The Trofimuk Institute of Petroleum Geology and Geophysics (IPGG SB RAS) takes its rightful place at the forefront of a new modeling paradigm. Sergey Markov PhD (phys.-math.), head of the Laboratory of mathematical modeling of multiphase processes in native and artificial multiscale heterogeneous media (IPGG SB RAS), spoke about key trends in this field, and about the work being done at the Institute, and its contribution to science and to the mining industry. Below are the answers to some questions for his expert witness.
Q: Today’s world society gives much consideration to the development of algorithms and software related to supercomputer technologies. What is the explanation for this?
A : First of all, the need to solve a broad range of high-resolution applied problems. With computing technologies developing fast, as early as in 2022 supercomputer performance reached the level of exaFLOPS, which is a quintillion (1018) floating-point operations per second. If a person completed one operation per second, it would take him/ her 32 billion years for to do what a machine can do in 1 second.
High-resolution geophysical survey applications provide a promising avenue for enhancing the accuracy and efficiency of geophysical data analysis, as well as constructing images of multiscale geological objects (hydrocarbon reservoirs) and digital core models. Mathematical methods (specifically, numerical modeling) is a powerful tool for conducting such research.
Q : Is it true that the growth in computer performance will automatically empower one to faster obtain quality results?
Not exactly. Specialists from supercomputer centers speak about the paradox consisting in the following: despite the availability of high-performance supercomputers, there are few tasks that use more than a hundred petaflops per variant. This is not at all the evidence suggesting the lack of formally set tasks (problem formulation) - they do exist. However, it is necessary to ensure that a set of criteria is adopted to check the plausibility of the problem (well-posed problem) and that the methodologies chosen are appropriate. The major challenge is associated with the problem of decomposition and loading its parts onto different groups of processor units (PU) for parallel computing. New numerical methods are needed to ensure efficiency of using a supercomputer.
Q: What is plausible methodologies?
A : Novel methods should allow taking into account both geometric and functional scalability that arise when solving geophysical problems. They include new methods of domain decomposition in the context of tasks distribution across the nodes of computing systems, as well as visualization methods. The diversity of proposed methods for discretization of numerical models, and the symbiosis of machine learning algorithms and numerical modeling indicate the active phase of solving this task.
It is worth nothing that, as such, the situation arose as early as in the early 1950s. It was a time when the foundations of applied mathematics and programming were laid, which largely contributed to the development of finite difference theory, Monte Carlo methods, and the fast Fourier transform. Today, we are witnessing a similar stage in the breakthrough technologies.
Q: To what extent are these advancements deployed at your laboratory of mathematical modeling of multiphase processes in native and artificial multiscale heterogeneous media?
A : We conduct research based on modern computational schemes of finite element methods and machine learning for numerical solutions of a wide range of geophysical applications. These include the study of physical properties of rock samples using digital core models; investigation of the impact of heat transfer on electromagnetic properties of the near-wellbore zone; identification of mineral deposits by electromagnetic surveys; wave processes in heterogeneous media.
An example of a problem solved at IPGG SB RAS for drilling mud injection into a porous oil-saturated sandstone sample
The influence of heat transfer processes on electrical properties in the near-wellbore zone is a problem also solved at IPGG SB RAS
The influence of heat transfer processes on electrical properties in the near-wellbore zone is a problem also solved at IPGG SB RAS
The specified problems are characterized not only by a high contrast of physical properties of the investigated objects, but also by space and time domain scalability. These challenges can be bypassed using nonconforming virtual finite element methods.
Q : What are their advantages?
A : These methods have great potential in terms of practical application, since they allow giving up the “classical idea” of the finite-element schemes and working instead with projector-based virtual spaces. This will allow significantly reducing the analog sizing task thus saving computational cost associated with spatial discretization. Hierarchical grid models are implemented using computing and data clustering techniques.
Q: What is the laboratory’s research outlook?
A: Our research was supported by the Russian Science Foundation. While implementing the new project, we will propose new methods and effective algorithms for solving applied problems of exploration geophysics in the high-north regions of Russia.
It is worth pointing out that modern science cannot be done in isolation. Only cross-disciplinary collaboration with involvement of industry professionals yields feasible results. Within the Institute, we interact with researchers from the laboratory of multiscale geophysics; we discuss scenarios for physical experiments with Nikita Golikov. The key points here are the problem formulation and geophysical insights inferred from the results interpretation. In this respect, the support and mentorship of Academician Mikhail I. Epov helps our laboratory immensely. We appreciate his assistance in setting new applied tasks and participation in discussions of the results obtained.
Published by IPGG Press Service
Illustrations provided by Sergey Markov