The article is devoted to the application of modern methods of generative image compression using variational autoencoders and neural network architectures. Special attention is paid to the analysis of existing approaches to image generation and restoration, as well as a comparative assessment of compression quality in terms of visual perception and metric indicators. The aim of the study is to systematize deep image compression methods and identify the most effective solutions based on the variational Bayesian approach. The paper considers various architectures, including conditional autoencoders and hypernetwork models, as well as methods for evaluating the quality of the data obtained. The main research methods used were the analysis of scientific literature, a comparative experiment on the architectures of generative models and a computational estimation of compression based on metrics. The results of the study showed that the use of variational autoencoders in combination with recurrent and convolutional layers makes it possible to achieve high-quality image recovery with a significant reduction in data volume. The conclusion is made about the prospects of using conditional variational autoencoders in image compression tasks, especially in the presence of additional information (for example, metadata). The presented approaches can be useful for developing efficient systems for storing and transmitting visual data.

Keywords: variational autoencoders, generative models, image compression, deep learning, neural network architectures, data recovery, conditional models

This article introduces the FCA (Filtration Capacitive Analysis) software suite, designed for interpreting well test data from complex well configurations, including horizontal wells with multistage hydraulic fracturing (MSHF) in low-permeability reservoirs. The system employs a modular client-server architecture, ensuring scalability and reliability. The implemented numerical methods address key challenges in well test interpretation, such as data preprocessing with advanced filtering techniques (e.g., Kalman filter, Savitzky-Golay filter, and convolutional autoencoders) and solving the inverse problem using filtration-capacitance invariants and the superposition principle. Practical application results demonstrate FCA's effectiveness in estimating formation flow capacity properties and well parameters. Validation against industry-standard simulator KAPPA Saphir confirmed FCA's higher sensitivity to model correctness, reducing interpretation ambiguity and ensuring reliable industrial deployment.

Keywords: well test interpretation, low-permeability reservoir, filtration-capacitance analysis, inverse problem, software architecture, client-server, convolutional autoencoder, superposition principle, misfit minimization, multistage hydraulic fracturing

The article focuses on developing data clustering algorithms using asymmetric similarity measures, which are relevant in tasks involving directed interactions. Two algorithms are proposed: stepwise cluster formation and a modified version with iterative center refinement. Experiments were conducted, including a comparison with the k-medoids method. The results showed that the fixed-center algorithm is efficient for small datasets, while the center-recalculation algorithm provides more accurate clustering. The choice of algorithm depends on the requirements for speed and quality.

Keywords: clustering, asymmetric similarity measures, clustering algorithms, iterative refinement, k-medoids, directed interactions, adaptive methods

This paper considers the modeling of user work with a multi-server database developed on the basis of microservice architecture. The subject area was analyzed, the main entities of the system were described, and the mechanisms of data transfer and service interaction using Docker and Apache Kafka were implemented. It was revealed that the development of a multi-server database allowed to achieve high scalability and fault tolerance of the system. The implementation of replication and sharding mechanisms provided even load distribution, and the use of Kafka message broker facilitated efficient data exchange between services. The testing confirmed the system's reliability under high load, as well as revealed its strengths and potential improvements.

Keywords: modeling, load balancing, Docker, Apache Kafka, microservice architecture, distributed systems, query optimization

The article is devoted to the method of calculating the effective frequency of the load impulse FWD, which allows establishing a correspondence between the laboratory-determined viscoelastic characteristics of asphalt concrete layers included in the road pavement and the modulus of elasticity calculated by backcalculation based on the deflection  bowl determined in field tests using FWD. The article proposes an algorithm for finding this frequency, presents the results of its calculations, and compares the elastic moduli obtained in the laboratory and by the results of the backcalculation. The conducted numerical experiment confirms that for the deflection cups generated within the viscoelastic calculation model, the difference in modules does not exceed 8%. At the end of the work, ways are proposed for the practical application of the calculated parameter and further improvement of the method on real deflection  bowl obtained on  road pavements.

Keywords: dynamic loading system, road surface, deflection bowl, asphalt concrete, elastic modulus, relaxation modulus, AMPT test, master curve, backcalculation, effective pulse frequency

The article presents numerical modeling of flow dividers (separators) with different hole diameters (3.5 mm, 7.0 mm, 14.0 mm) to prevent cavitation damage. The hole diameters, number, and rows in the separators have equivalent significance, as they determine the distribution of local velocities and pressures in the flow. This minimizes the risk of vapor bubble formation and subsequent collapse, which can lead to erosion of metal surfaces. For clarity, the simulation results are presented in the form of pictures of the distribution of pressure and velocities in each of the separators with different diameters. In order to prevent cavitation, the authors have presented a design of a "short-stroke" valve in which it is allowed to use flow dividers with enlarged holes.

Keywords: valve, cavitation, distribution pattern, separator, holes, rotation, simulation, flow divider

An ensemble of models for predicting the position of a mobile robot moving in an unstructured environment is presented. An architecture has been developed that integrates a kinematic motion model with trainable models utilizing elevation map data and semantic segmentation. The principles for constructing a spatial feature map are described, incorporating geometric characteristics such as the terrain roughness index and a fuzzy traversability index. A modular structure of the following blocks is proposed: data preprocessing, geometric property computation, segmentation, and decision-making. Test results demonstrate the advantage of combining kinematic and sensor-based models for autonomous navigation in complex environments.

Keywords: traversability model, elevation map, point cloud, kinematic model, segmentation, machine learning, feature map

The article is devoted to assessing the influence of possible inhomogeneities in the layers of road pavements on the results of backcalculation of elastic moduli based on testing data of falling weight deflectometres (FWD). The article discusses the influence of different locations of theoretically specified inhomogeneities of structural layers within the roadway. Additionally, the influence of the location of the edge of the pavement on the results of calculating elastic moduli by backcalculation is considered. The conducted numerical experiment confirms that possible inhomogeneities of the road pavement can significantly influence the results of reverse calculations, and as a consequence, the decision-making on the appointment of repair measures. The boundaries are also determined at which the presence of a shoulder with a pavement design that differs from the roadway significantly distorts the results obtained. At the end of the work, ways to practically take into account inhomogeneities and further improve the method of inverse calculation of the elastic moduli of non-rigid road pavements are proposed.

Keywords: dynamic loading system, road pavement,structural layers, deflection bowl, asphalt concrete, elastic modulus, backcalculation, heterogeneity of layers, roadway, roadside

The calculation of wind loads on curtain facades and their fastening elements for high-rise buildings and structures using engineering methods and various numerical techniques remains an important task to this day. The corner sections of the building, where the greatest negative wind pressure occurs, are of particular interest. Incorrect calculation of wind suction can lead to the separation of panels during strong winds. The article calculates the peak wind load using a numerical method for a rectangular building with an aspect ratio of 0.6. Numerical calculations of the two-dimensional flow around the building profile in the ANSYS Fluent program using the k-e Realizable turbulence model were used to obtain the coefficients of drag, lateral force, and the distribution of the pressure coefficient at maximum lateral force. The calculations showed that the wind suction at the edge of the building exceeds the standard value by approximately 30%. The results obtained in the article should be taken into account when designing the facade.

Keywords: peak wind load, wind suction, rectangular buildings, peak negative aerodynamic pressure coefficient

The article discusses the problems of generating and updating software documentation using large language models. An overview of existing approaches is presented, including code summarization, systems using augmented generation approaches, assistants embedded in the development environment, and their limitations in terms of loss of architectural context and the occurrence of structural hallucinations. The concept of a graphically augmented documentation system is proposed, where the "source of truth" is a directed graph of knowledge about the code, built by static code analysis and analysis of library dependencies. An algorithm for constructing a graph is described, including node extraction, library bytecode analysis, and semantic link classification. The effectiveness of the approach was confirmed by experimental implementation on an industrial microservice, where the system demonstrated the ability to correctly restore the context and generate meaningful documentation without distorting the facts.

Keywords: automatic documentation, large language models, knowledge graph, augmented text generation, static analysis, semantic search, vector representation, microservice architecture, program structure interface, bytecode, technical documentation

The article examines a functional-dynamic model of implementing intelligent digital platforms and solutions, whose governing role in the development of a macroeconomic system is taken into account using a feedback mechanism. The relevance of the study is demonstrated in the context of active digital transformation of industries. The mathematical form of the model under consideration is a system of nonlinear differential equations of an evolutionary type, similar to dynamic models of the development of biological communities. An analysis of a macrosystem influenced by innovative technologies is carried out. As such a system, a two-sector macrostructure is considered, simulating the impact through the implementation and use of intelligent digital platforms (IDP) of two related industries, which are the transport and logistics and manufacturing sectors. The objective of the work is to study the stable states of such a structure. The model allows for taking into account the influence of investments in IDPs based on the principle of their proportionality to the growth rates of return on assets in these industries. In the work, quantitative estimates of the parameters of the original model are adjusted. An analysis of the macrosystem is carried out under conditions of different development rates of the interacting industries. The stability of the system according to Lyapunov is studied. An asymptotic approximation ‒ a solution to the problem ‒ was constructed using A.B. Vasil’eva's boundary layer decomposition method. The results describe the process of self-organization in a stable model of interaction between two related industries, supported by integrated digital platforms.

Keywords: functional-dynamic model, intelligent digital platforms, two-sector macrostructure, transport and logistics industry, production, sustainability, inter-industry interaction, asymptotic analysis, boundary layer function method

A carbon fiber reinforced steel beam was tested, which was subjected to cyclic loads of varying intensity during bending during glue hardening and periodically tested in static mode to determine the increase in stiffness. Tests have shown that adhesion occurs at higher loads, the adhesion strength decreases, and when the shear stress in the adhesive layer is exceeded, adhesion does not occur. The flexibility of the adhesive layer also reduces the cross-section characteristics, but by no more than 7%. Lap shear tests performed on samples cut from reinforced beams confirmed the results of bending tests, showing that the greatest decrease in adhesion strength occurs at the ends of the beams, where the sliding and shear stresses are greatest.

Keywords: carbon fiber reinforcement, cyclic load, adhesive joint strength, lap shear test

Verification and debugging of algorithms and related software that implements the search for such dangerous phenomena for aircraft flights as wind gusts and turbulence areas can be implemented in radar signal simulators, including using the concept of accessing databases storing test wind fields by coordinate components. The practical value of this approach is to minimize the number of expensive flight tests in difficult weather conditions. After implementing database data interpolation, continuous fields can be obtained, including predicted radar parameters, the processing of which, depending on the changing parameters of the locator: beam width, probe pulse duration, leads to different estimates, including the measured parameters of the movement of air masses. This article describes an approach to simulation modeling that makes it possible, by generating radio signals, the primary source of which are continuous interpolated functions of air mass motion parameters, to obtain either averaged radial velocity values in resolution elements or its standard deviation.  As a result, it allows us to test signal processing algorithms for detecting wind shifts or turbulence in weather navigation radars. The results of verification of the procedure for processing radio signals generated using the proposed approach are presented, confirming the correctness of the formation and detection of simulated fields of turbulent regions.

Keywords: on-board radar, meteorological navigation, simulation, algorithms, parameter estimation

For accurate modeling of the stress-strain state of the lithospheric plate, it is necessary to correctly define boundary conditions that reflect the interaction with the geological environment. The Dirichlet edge problem, in this context, involves setting displacements at the boundary of the calculated region. The problem is that the true displacements at the craton boundary are generally unknown and can change over time under the influence of tectonic processes and load changes.

Keywords: boundary conditions, stress-strain state, mathematical modeling, model, lithospheric plate, finite element method, geotectonics, stretching, compression, computer modeling, asthenosphere

The article is devoted to the study of the problem of estimating unknown parameters of linear regression models using the least absolute deviations method. Two well-known approaches to identifying regression models are considered: the first is based on solving a linear programming problem; the second, known as the iterative least-squares method, allows one to obtain an approximate solution to the problem. To test this method, a special program was developed using the Gretl software package. A dataset of house prices and factors influencing them, consisting of 20640 observations, was used for computational experiments. The best results were obtained using the quantreg function built into Gretl, which implements the Frisch-Newton algorithm; the second result was obtained using an iterative method; and the third result was achieved by solving a linear program using the LPSolve software package.

Keywords: regression analysis, least absolute deviations method, linear programming, iterative least squares method, variational weighted quadratic approximation method

With the rapid growth of information on the internet, the accumulation of large databases, and the constant influx of data from various sensors and intelligent systems, it is becoming increasingly difficult for users to find what they are really looking for. Therefore, the development of automatic summarization methods is considered a crucial task in natural language processing. These needs have motivated the development of various methods and approaches for extracting semantic and semantic information from documents, classifying it, and systematizing it. This article develops the architecture of a hybrid-syntactic fuzzy system for extracting semantic features from text and presents its mathematical formalization. The author's method enables a transition from empirical assessments of word importance to a rigorous formalized calculation of their semantic weight.

Keywords: semantics, sentence, extraction, fuzzy logic, comparison, data

An oil field is a complex system the operability of which depends on the power supply reliability. The main disturbance in the electrical system is voltage sag which causes transient processes which can lead to a halt in oil production.
The article discusses the transients modeling in the oil production well electrical system, consisting of a transformer, a cable line and a submersible induction motor. The mathematical model has been compiled for calculating transient processes in such systems while each element is described as a separate module containing algebraic and differential equations which allows modeling dynamic and steady-state modes of operation. Dependences of longitudinal and transverse components of stator current and rotor speed of submersible induction motor at start-up, voltage sag and power supply disconnection are obtained.

Keywords: transient processes, electrical system, oil production, submersible induction motor, voltage sag, mathematical modeling

To evaluate the noise suppression efficiency of railway transport screens, the COMSOL Multiphysics software package (module "Pressure Acoustics, Frequency Domain") was used in this article. Using the finite element method, calculations were carried out and the sound insulation characteristics of noise protection screens were assessed, and a comparative analysis of the characteristics of various screen models was also carried out. Based on the effects of acoustic interference, an interference-type noise suppression device is proposed, in which filling the upper part with a porous material improves the diffraction sound absorption of the noise protection screen, thereby enhancing its acoustic characteristics.

Keywords: COMSOL software, finite element method, transport, noise barriers, acoustic interference, acoustic characteristics

The use of machine learning when working with text documents significantly increases the efficiency of work and expands the range of tasks to be solved. The paper provides an analysis of the main methods of presenting data in a digital format and machine learning algorithms, and a conclusion is made about the optimal solution for generative and discriminative tasks.

Keywords: machine learning, natural language processing, transformer architecture models, gradient boosting, large language models

The development of digital learning platforms, electronic document management systems, and web-based systems that process natural language text information has led to an increase in the volume of content and/or arrays of processed full-text documents. This, in turn, has increased the demand for highly effective natural language processing methods capable of capturing text semantics. This article proposes a hybrid architecture based on the integration of probabilistic and fuzzy logic that effectively addresses semantic ambiguity issues by integrating stochastic and fuzzy logic channels that take into account both statistical patterns and linguistic uncertainty.

Keywords: integration, semantics, interpretation, natural language, uncertainty, patterns

A stochastic model of the optimal functional extrapolator of a fractal L–Markov process with a quasi-rational spectrum is constructed. When developing the model, methods of spectral and fractal analysis of random processes, the theory of functions of a complex variable, methods for calculating stochastic integrals, and the theory of stochastic differential–difference equations connecting processes with a quasi-rational spectrum with processes with a rational spectrum were used.; as well as an original technique for constructing spectral characteristics of extrapolation, developed by the famous mathematician A. Yaglom. Using the Levinson–McKean theorem, it is established that the random processes studied in this paper are L–Markovian in nature. The fulfillment of the conditions of Mandelbrot's theorem on the shape of the spectral density of fractal random processes, as well as the values of the Hearst exponents and the fractality index, suggest that the random process under study is fractal and, moreover, persistent. It is proved that the optimal extrapolator constructed over the entire past of the process can be represented as the sum of a linear combination of the values of the process itself at three time points in the case of 0 < τ < 1 and at two time points in the case of 1 < τ < 2 and an integral with an exponentially decaying weight function extended to (– ∞; ∞). In the first case, the L – boundary of the L–Markov process under study consists of three points L = {t; t – 2; t + τ – 2}, and in the second case it consists of two points L = {t; t + τ – 2}, where τ is the lead time.

Keywords: extrapolation, L –Markov process, fractality, trend tolerance, spectral characteristic, optimal extrapolator

The paper presents a technique for integrating three-dimensional solid-state models developed in the CAD SolidWorks environment into the MATLAB/Simulink dynamic modeling environment. A key element of the research is the use of the Simscape Multibody module to transform geometric data and kinematic relationships into a multi-mass dynamic model. The technique has been tested using the example of creating a virtual prototype of a steering wheel platform.

Keywords: SolidWorks, MATLAB/Simulink, Simscape Multibody, Virtual prototype, steering kinematics, end-to-end design, multi-mass dynamic modeldiversification, production and technical goals to ensure production flexibility

The article is devoted to the study and application of thin films of iron-yttrium garnet (YIG) in microwave devices (tunable generators and filters) on magnetostatic waves. The properties of films, their physical principles of propagation, advantages over YIG-sphere, the problem of arising temperature frequency shifting (TDS) and its solutions are considered. The concept of spin-wave propagation is given. The operation principles of devices based on magnetostatic waves, their prospects and competitive advantages are described. There was performed mathematical modeling of cubic and uniaxial anisotropy of the YIG-film to reduce the TDS. All of that were confirmed by special tests .

Keywords: Yttrium-iron garnet, generators, phase shifters, delay line, tunable filters, anisotropy, spin-waves, liquid phase epitaxy.

This article presents a comprehensive analysis of Russian-language texts utilizing neural network models based on the Bidirectional Encoder Representations from Transformers (BERT) architecture. The study employs specialized models for the Russian language: RuBERT-tiny, RuBERT-tiny2, and RuBERT-base-cased. The proposed methodology encompasses morphological, syntactic, and semantic levels of analysis, integrating lemmatization, part-of-speech tagging, morphological feature identification, syntactic dependency parsing, semantic role labeling, and relation extraction. The application of BERT-family models achieves accuracy rates exceeding 98% for lemmatization, 97% for part-of-speech tagging and morphological feature identification, 96% for syntactic parsing, and 94% for semantic analysis. The method is suitable for tasks requiring deep text comprehension and can be optimized for processing large corpora.

Keywords: BERT, Russian-language texts, morphological analysis, syntactic analysis, semantic analysis, lemmatization, RuBERT, natural language processing, NLP

A method has been developed for synthesizing antenna arrays with element failures using a convolutional artificial neural network with two encoders. A neural-network block architecture is proposed for computing the radiation pattern from the amplitude distribution of currents over the antenna-array aperture, enabling unsupervised training of the artificial neural network. The results obtained confirm the feasibility of the developed method.

Keywords: Antenna array synthesis, antenna element failures, radiation pattern, artificial neural network, unsupervised learning.