The work is devoted to the development and analysis of computer vision algorithms designed to recognize objects in conditions of limited visibility, such as fog, rain or poor lighting. In the context of modern requirements for safety and automation, the task of identifying objects becomes especially relevant. The theoretical foundations of computer vision methods and their application in difficult conditions are considered. An analysis of image processing algorithms is carried out, including machine learning and deep learning methods that are adapted to work in conditions of poor visibility. The results of experiments demonstrating the effectiveness of the proposed approaches are presented, as well as a comparison with existing recognition systems. The results of the study can be useful in the development of autonomous vehicles and video surveillance systems.

Keywords: computer vision, mathematical modeling, software package, machine learning methods, autonomous transport systems

In systems for monitoring, diagnostics and recognition of the state of various types of objects, an important aspect is the reduction of the volume of measured signal data for its transmission or accumulation in information bases with the ability to restore it without significant distortion. A special type of signals in this case are packet signals, which represent sets of harmonics with multiple frequencies and are truly periodic with a clearly distinguishable period. Signals of this type are typical for mechanical, electromechanical systems with rotating elements: reducers, gearboxes, electric motors, internal combustion engines, etc. The article considers a number of models for reducing these signals and cases of priority application of each of them. In particular, the following are highlighted: the discrete Fourier transform model with a modified formula for restoring a continuous signal, the proposed model based on decomposition by bordering functions and the discrete cosine transform model. The first two models ideally provide absolute accuracy of signal restoration after reduction, the last one refers to reduction models with information loss. The main criteria for evaluating the models are: computational complexity of the implemented transformations, the degree of implemented signal reduction, and the error in restoring the signal from the reduced data. It was found that in the case of application to packet signals, each of the listed models can be used, the choice being determined by the priority indicators of the reduction assessment. The application of the considered reduction models is possible in information and measuring systems for monitoring the state, diagnostics, and control of the above-mentioned objects.

Keywords: reduction model, measured packet signal, discrete cosine transform, decomposition into bordering functions, reduction quality assessment, information-measuring system

The current situation in the practice of designing complex technical systems with metrological support is characterized by the following important features: a) the initial information that can actually be collected and prepared at the early stages of design for solving probabilistic problems turns out, as a rule, to be incomplete, inaccurate and, to a high degree, uncertain; b) the form of specifying the initial information (in the form of constraints) in problems can be very diverse: average and dispersion characteristics or functions of them, measurement errors or functions of them, characteristics specified by a probability measure, etc. These circumstances necessitate the formulation and study of new mathematical problems of characterizing distribution laws and developing methods and algorithms for solving them, taking into account the constraints on the value and nature of change of the determining parameter (random variable) of a complex technical system. As a generalized integral characteristic of the determining parameter, the law of its distribution is chosen, which, as is commonly believed, fully characterizes the random variable under study. The purpose of this work is to develop a method that allows constructing distribution laws of the determining parameter of a complex technical system using the minimum amount of available information based on the application of Chebyshev inequalities. A method for characterizing the distribution law by the property of maximum entropy is presented, designed to model the determining parameter of complex technical systems with metrological support. Unlike the classical characterization method, the proposed method is based on the use of Chebyshev inequalities instead of restrictions on statistical moments. An algorithm for constructing the distribution function of the determining parameter is described. A comparison is given of the results of constructing distribution laws using the developed method and using the classical variational calculus.

Keywords: Chebyshev inequalities, complex technical system, design, determining parameter, characterization of distribution law

In operational diagnostics and recognition of states of complex technical systems, an important task is to identify small time-determined changes in complex measured diagnostic signals of the controlled object. For these purposes, the signal is transformed into a small-sized image in the diagnostic feature space, moving along trajectories of different shapes, depending on the nature and magnitude of the changes. It is important to identify stable and deterministic patterns of changes in these complex-shaped diagnostic signals. Identification of such patterns largely depends on the principles of constructing a small-sized feature space. In the article, the space of decomposition coefficients of the measured signal in the adaptive orthonormal basis of canonical transformations is considered as such a space. In this case, the basis is constructed based on a representative sample of realizations of the controlled signal for various states of the system using the proposed algorithm. The identified shapes of the trajectories of the images correspond to specific types of deterministic changes in the signal. Analytical functional dependencies were discovered linking a specific type of signal change with the shape of the trajectory of the image in the feature space. The proposed approach, when used, simplifies modeling, operational diagnostics and condition monitoring during the implementation of, for example, low-frequency diagnostics and defectoscopy of structures, vibration diagnostics, monitoring of the stress state of an object by analyzing the time characteristics of response functions to impact.

Keywords: modeling, functional dependencies, state recognition, diagnostic image, image movement trajectories, small changes in diagnostic signals, canonical decomposition basis, analytical description of image trajectory

A combined theoretical and practical study of the burner device parameters has been performed. The flow characteristic of the fuel supply system has been determined. Aerodynamic studies of the burner device characteristics have been conducted, axial velocity fields have been constructed, and critical parameters of the air supply unit design have been identified. The temperatures of in-chamber processes have been experimentally determined. A mathematical model of chemical reactions of the torch has been developed, and the dependence of diesel fuel toxicity on the excess air coefficient has been constructed. The effect of water vapor on the burner device operation has been determined.

Keywords: burner device, axial velocity field, intra-chamber processes, thermochemical parameters, mathematical modeling, toxicity

The study is devoted to improving the methodology of calculation of structures for the impact of explosive loads, including aerial shock waves in UAV attacks. Modern modeling methods are analyzed, a calculation algorithm using non-linear dynamic approaches is proposed. It is shown that the use of spherical discrete elements allows a more accurate assessment of debris destruction and formation. The calculations presented confirm the effectiveness of the proposed approach.

Keywords: Explosive loads, air shock wave, unmanned aerial vehicles, dynamic calculation, destruction of structures, simulation of explosive effects, application of spherical discrete elements, calculation algorithm, formation of debris, non-linear methods

This article discusses numerical methods used to estimate the parameters of a family of Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models, which are widely used for analyzing and predicting financial time series with variable variance. The paper provides a comparative analysis of numerical methods for estimating GARCH effects, which are based on the gradient descent method of adaptive algorithms, various variations of quadratic methods based on the Newton method, as well as alternative methods based on the simplex method, linear and quadratic interpolation. The analysis is carried out on the basis of synthetic data and on real data on quotations of the Moscow Exchange stock index using the Python 3 programming language and libraries scipy, numpy, matplotlib and others. The results of the study show that the specifics of the financial time series problem are sensitive to the choice of numerical methods for solving the optimization problem of maximizing the likelihood function. Numerical experiment has shown that using the Nelder-Meade method to evaluate GARCH effects gives the best results for solving the problem of maximizing the likelihood function.

Keywords: mathematical modeling, numerical methods, maximum likelihood method, gradient descent, Newton's method, mathematical modeling, GARCH, time series, stock market, news flows

A finite element model of the deformation zone during cold drawing on a movable mandrel has been developed and justified. This makes it possible to determine the state of the metal, calculate its damage and the shape of the die channel, while the configuration of the first transition is taken as the initial one to obtain the second transition.

Keywords: Simulation of the process of long-mandrel drawing of profile pipes

The article solves the problem of automated generation of user roles using machine learning methods. To solve the problem, cluster data analysis methods implemented in Python in the Google Colab development environment are used. Based on the results obtained, a method for generating user roles was developed and tested, which allows reducing the time for generating a role-based access control model.

Keywords: machine learning, role-based access control model, clustering, k-means method, hierarchical clustering, DBSCAN method

This study analyzes classical machine learning methods applied to the prediction of corporate bond yield spreads. Both linear methods, such as Principal Component Analysis and Partial Least Squares, and nonlinear methods, such as copula regression and adaptive regression splines, are examined. The study also explores the potential application of Random Forest models and classical neural networks. It includes a description of the data used for forecasting and presents some results of the empirical analysis. The findings have the potential to significantly impact practitioners and the scientific community striving to improve forecasting accuracy and optimize investment strategies.

Keywords: Machine Learning, Financial Engineering, Stock Market Modeling, Bond Market

In order to optimize the operation of dust-settling chambers of steelmaking furnace emission purification systems and increase the overall efficiency of the cleaning system, the movement of gas-air flows and dust particles of different diameters inside dust-collecting chambers was studied using the SolidWorks software product with the FlowSimulation application, which allowed us to investigate the influence of a number of factors, for example, fractional composition, the condition of the working surfaces of chambers, on the movement of gas-air the flow.

Keywords: steelmaking furnace, gas-air flow, dust-settling chamber, cleaning efficiency, dust, dispersed composition, modeling

The work includes an analysis of the mathematical apparatus determining the influence of parasitic parameters of the capacitor, the topology of the printed circuit board on the effective range of the decoupling capacitor. A mathematical apparatus is presented that determines the shift in the resonant frequency of the connected decoupling capacitor, taking into account the parasitic parameters of the topology.

Keywords: power distribution system, decoupling capacitor, self-resonance frequency, anti-resonance frequency, effective range, parasitic parameters, topology

This article is devoted to the development of a collision detection technique using a polygonal mesh and neural networks. Collisions are an important aspect of realistically simulating physical interactions. Traditional collision detection methods have certain limitations related to computational accuracy and computational complexity. A new approach based on the use of neural networks for collision detection with polygonal meshes is proposed. Neural networks have shown excellent results in various computer vision and image processing tasks, and in this context they can be effectively applied to polygon pattern analysis and collision detection. The main idea of ​​the technique is to train a neural network on a large data set containing information about the geometry of objects and their movement for automatic collision detection. To train the network, it is necessary to create a special module responsible for storing and preparing the dataset. This module will provide collection, structuring and storage of data about polygonal models, their movements and collisions. The work includes the development and testing of a neural network training algorithm on the created dataset, as well as assessing the quality of network predictions in a controlled environment with various collision conditions.

Keywords: modeling, collision detection techniques using polygonal meshes and neural networks, dataset, assessing the quality of network predictions

The step-by-step construction of a computer model of the process of long-angle drawing of profile pipes is considered. The minimum dimensions of the blanks have been determined, the use of which ensures the necessary dimensions of the finished product. The scheme of applying a deforming force with size adjustment in the current state during step-by-step deformation is taken into account. Geometric and finite element models have been obtained that make it possible to find all the parameters of the deformation site during the drawing process.

Keywords: dimensions of the workpiece, profile pipe, boundary conditions, load application, physical model, finite element grid

The paper considers the experience of using neural networks in construction. The widespread coverage of AI success in various areas of construction has led to an increase in business and public interest in the successful implementation of AI in various construction areas. Examples of the use of neural networks in the construction expertise of the working documentation "Metal structures" and "Metal detailed structures" are given. The process of solving the assigned tasks by an expert builder in comparison with the answers received by the neural network is described. A comparative analysis of the quality of the results obtained by the expert builder and artificial intelligence is given. As part of this study, the main algorithms for training neural networks that are applicable to solving the problem were analyzed. Particular attention is paid to algorithms capable of efficiently handling parameter variations and new configurations not represented in the training dataset. The use of these algorithms will provide increased accuracy when scaling the solution. A neural network forecast for this area of construction expertise is given.

Keywords: neural network, construction, construction expertise, expert builder, comparative analysis, training sample, neural network forecast

The simulation model of Apache HTTP Server as a mass service system is considered, the parameters of the corresponding system and Apache HTTP Server are compared using GPSS World environment. The comparison of the simulation model with a real web server is based on the construction of a test server. using Apache JMeter application, which can be used to simulate high load on the server. Query generation and statistics collection was done by Apache JMeter application. A comparison of both reports was given, differences in characteristics were pointed out, and assumptions about the reasons for the differences were outlined. The model can be applied to establish requirements for Apache HTTP Server in order to optimise its performance.

Keywords: simulation modelling, mass service system, efficiency characteristics, test server, flow of requests, service channels, queue

At the moment, quantum key distribution (QKD) technology guarantees the highest level of data exchange security, which makes QKD networks one of the most promising areas in the field of computer security. Unfortunately, the problem of topology optimization when planning and extending QKD networks has not attracted enough attention. This paper reviews approaches that use analytical models in the topology optimization problem of quantum key distribution networks. Different methods that solve problems of network capacity and security maximization and cost minimization are reviewed, the utilized algorithms are described, and conclusions about possible further research in this area are drawn.

Keywords: quantum key distribution, mathematical modeling, network topology, analytical modeling, topology optimization

The article analyzes the impact of transformation types on the learning quality of neural network classification models, and also suggests a new approach to expanding image sets using reinforcement learning.

Keywords: neural network model, training dataset, data set expansion, image transformation, recognition accuracy, reinforcement learning, image vector

This paper investigates the effectiveness of the distance fields method for building 3D graphics in comparison with the traditional polygonal approach. The main attention is paid to the use of analytical representation of models, which allows to determine the shortest distance to the objects of the scene and provides high speed even on weak hardware. Comparative analysis is made on the possibility of wide model detailing, applicability of different lighting sources, reflection mapping and model transformation. Conclusions are drawn about the promising potential of the distance field method for 3D graphics, especially in real-time rendering systems. It is also emphasized that further research and development in this area is relevant. Within the framework of this work, a universal software implementation of the distance fields method was realized.

Keywords: computer graphics, rendering, 3D graphics, ray marching, polygonal graphics, 3D graphics development, modeling, 3D models

The article describes the methodology for developing a client-server application intended for constructing a virtual museum. The creation of the server part of the application with the functions of processing and executing requests from the client part, as well as the creation of a database and interaction with it, is discussed in detail. The client part is developed using the Angular framework and the TypeScript language; the three-dimensional implementation is based on the three.js library, which is an add-on to WebGL technology. The server part is developed on the ASP.NET Core platform in C#. The database schema is based on a Code-First approach using Entity Framework Core. Microsoft SQL Server is used as the database management system.

Keywords: client-server application, virtual tour designer, virtual museum, three.js library, framework, Angular, ASP.NET Core, Entity Framework Core, Code-First, WebGL

The work is devoted to the analysis of machine learning methods for solving problems of automatic document processing. The study considers such methods as classification, information extraction, pattern recognition and natural language processing and their application in the analysis of text data. An analysis of existing algorithms and models, including linear models, decision trees, support vector methods, and a comparison of their effectiveness depending on various conditions and parameters is carried out. Particular attention is paid to the problems that specialists face when using machine learning methods in working with documents, such as data quality, the need for pre-processing and tuning of model parameters. Prospects for further research in this area and examples of possible integration of modern machine learning methods to improve the efficiency and accuracy of automatic document processing in various industries are given.

Keywords: machine learning, automatic document processing, computational experiment, artificial intelligence, classification models, software package

Flight safety is one of the most important priorities of civil aviation. In recent years, there has been a downward trend in the number of aviation accidents, which is associated with the introduction of new technologies and training methods. One of these methods is computer simulator training (CTP).KTP is a training method in which LE specialists practice skills and procedures in a virtual environment that simulates real flight conditions. KTP allows you to increase the effectiveness of training, reduce the risk of errors and ensure that training meets modern safety requirements aviation simulators, development of simulator systems, simulators of aviation instrumentation.

Keywords: diversification of management, production diversification, financial and economic purposes of a diversification, technological purposes of ensuring flexibility of production

Models and a software package have been developed that allow for the analysis of counterparties for the probability of fulfilling government contracts. A comparative analysis of machine learning models has been conducted: logistic regression, decision forest, clustering, and neural network. A software package has been developed that allows for contract forecasting. A computational experiment has been conducted to analyze counterparties taking into account contracts that have been fulfilled or not completed by them. The best model has been established, demonstrating a forecast accuracy of 97.89% by the accuracy metric.

Keywords: mathematical modeling, cybersecurity, intelligent models, financial sector, government contracts, information infrastructure

Multilayer edge commutation in 3D integration technologies can simplify the design of microassemblies and reduce the length of edge electrical connections. However, this commutation is vulnerable to thermomechanical stresses and requires preliminary analysis of the product design. This paper shows the results of modeling various variants of multilayer edge commutation for 3D microassemblies, differing both in the dielectric material used at the edge redistribution layer and in the material for sealing the microassembly volume. It has been established that the lowest values ​​of thermomechanical stresses in commutation are characteristic of materials whose temperature coefficient of linear expansion is as close as possible to this parameter of conductors. At the same time, the use of composite dielectrics in redistribution layers leads to a more significant decrease in stresses than the use of more thermally stable unfilled polymers.

Keywords: 3D integration, packaging, thermomechanical stresses, polyimide, redistribution layer

The article is devoted to the development of an algorithm for three-dimensional terrain reconstruction based on single satellite images. The algorithm is based on the algorithmic formation of three-dimensional models based on the output data of two deep learning models to solve the problems of elevation restoration and instance segmentation, respectively. The paper also presents methods for processing large satellite images with deep learning models. The algorithm proposed in the framework of the work makes it possible to significantly reduce the requirements for input data in the problem of three-dimensional reconstruction.

Keywords: three-dimensional reconstruction, deep learning, computer vision, elevation restoration, segmentation, depth determination, contour approximation