Summary: TBA

Summary: TBA

Summary: The author presents an algorithm for numerical modeling of seismic wave field propagation, which consists in combining a difference scheme on shifted grids and the NDM-net neural network to suppress numerical errors in the modeling data caused by both numerical dispersion and discretization of the velocity model of the medium. To assess the accuracy of the algorithm results, numerical experiments were performed using three two-dimensional isotropic media with fundamentally different structures. It is shown that the algorithm allows reducing the numerical modeling time by up to six times compared to the classical method while maintaining an acceptable level of accuracy of the numerical solution. An algorithm for optimizing the training data set using cluster analysis and three metrics was also developed and implemented: the distance between sources, the difference in values ​​​​between seismograms, and the divergence of velocity models of the medium. When using statistical analysis of the results of applying the trained NDM-net neural network on random data samples, a linear combination of the three introduced metrics is introduced to optimize the training data set. Numerical experiments using all types of data samples show that the use of the NDM-net neural network trained with an optimized sample allows suppressing the numerical error by 6 times, while with other types of samples by 4-5 times.

Аннотация: В работе представлен алгоритм численного моделирования распространения сейсмических волновых полей, заключающийся в комбинировании разностной схемы на сдвинутых сетках и нейронной сети NDM-net для подавления численных ошибок в данных моделирования, вызванных как численной дисперсией, так дискретизацией скоростной модели среды. Для оценки точности результатов алгоритма выполнены численные эксперименты с использованием трёх двумерных изотропных сред с принципиально разной структурой. Показано, что алгоритм позволяет до шести раз сократить время численного моделирования по сравнению с классическим  методом при сохранении приемлемого уровня точности численного решения. Также разработан и реализован алгоритм оптимизации обучающей выборки данных с использованием кластерного анализа и трёх метрик: расстояния между источниками, разница в значениях между сейсмограммами и расхождение скоростных моделей среды. При использовании статистического анализа результатов применения обученной нейронной сети NDM-net на случайных выборках данных вводится линейная комбинация трёх введённых метрик для оптимизации обучающей выборки данных. Численные эксперименты с использованием всех видов выборок данных показывают, что применение нейронной сети NDM-net, обученной при оптимизированной выборке позволяет подавить численную ошибку в 6 раз, в то время как при других типов выборок в 4-5 раза.

Summary: TBA

Summary: Recent studies in multiphysical simulation of fluid-saturated rocks involve a modern model of poroelasticity based on a thermodynamically compatible system of conservation laws. The model has some significant advantages in comparison with well-known Biot model of poroelasticity. However, application for some problems (numerical rock sample loading and upscaling of porous fluid-saturated media) requires further investigation of the model, in particular, quasi-static modification for system of governing equations for small-amplitude propagation. Presentation describes the derivation of quasi-static equations and draw attention to challenges appearing in this process.

Summary: TBA

Summary: The report is devoted to such properties of rocks and rock-fluids as absolute permeability, relative phase permeability, fluid saturation, wettability, surface and interfacial tension, contact angle, and capillary pressure. These properties are an important aspect for understanding the structure of the reservoir; they allow you to estimate the amount of hydrocarbons in the reservoir and also make it possible to control the production process.

It's presented the methods for measuring these parameters in the laboratory, as well as the methods for numerical simulation of one-, two-, and three-phase flows in porous media.

Summary: The report will present the developed computational technologies for modeling the propagation of seismic waves in a complex environment as part of active vibroseismic monitoring of volcanic structures. Based on the finite-difference approximation of 2D and 3D equations of the dynamic theory of elasticity, parallel algorithms and a set of programs adapted to the main modern computing architectures have been developed.

The effectiveness of using various approaches to parallelization and optimization of the original algorithm has been studied. Several geological and geophysical models of the internal structure of a magmatic stratovolcano are considered, and numerical modeling of vibration transmission for these media models is carried out with reference to the location of the Elbrus volcano. In each case, the calculated seismic field inside the volcanic edifice has a complex structure, depending on the geometry of the objects under study and their rheological characteristics. Based on the conducted research, it is concluded that it is possible to monitor magmatic volcanoes using precision vibroseismic sources and seismic observation systems.

Summary: The report presents the potential of using deep learning to improve the accuracy of seismic modeling, as well as reduce computational costs. NDM-net (Numerical Dispersion Mitigation) is used to suppress numerical dispersion in seismograms. As a training sample, an exact solution is used on only a small part of the sources, after which NDM-net is applied to the entire dataset calculated on a coarse grid. A number of numerical experiments and statistical analysis were carried out to assess the sensitivity of the prediction error to different metrics. It is shown that the method of hierarchical clustering based on a combined distance matrix is applicable for the construction of a training sample.

Summary: The report presents the method published by Ugo Boscain, et al. "Highly corrupted image inpainting through hypoelliptic diffusion". 2018 that proposed an algorithm for recovering damaged images based on the use of a second-order differential operator of a special form. We discuss the possibility of extending this approach to the reconstruction of 3D seismic data.

Summary: The double root equation is a special approximation of the wave equation that describes only a singly reflected wave field and has found application in the field of seismic migration. The report builds a gradient algorithm for refining the migration velocity model based on the high-frequency asymptotics of this equation. The performance of the algorithm is demonstrated on two sets of non-noisy synthetic data in a two-dimensional formulation of the problem. The report discusses the advantages and disadvantages of the proposed approach, and outlines ways for its further development.

Summary: The talk will review a number of approaches used in the mathematical modeling of fractured media. The main properties of sub-Riemannian geometry are presented in the case of two-dimensional distributions in three-dimensional space. The similarity between the sphere in this geometry and the permeability front in one of the models of fractured media is indicated

Summary: When conducting filtration experiments on rock core samples in reservoir conditions in terms of temperature and pressure, formation water is pumped through the test sample and displaces oil (or an oil model). One of the options for calculating the volume of oil displaced by water is the use of an acoustic separator at the outlet of the sample, where the ultrasonic signal reflected from the water/oil fluid interface allows you to directly measure the height of the liquid column and thus calculate the amount of oil displaced from the sample. This presentation discusses the construction of an automatic system for measuring the amount of oil released from a sample by measuring the liquid level in the separator, where the level is measured by a high-frequency ultrasonic transducer and the first entry in the signal reflected from the interface of the media is distinguished using the Bayer algorithm

Summary: As part of the autumn school of the Euler Institute on Quantum Algorithms, one of the problems was to determine the kinetic coefficients of chemical reactions based on quantum computing. The report discusses the solution of the inverse problem by reducing it to the quantum optimization problem (PUBO/QUBO). The presented algorithm makes it possible to reduce the number of time intervals at which concentrations are determined, as well as the number of binary variables in the PUBO problem. The solution of the problem is compared with such classical methods as the method of least squares.

Summary: We present  an algorithm that implements the discrete element method for numerical simulation of large deformations including those leading to destruction of poroelastic materials. In the discrete element method, a continuous medium is represented by a set of model particles with given interaction laws. Under the action of external forces and forces of interaction of particles, they can change their position in space according to the equations of motion for a discrete set of particles, which allows effectively take into account large deformations. The pecularity of the Discrete Element Method is the absence of theoretically substantiated relationships between the model parameters at the level of interaction of individual particles and the macroscopic mechanical parameters of the medium. The paper presents a systematic numerical study of the influence of model parameters on the effective Young's modulus and breaking point of the simulated elastic body.

Summary: When a seismic wave propagates throught fractured porous fluid-saturated media, it induces fluid flow as between fractures and background rock, as between intersecting fractures. This fluid flow impact on seismic characteristics (in particular, frequency-dependent wave attenuation) can be used to estimate fractured hydrocarbon reservoir transport properties and fluid mobility. Transport properties of fractured reservoir are mostly defined by long connected fractures chains, and this global fracture connectivity influence on wave attenuation should be estimated. For this purpose the algorithm of discrete fracture network generation with defined percolation length is developed and implemented using simulated annealing method. Used objective function includes the probability of existence of continuous path within connected fractures on given distance throughout fractured medium. Also finite-difference approximation of Biot dynamic equations on staggered grid is used to develop and implement the algorithm of numerical estimation of seismic attenuation in anisotropic fractured porous fluid-saturated media. Results of numerical experiments of plane P-wave propagation using developed algorithms demonstrate the influence of global fracture connectivity, microscale anisotropy and fracture-filling material physical properties on frequency-dependent seismic wave attenuation.

Summary: High-resolution seismic imaging is one of the key technologies for a quantitative description of the Earth interior, improving our knowledge regarding tectonic evolution, territory and energy management of our modern society with mineral and fossil resources, and more or less permanent deposits of unwanted wastes in giant volumes of the upper crust. Such remote-sensing investigation is based on seismic records which are oscillatory signals. Seismic waves interact in a rather complex way with materials during its travel inside the Earth with two endmember wave-matter probings: transmission interaction and reflection interaction which will be described.

Extracting information from these coherent time series requires careful space-time data analysis and accurate wave propagation simulation in heterogeneous models. Exploiting either exact wavefield with the needed calibration for the true ground-motion amplitude, or full waveform with less requirement on network deployment, or phase measurement as a more robust observable are different investigations performed in seismology and in seismic. High-resolution imaging requires massive data and powerful computers and skills for using them in a pertinent way. 

Where should we put our efforts for improving such quantitative imaging and what are limits we may expect from wave physics in the translucent Earth we live on will be discussed during this presentation?

The talk is in Russian