Journal archive

The article presents the results of a study on optimizing the energy consumption of Industrial Internet of Things devices providing telemetry and included in the control loops of electric power systems. The relevance of the study lies in the emerging need to increase the battery life of mobile telemetry devices of industrial and technological systems, which helps to reduce the costs of their maintenance and support in working condition. An algorithm for controlling the parameters of the processor core of mobile devices with ARM architecture processors is proposed, which ensures higher energy efficiency of such devices used in the Industrial Internet of Things of the EPS. The novelty of the obtained results is the proposed algorithm for software control of the configuration of the ARM processor core parameters of mobile telemetry devices of the EPS, ensuring its higher energy efficiency, which is achieved by using data mining methods in its structure – a bidirectional neural network of long short-term memory and a fuzzy logical inference system. The choice of this network architecture is due to its ability to identify relationships in temporal sequences of device parameters by viewing the sequence in two directions at once – from the beginning to the end and vice versa. The network operates in the energy consumption plan classification mode, the results of which are then fed to the input of the fuzzy logic inference system to predict the optimal parameters of the ARM processor, which together forms a neuro-fuzzy model for managing the energy consumption of IIoT devices. Using machine learning libraries in the Python language in the Google Colab environment, model experiments were conducted, as a result of which the classification accuracy using a bidirectional neural network exceeded 0.8, and the standard deviation was 0.058 when predicting the parameters of the ARM processor based on the fuzzy logic inference system.

The article presents an approach to intelligent management of the training of groups of specialists in organizational and technical systems based on predictive modeling of the achievability of goals and cluster analysis, which allows solving the problem of qualitative planning of the process of training groups of specialists in time-limited conditions by adaptive formation of its structure and choosing a rational amount of educational information corresponding to the capabilities of trained specialists in its processing and assimilation. For this purpose, based on cluster analysis, it is proposed to form homogeneous classes of specialists taking into account their individual cognitive characteristics in order to subsequently build reference scenarios for each class of specialists of the necessary complexity by changing the volume of training content. The reference scenario is the basis for building a scenario-information model of the process of training groups of specialists in organizational and technical systems, which is a static model of this process and contains the whole set of possible scenarios. Based on the proposed scientific and methodological apparatus, the scenario-information model is transformed into a modified soft temporal Petri net for modeling the process of training groups of specialists. The simulation results underlie management decisions in the intelligent management of the process of training groups of specialists in organizational and technical systems.

To establish regularities of polymerization processes, it is advisable to use methods of mathematical modeling. On the basis of mathematical description of the process it is possible to determine the optimal conditions of its course, providing an increase in its efficiency, as well as the quality of produced polymers. One of the problems of optimization of polymerization processes is the problem of determining the initial composition of the reaction mixture, which ensures the achievement of given quantitative or qualitative indicators of the process. The aim of the work is to develop an algorithm for determining the initial concentrations of polymer synthesis reaction. The statement of the problem of searching for optimal concentrations of components of the polymer synthesis process and a numerical algorithm for its solution are given. Since the kinetic model of the polymerization process is an infinite system of ordinary nonlinear differential equations, its solution using classical optimization methods encounters computational difficulties. Therefore, it is proposed to apply the method of artificial immune systems to calculate the optimal initial composition of the reaction mixture of the polymerization process. This method allows to overcome local extrema in multidimensional search spaces and it is easy to implement for a particular process when the number of optimized parameters is changed. The developed algorithm based on the method of artificial immune systems includes a procedure for converting an infinite system of differential equations describing the kinetics of the polymerization process to a finite form using the method of moments. The algorithm has been tested on an industrially significant polymerization process of butadiene on a neodymium-containing catalytic system. The optimum concentrations of reagents that provide the polymer with the desired properties have been calculated.

The control of complex technical objects is often hampered by the lack of information necessary to build probabilistic models of the control process. An alternative approach is based on models in which elements of fuzzy set theory are used to identify an effective control strategy from the point of view of the selected criterion. The disadvantages of fuzzy situational models include the subjective nature due to the expert data used for their construction, as well as the inability to distinguish in sufficient detail the reactions of an object to the same control actions, which is most clearly manifested in complex technical objects, the functioning of which is influenced by many situational factors. The paper proposes an approach to the construction of fuzzy situational models based on precedents and allows for the ambiguity of the transition with the same control action from one situation to another in cases where it is impossible to ensure the stability of both the uncertainty factors of the external environment and the parameters of the object itself. The method of constructing models based on the consideration of accumulated historical data regarding the operated object is described. This principle largely eliminates the subjectivity inherent in expert methods. In addition, programs implementing the method are distinguished by significantly shorter execution time, which is important for process control systems. A Python program has been developed for conducting computational experiments. The verification, which was carried out using the example of a control system for a conveyor-type roasting machine, confirmed the operability of the proposed method and model.

Correct classification of land plots by their types, for example, such as forest, agricultural, urbanized, water bodies, and others, is relevant for remote sensing of the Earth and the development of geoinformation technologies. The accuracy and reliability of the results of such categorization are of paramount importance for the efficient use of natural resources, rational land use, and environmental monitoring. The article presents an approach to solving the problem of categorizing land plots based on satellite images by applying a modified standard model of a convolutional neural network. The main attention is paid to the modification of the network architecture in order to improve the accuracy of land plot classification. The authors propose an approach to training and optimizing the network in order to solve this problem. The stages of data preparation are discussed in detail, including preprocessing satellite images, annotating them, and creating high-quality training samples. The presented approaches to network training and optimization include the use of modern regularization techniques, adaptive learning methods, and class balancing strategies, which allows efficient processing of both large amounts of data and more limited sets of specific information. To test the approach’s operability and obtain the values of quality indicators, experiments were conducted to train and test the model on various sets of satellite image data. The results of the experiment suggest that the accuracy of categorization achieved on the basis of the created model meets the requirements of the Federal Service for State Registration, Cadastre and Cartography for studying remote Territories for the suitability of land for their rational use, and the proposed method can be used to solve practical problems.

A probabilistic model of the dynamics of sales of goods on the marketplace, which takes into account the impact of delivery duration on the consumer’s decision to purchase this product, is proposed. It was assumed that the quantity of goods on the marketplace is unlimited, and consumers, all other things being equal, buy goods from sellers offering the shortest delivery time. In addition, it was assumed that each consumer purchases only one unit of the product. The cases of free and paid delivery of the product to the consumer are considered separately, as well as the general case where there are consumers choosing both free and paid delivery on the marketplace. The following economic factors were used as random variables determining the dynamics of goods sales: moments of consumers’ purchasing activity, duration of delivery and maximum possible duration of delivery for a consumer, purchasing power, price for goods, number of sellers and consumers of goods on the marketplace. It is shown that the quantity of goods sold on the marketplace by a certain moment of time is a random variable, the true value of which at a large number of consumers differs little from the expected value of this random variable. The analytical expression for time dependence of the expected value of the quantity of sold goods for the cases of free and paid delivery, as well as in the presence of both types of delivery on the marketplace is found. Numerical modeling of this dependence for all considered types of delivery has been carried out. In this case, we used partial laws of probability distributions of the above mentioned economic factors, simulating certain behavior of consumers and sellers on the marketplace. In addition, in the case of paid delivery the dependence of delivery cost on its duration found on the basis of empirical data was used in numerical modeling. We analyzed the influence of the parameters of the probability distribution law of delivery duration on such characteristics of sales volume dynamics as the maximum possible sales volume and the speed of approaching the expected value of the number of sold goods to the maximum sales volume.

One of the promising ways to reduce the dependence of domestic industry on the supply of critical goods, components and raw materials necessary for the construction and effective functioning of multi-stage production and technological chains is to intensify processes for their import substitution, including through various measures of state support. However, the critical need for a wide range of products requires the selection of the most “promising” projects for inclusion in program-target documents using a set of criteria (sometimes even non-financial). As a result, there arises an urgent scientific and practical task of developing approaches to the formation of multi-projects (a set of projects) that can qualify for state support under various programs to reduce import dependence and overcome the technological backwardness of Russian industry, based on the use of modern economic and mathematical methods. In the application to this task, a multi-project can be represented as a “set” (unrelated projects), a “chain” (rigid sequence of projects) or a “network” (projects with complex logical-temporal interrelations). The specifics of each type determine the conditions and impose restrictions on the processes of selecting components for inclusion in their composition, which consists of finding the best combination of projects and/or programs, i. e., it is reduced to a task of conditional multidimensional optimization. In the absence of a requirement to find a “strictly optimal” composition, one can use metaheuristic methods that are capable of finding solutions close to these in an “acceptable” time. Among them, the largest and most well-known class are swarm intelligence algorithms based on the principles of collective behavior of a population of living organisms. To form the composition of multi-projects, the article proposes to use algorithms inspired by the collective behavior of a pack of wolves (Grey Wolf Optimizer) and a school of fish (Fish School Search) to satisfy their food needs. To increase the efficiency of their use for solving the task of finding the best composition of a “set” and “network” of projects, their hybridization with fuzzy logic methods (in particular, fuzzy clustering and fuzzy-logical inference) was proposed.

The article presents the results of implementing an integrated approach to quality control of linear electric motor inductors at the production stage. The approach is based on the use of mathematical modeling methods and computational experiment. To assess the quality of inductors, it is proposed to use two types of indicators: differential and integral. Differential indicators are determined by conducting a series of experiments with each tooth division of the inductor. Integral indicators allow us to assess the quality of the inductor magnetic circuit as a whole, for which criterion relationships are used. To test the effectiveness of the proposed approach and the reliability of the obtained indicators, a simulation mathematical model of an electric motor inductor with detailing down to one tooth division was developed. Based on this model, an algorithm and a computer program were created that implement the proposed method of quality control during technical testing. The program is written in the Maple software product’s own interpreted programming language and requires the user to enter the parameters of the inductor and power source. As a result of the program, the user receives a conclusion on the quality of the inductor magnetic circuit, as well as numerical values of the differential and integral quality indicators. The materials of the prototype inductor with a high-quality and low-quality magnetic core were used to test the quality criteria using mathematical modeling and computational experiment methods. The results of the experiment confirmed the feasibility of implementing the proposed approach and showed that it adequately reflects the technical quality of inductor magnetic cores. The software product created as a result of the study is intended for technical control of inductor magnetic cores during their production and information support for measures to improve their quality.

Intelligent process control systems allow analyzing incoming information about the control object and the external environment at a qualitatively new level and, on this basis, increasing the efficiency of the entire production. The article presents the results of a study aimed at developing a software model of an intelligent control system for complex processes of small-scale ore processing. The complexity of the processes lies in the interrelationship of many variables describing the technological units involved in processing. An analytical description of such interrelations allows describing processes with high accuracy, but leads to complex calculations that are difficult to apply in real conditions. The use of new computational algorithms from the group of intelligent methods made it possible to overcome this contradiction, ensuring, on the one hand, good accuracy of solutions, and on the other hand, making it possible to automate the adjustment of the control system to changing input effects and external conditions. The basis of the proposed software model is a deep neural network of long short-term memory, solving the regression problem when analyzing input data and calculating control effects. The novelty of the research results is the structure of the software model of the intelligent control system, including a neural network as a regulator and a hierarchical system of fuzzy inference for a generalized assessment of the control qualities. An original feature of the software structure is the use of a calculator of derivatives of different orders to feed them to the input of the neural network, which helps to expand its receptive field and increases the accuracy of its results. Testing of the proposed structure of the software model was carried out in the MatLab-Simulink environment. The results of the simulation experiment showed that, unlike the PID controller, the use of a deep neural network as a regulator allows for successful compensation of the influence of external factors on the quality of control.

The work is devoted to the creation of a hybrid finite element model of the process of additive synthesis of a product by the Wire Arc Additive Manufacturing/3D Metal Print method, supplemented by a block of wave strain hardening of the synthesized product. The main result obtained is the proposed original approach to the mathematical modeling of fundamentally new processes of synthesis and hardening of a wide range of products manufactured from various materials in the form of wire, which can be extended to the entire class of additive synthesis processes with periodic or post-processing. The main feature of the approach is the hybridization of calculation models, while, unlike known solutions, significant attention is paid to the processes of exchange and transmission of information between the models of synthesis and hardening processes. A distinctive feature of the hybrid model is the need for a periodic exchange of information on the geometric and temperature parameters of the synthesized areas at a specific point in time. The novelty of the hybrid model is in the application of coordinated approaches to solving thermal and mechanical problems in a three-dimensional formulation in the ANSYS environment, taking into account the natural dynamic heat flows that form in the synthesized product and the installation table modeled by the finite element method. Calculation of continuously changing dynamic temperature fields in the synthesized product with additional optimization and visualization capabilities is an important structural part of the model. The studies performed allowed us to identify a significant range of the ratio of the product and table volumes 30 ≤ Vс/Vи ≤ 100 and patterns in the formation of temperature fields when changing the table parameters.