customizations. A general structure based on oriented edges is defined to address the general case. When only a limited number of polyhedron types is present in the grid (zoo grids), memory usage and rendering time are reduced by using a catalog of cell types generated automatically. This data structure is further optimized to deal with stratigraphic grids made of hexahedral cells. The visualization method is applied to several gridded subsurface models conforming to geological structures.

Modeling Channel Forms and Related Sedimentary Objects Using a Boundary Representation Based on Non-uniform Rational B-Splines

This paper aims at providing a flexible and compact volumetric object model capable of representing many sedimentary structures at different scales. Geobodies are defined by a boundary representation; each bounding surface is constructed as a parametric deformable surface. A three-dimensional sedimentary object with a compact parametrization which allows for representing various geometries and provides a curvilinear framework for modeling internal heterogeneities is proposed. This representation is based on non-uniform rational basis splineswhich smoothly interpolate between a set of points. The three-dimensional models of geobodies are generated using a small number of parameters, and hence can be easily modified. This can be done by a point and click user interaction for manual editing or by a Monte-Carlo sampling for stochastic simulation. Each elementary shape is controlled by deformation rules and has connection constraints with associated objects to maintain geometric consistency through editing. The boundary representations of the different sedimentary structures are used to construct hexahedral conformal grids to perform petrophysical property simulations following the particular three-dimensional parametric space of each object. Finally these properties can be upscaled, according to erosion rules, to a global grid that represents the global depositional environment.