Trudy NGTU im. R.E. Alekseeva

This article presents a neural network system for location identification from a photograph that employs a cascade of filters to recognize key features: text language, landscape type, vegetation variety and road surface characteristics. The distinctive aspect of this approach lies in combining the outputs of all filters using an original probabilistic method, which allows to significantly narrow the search area. The system successfully identifies distinctive topological features of different geographic regions, such as the red roads of Australia, the coniferous forests of Russia and Canada, or the tropical vegetation of South America. Testing on an extensive dataset of photographs confirms the high efficiency of the method, with the system correctly identifies the country or region of capture in most cases. This approach opens new possibilities for applications where metadata-free geolocation is crucial ‒ from travel services to historical research. Further development of the system involves adding new filters to achieve even more precise location identification.

The article discusses numerical integration strategies designed for solving stiff equations. These strategies are based on the use of a finite-difference formula that implements an implicit integration method. It is proved that the problem of constructing an optimal algorithm is a multi-stage problem. The implemented version of the optimal algorithm is described. Possible modifications of the algorithm, which do not require solving auxiliary problems, are proposed. Here, the optimal numerical integration algorithm refers to an algorithm that minimizes the number of computations of the right-hand sides of differential equations system, subject to constraints determined by the tolerance of computations. The results of computational experiments on the use of improved strategies compared to those previously described are presented.

The article is devoted to the validation of a fuel compact computational model of the RASNAR-GAZ computer program. This program is designed to calculate non-stationary modes in nuclear power plants with a gas coolant. Experimental data from reactor tests of spherical fuel element of a high-temperature gas-cooled reactor (HTGR) in the pulsed mode of operation of a pulsed graphite reactor are used for this purpose. It is shown that it is possible to apply the computational model of the fuel compact in the form of a homogeneous mixture with equivalent parameters of heat capacity, thermal conductivity and density to the heterogeneous structure of fuel composition particles dispersed in a graphite matrix, forming fuel compacts. The non-stationary equation of thermal conductivity of a fuel element in the model used was described in cylindrical geometry. The validation showed that the RASNAR-GAZ program describes temperature changes both in the center and outside of the fuel compact with sufficient accuracy for engineering calculations, and the deviations do not exceed the values specified in the certification passport.

The article is devoted to experimental studies of the influence of geometric parameters of mixing blades of TVS-Kvadrat mixing spacer grids on hydraulic resistance. The research methodology is based on modeling the flow of water coolant in the reactor core in the region of self-similar flow based on the theory of hydrodynamic similarity. The influence of the blade inclination angle in the range from 15° to 40° with a constant area of the flow cross-sectional area, as well as the influence of area of the flow section obstruction by the blades in the range from 6.8 to 10.6 mm² with a constant inclination angle of 25° on the hydraulic resistance coefficient of mixing spacer grids were studied. All studies were conducted for a corridor tube bundle with a relative pitch S/d=1.3 at Reynolds criterion values from 10.000 to 100.000. As a result of the studies, integral dependencies were obtained that characterize the influence of the geometric parameters of the blades on the hydraulic resistance. The obtained dependencies will allow predicting the hydraulic resistance, as well as making substantiated choices of the geometric parameters of mixing blades when designing and creating new designs of mixing devices. The set of experimental data served as the basis for creating a database that can be used to validate calculation codes and programs for complex mathematical modeling using modern computers.

The article is devoted to increasing the topology optimization efficiency based on the finite element method in the design of vehicle parts by reducing the number of load cases. To achieve this, an analysis of the influence of various methods of automated load selection was carried out. As an example, the design of a steering knuckle for an all-wheel drive unmanned transport and technological vehicle with a gross weight of 800 kg is presented. The results of the verification calculations confirm that it is sufficient to use only a limited number of load cases, which are decisive, to obtain the optimal power circuit. The proposed method allows to speed up the topological optimization process more than twice when minimizing compliance, in addition, halving the number of load cases with stress constraints reduces computational time by 2.6 times. The obtained results can be used in the design of highly loaded parts of minimum mass for vehicles.

The paper proposes a number of algorithms for the control system of an unmanned vehicle with adaptive cruise control and automatic emergency braking functions. This system allows to maintain a distance from the vehicle in front, make an emergency stop in front of an unexpected obstacle and maintain a set speed. The described adaptive cruise control algorithm is aimed at improving the road safety. The advantage of the automated control system over an operator is the elimination of the human factor, as well as high reaction speed and accuracy of actions. A mathematical model of the motion of an unmanned vehicle is presented. The model is developed to test and adjust the proposed control system implementing the functions of adaptive cruise control. The system can generate the control action necessary for the safe movement of the vehicle, eliminating the possibility of a collision using data on the speed of the obstacle and the distance to the obstacle. Computer simulation shows the system's performance under basic scenarios. Data on the experimental study of equipment designed to test the proposed algorithms in practice, as well as the results of development a fully functional control system for the movement of an unmanned vehicle are provided.