Voici les éléments 1 - 3 sur 3
  • Publication
    Accès libre
    Automatic Parameter Tuning of Multiple-Point Statistical Simulations for Lateritic Bauxite Deposits
    (2018)
    Yasin Dagasan
    ;
    ; ;
    Oktay Erten
    ;
    Erkan Topal
    The application of multiple-point statistics (MPS) in the mining industry is not yet widespread and there are very few applications so far. In this paper, we focus on the problem of algorithmic input parameter selection, which is required to perform MPS simulations. The usual approach for selecting the parameters is to conduct a manual sensitivity analysis by testing a set of parameters and evaluating the resulting simulation qualities. However, carrying out such a sensitivity analysis may require significant time and effort. The purpose of this paper is to propose a novel approach to automate the parameter tuning process. The primary criterion used to select the parameters is the reproduction of the conditioning data patterns in the simulated image. The parameters of the MPS algorithm are obtained by iteratively optimising an objective function with simulated annealing. The objective function quantifies the dissimilarity between the pattern statistics of the conditioning data and the simulation image in two steps: the pattern statistics are first obtained using a smooth histogram method; then, the difference between the histograms is evaluated by computing the Jensen–Shanon divergence. The proposed approach is applied for the simulation of the geological interface (footwall contact) within a laterite-type bauxite mine deposit using the Direct Sampling MPS algorithm. The results point out two main advantages: (1) a faster parameter tuning process and (2) more objective determination of the parameters.
  • Publication
    Accès libre
    Stochastic heterogeneity modeling of braided river aquifers: a methodology based on multiple point statistics and analog data
    In this thesis a new pseudo-genetic method to model the heterogeneity of sandy gravel braided-river aquifers is proposed. It is tested and compared with other modeling approaches on a case study of contaminant transport. Indeed, in Switzerland or in mountainous regions, braided-river aquifers represent an important water resource that need to be preserved. In order to manage this resource, a good understanding of groundwater flow and transport in braided-river aquifers is necessary. As the complex heterogeneity of such sedimentary deposits strongly influences the groundwater flow and transport, groundwater behavior predictions need to rely on a wide spectrum of geological model realizations.
    To achieve realistic sedimentary deposits modeling of braided river aquifers, the proposed pseudo-genetic algorithm combines the use of analogue data with Multiple-Point Statistics and process-imitating methods. The integration of analogue data is a key feature to provide additional, complementary and necessary information in the modeling process. Assuredly, hydrogeologist are often subject to field data scarcity because of budget, time and field constraints. Multiple-Points Statistics recent algorithms, on one hand, allow to produce realistic stochastic realizations from training set with complex structures and at the same time allow to honor easily conditioning data. On the other hand, process-imitating methods allow to generate realistic patterns by mimicking physical processes.
    The proposed pseudo-genetic algorithm consists of two main steps. The first step is to build main geological units by stacking successive topography realizations one above the other. So, it mimics the successive large flood events contributing to the formation of the sedimentary deposits. The successive topographies are Multiple-Point Statistics realizations from a training set composed of Digital Elevation Models of an analogue braided-river at different time steps. Each topography is generated conditionally to the previous one. The second step is to generate fine scale heterogeneity within the main geological units. This is performed for each geological unit by iterative deformations of the unit bottom surface, imitating so the process of scour filling. With three main parameters, the aggradation rate, the number of successive iterations and the intensity of the deformations, the algorithm allows to produce a wide range of realistic cross-stratified sedimentary deposits.
    The method is tested in a contaminant transport example, using as reference Tritium tracer experiment concentration data from MADE site, Columbus, Mississippi, USA. In this test case, an assumption of data scarcity is made. Analogue data are integrated in the geological modeling process to determine the input parameters required -- characteristic dimensions and conductivity statistical properties -- for two variants of the proposed pseudo-genetic algorithm as well as for multi-gaussian simulation and object based methods. For each conceptual model, flow and transport simulations are run over 200 geological model realizations to cover a part of the uncertainty due to the input parameters. A comparison of the plume behavior prediction is performed between the different conceptual models.
    The results show that geological structures strongly influence the plume behavior, therefore the choice or the restriction to specific conceptual models will impact the prediction uncertainty. Though little information are available for the modeler, it is possible to achieve reasonable predictions by using analogue data. Of course, with limited information, it is impossible to make an accurate prediction to match the reference, and none of each conceptual model produces better predictions but all are useful to cover the uncertainty range. The results also underline the need to consider a wide exploration of the input parameters for the various conceptual models in order to recover the uncertainty.
  • Publication
    Accès libre
    Probability aggregaton methods and multiple-point statistics for 3D modeling of aquifer heterogeneity from 2D training images
    (2011)
    Comunian, Alessandro
    ;
    ;
    Multiple-point statistics (MPS) is a rising method for the characterization of heterogeneity. Its strength and its Achilles' heel lie in the training image, which is the conceptual model of geological heterogeneity on which MPS simulations are based. Indeed, on one side the use of the training image allows great flexibility when for example for bi-dimensional (2D) simulation a training image can be provided by a photo-mosaic of an outcrop or by a sketch drawn by a geologist. On the other side, in three-dimensions (3D) a training image is rarely available.

    When the information provided by a 3D image is not accessible, then one must somehow use probabilistic information which comes from lower dimen- sion sources, like for example 2D training images. If different 2D sources of information are available, one possibility is to aggregate the corresponding probability information. This problem is very general and several methods exist. Two main categories of methods are distinguished: those based on the sum (convex) and those based on the multiplication (non-convex). When the weighting factors can be determined from some training data, the best reliabilities are obtained with the Beta-transformed linear pool and the Bor- diey's formula. Instead, when training data are not accessible, reasonably reliable results can be obtained with the Bordley's formula and with the Markovian-type categorical prediction.

    One convex method and one non-convex method are tested for the ag- gregation of information coming from 2D training images. For the tests, one 3D image of a micro-computed tomography of a sandstone and one 3D realization of a fluvio-glacial environment are used as references. Two di- mensional slices of the reference 3D images are used as training images for providing the information to be aggregated with the methods cited above, but also for the simulation with two novel method proposed here. One of this methods is baaed on sequential 2D simulations conditioned by the data computed during the previous simulation steps (method s2Dcd). With this last method it is possible to obtain , without the use of a 3D training im- age, 3D simulations which can be considered close to the reference images according to most comparison criteria considered. Moreover, while the re- sults obtained with the method s2Dcd are close to the results obtained with a MPS simulation which make use of a 3D training image, the CPU time required by s2Dcd is from two to four orders of magnitude smaller than with a traditional 3D simulation. This computational efficiency is a step forward for the introduction of MPS in frameworks which require a great number of realizations in a reasonably restricted amount of time, like for example Monte Carlo methods or stochastic inverse problems.

    Other techniques exists to deal with the simulation in 3D when a 3D training image is not available. One of this techniques, developed in this thesis' framework, is applied for the simulation of the image of the fluvio- glacial aquifer analog used as reference for the tests depicted above. It is a hierarchical technique: six parallel outcrops mapped during a quarry excava- tion serve to recognize geological features at different scales and on different depositional layers. Once the complexity of the observed heterogeneity is simplified, object based techniques are used to simulate 3D training images containing simple shapes. Maps of the orientation of the main geological structures are created by interpolating orientations derived from morpholog- ical analysis. All these information are then included in a MPS simulation framework. The geological heterogeneity reproduced using this technique is realistic and can provide an high-resolution benchmark for fluid flow and transport problems.

    In summary, this thesis demonstrates that is is possible to apply MPS methods obtaining credible results from a geological point of view even in absence of a 3D training image.